Breast microcalcifications are a common mammographic finding. Microcalcifications are considered suspicious signs of breast cancer and a breast biopsy is required, however, cancer is diagnosed in only a few patients. Reducing unnecessary biopsies and rapid characterization of breast microcalcifications are unmet clinical needs. In this study, 473 microcalcifications detected on breast biopsy specimens from 56 patients were characterized entirely by Raman mapping and confirmed by X-ray scattering. Microcalcifications from malignant samples were generally more homogeneous, more crystalline, and characterized by a less substituted crystal lattice compared with benign samples. There were significant differences in Raman features corresponding to the phosphate and carbonate bands between the benign and malignant groups. In addition to the heterogeneous composition, the presence of whitlockite specifically emerged as marker of benignity in benign microcalcifications. The whole Raman signature of each microcalcification was then used to build a classification model that distinguishes microcalcifications according to their overall biochemical composition. After validation, microcalcifications found in benign and malignant samples were correctly recognized with 93.5% sensitivity and 80.6% specificity. Finally, microcalcifications identified in malignant biopsies, but located outside the lesion, reported malignant features in 65% of in situ and 98% of invasive cancer cases, respectively, suggesting that the local microenvironment influences microcalcification features. This study confirms that the composition and structural features of microcalcifications correlate with breast pathology and indicates new diagnostic potentialities based on microcalcifications assessment.

Significance:

Raman spectroscopy could be a quick and accurate diagnostic tool to precisely characterize and distinguish benign from malignant breast microcalcifications detected on mammography.

The clinical presentation of breast cancer has profoundly changed in recent years, thanks to the wide adoption of screening mammography. Therefore, a shift toward early-breast cancer has been observed. In this context, breast microcalcifications are a common finding on mammography and they even increased following the evolution of digital imaging techniques, but only a small proportion of them reveals a breast cancer (1, 2). This results in a positive predictive value (PPV) for biopsy with microcalcifications ranging between 20% and 40% (3) and it contributes to the overall high-false positive rates of mammography, accounting for $2.8–4 billion per year in the United States only (4, 5). Nowadays, the description of microcalcifications in clinics is only based on morphologic features in mammograms or by their general appearance on stained histologic samples (6, 7). Microcalcifications are classified as Type I and Type II. Type I microcalcifications (calcium oxalate, CaO) is an uncommon type of mineralized component, mainly associated with benign lesions; Type II microcalcifications, composed by calcium phosphate [mainly by hydroxyapatite (HA)], are the most frequently observed, both in benign and malignant samples (6, 8). Several computational efforts have been attempted to improve the assessment of microcalcifications during mammography using deep learning and other computer-aided tools (9), but these are not able to discriminate their molecular composition. Only recently, a contrast-phase X-ray mammography approach reported the possibility to study microcalcifications composition and to distinguish Type I and Type II microcalcifications (10), but what is missing is the assessment of Type II microcalcifications malignancy.

Raman spectroscopy (RS) is a nondestructive approach able to investigate the biochemical composition of biological samples thanks to the simple detection of vibrational properties of molecules by light scattering, without using labels or dyes (11). Moreover, RS is particularly adequate to study mineral components (12) and it is compatible with real-time in vivo diagnostic evaluations (13–16). Haka and colleagues reported for the first time that RS is able to recognize biochemical differences between benign and malignant microcalcifications (74 and 16 calcifications, respectively) by carrying out single acquisitions at selected sites inside the calcified lesion (8). Subsequently, Raman-based tools were proposed to study microcalcifications in tissue. Stone and colleagues reported both space-offset RS (SORS) and transmission-RS approaches potentially able to study microcalcifications without penetrating the tissue, and tests were performed on phantom samples (17, 18). Saha and colleagues developed a RS-based optical fiber probe potentially able to investigate microcalcifications penetrating the tissue in a semi-invasive approach (19, 20). The optical probe was tested on ex vivo tissue samples confirming the potentiality of RS-based approaches to assess microcalcifications. In parallel to RS, Baker and colleagues used infrared spectroscopy on histologic samples reporting a decrease of carbonate content of HA passing from benign to invasive samples (21). Even considering the mentioned literature, the current knowledge about microcalcifications composition is still limited. No previous Raman-based studies investigated the detailed composition of microcalcifications by studying the mineral composition over their entire surface; previous RS-based studied considered a relatively low number of calcified regions and the investigation of microcalcifications associated with invasive carcinoma was scarce or absent. Here, we report the currently most extended biochemical characterization of breast microcalcifications, thanks to the use of a Raman mapping approach on core biopsies. The aim of this study is to clarify the association between microcalcifications composition and the pathologic features of breast lesions and the surrounding tissue, validating some specific evidences by X-ray scattering approaches. Moreover, microcalcifications surrounding malignant lesions were investigated to understand the influence of nearby cancerous tissue on their biochemical features.

Samples

Fifty-six patients affected by suspicious breast microcalcifications on screening mammography, undergoing a core biopsy and treated at the Breast Unit of ICS Maugeri (Pavia, Italy) from 2018 to 2019, were included. Patients with mass-like lesions or previous breast surgery were excluded from the study. All patients signed a written informed consent before inclusion in the study, which was authorized by the Ethical Committee of the Institution (protocol 2281/2018 CE), which approved the study in compliance with the Declaration of Helsinki. In details, according to the B-categories defined by the UK National Health Service Breast Screening Program (NHSBSP; ref. 22), 6 patients reported normal tissue or minimal changes (B1), 9 patients reported benign lesions (B2), 8 patients reported lesions of uncertain malignancy (B3), 17 reported in situ carcinoma (B5a), and 16 reported invasive carcinoma (B5b; Table 1). Detailed characteristics of patients are reported in Supplementary Table S1.

Table 1.

Number of microcalcifications detected from different subjects.

B categoryHistological classificationPatientsTotal MCMC found outside the lesionRepresentative MC
B1 “normal tissue” NOR 65 — 65 
Total  6 65  65 
B2 “benign” FAD 30  30 
 FIB  
 FNE 11  11 
 UDH 18  18 
Total  9 67  67 
B3 “of uncertain malignancy” PAP  
 FEA (DIN1A) 32  32 
 ADH (DIN1B) 27  27 
Total  8 61  61 
B5a “carcinoma in situ” DCIS DIN1C 19 11 
 DCIS DIN2 32 8 24 
 DCIS DIN3 46 19 27 
Total  17 97 38 59 
B5b “invasive carcinoma” IDC G2 12 133 92 41 
 IDC G3 2 
 ILC G2 8 
 ILC G3 33 19 14 
 IMC 0 
Total  16 184 121 63 
Grand Total  56 474 159 315 
B categoryHistological classificationPatientsTotal MCMC found outside the lesionRepresentative MC
B1 “normal tissue” NOR 65 — 65 
Total  6 65  65 
B2 “benign” FAD 30  30 
 FIB  
 FNE 11  11 
 UDH 18  18 
Total  9 67  67 
B3 “of uncertain malignancy” PAP  
 FEA (DIN1A) 32  32 
 ADH (DIN1B) 27  27 
Total  8 61  61 
B5a “carcinoma in situ” DCIS DIN1C 19 11 
 DCIS DIN2 32 8 24 
 DCIS DIN3 46 19 27 
Total  17 97 38 59 
B5b “invasive carcinoma” IDC G2 12 133 92 41 
 IDC G3 2 
 ILC G2 8 
 ILC G3 33 19 14 
 IMC 0 
Total  16 184 121 63 
Grand Total  56 474 159 315 

Note: Samples were classified according to the B categories defined by the UK NHSBSP (22) and according to histologic classification. “Microcalcifications outside the lesion” were microcalcifications detected in the tissue surrounding B5a or B5b lesions and categorized with a lower diagnostic malignancy. “Representative microcalcifications” were microcalcifications detected inside the malignant lesion specified by the diagnostic category (B5a or B5b). Bold terms indicate total numbers for each category.

Abbreviations: DCIS, ductal carcinoma in situ; FNE, fat necrosis; ILC, invasive lobular carcinoma; IMC, invasive mucinous carcinoma; MC, microcalcification,NOR, normal tissue; PAP, papillary lesion.

Tissue preparation

Tissue slices were generated from formalin-fixed, paraffin-embedded tissue blocks. For each patient, a 10-μm slice was microtomed and mounted on mirrored stainless steel slides specific for Raman measurements (Renishaw plc). In parallel, for each tissue sample, a contiguous 6-μm tissue slice was colored using standard hematoxylin and eosin stain for standard diagnostic evaluation. To correctly assess the concordance of histopathology with specific Raman features, microcalcifications located inside in situ or invasive carcinoma were primarily considered in B5a and B5b samples, respectively, and defined among “representative” microcalcifications. Microcalcifications detected in samples reporting cancer but locally surrounded by tissue that could be categorized with a lower B-category (i.e., noncancerous tissue in B5a samples and both noncancerous tissue or in situ carcinoma in B5b samples) were analyzed separately. The 10-μm slices selected for Raman analyses were then deparaffinized using a simple protocol optimized starting from a previously reported method (23). Briefly, tissue slices mounted onto mirrored steel slides were dewaxed by two baths of hexane 95% (Merck KGaA), two baths of ethanol absolute (Merck KGaA), and a final bath of ethanol 95%, repeating these steps three times, followed by air drying for 2 hours.

Raman spectroscopy

The experimental and data analysis workflow is illustrated in Fig. 1. A commercial confocal Raman Microscope (InVia Reflex, Renishaw plc) was used to perform Raman mapping acquisitions. A 785 nm laser source with round shape spot, powered with around 90 mW, was coupled with a N-Plan 100x (NA 0.75, WD 0.37) Leica objective and with a 1,200 l/mm grating, centered around 1,250 cm−1 thus obtaining a spectral range between 700 and 1,760 cm−1. The final power was then filtered to reach around 20 mW on the sample. The detector is a CCD (1,024 × 256 pixels) sensitive between 400 and 1,060 nm and cooled at −70°C. The instrument was daily aligned and intensity calibrated using automated procedures implemented in the instrument start-up process. The wavelength shift calibration was periodically performed using multiple standards (polystyrene, paracetamol, and silica) and daily checked by an automatic procedure using the silica band at 520 cm−1.

Figure 1.

Experimental workflow and components. Overview of the experimental workflow (left) and representative Raman spectra of components detected in each Raman map and considered in this study (right). The Raman map reported as example was obtained from the analysis of a B5a sample and reports the presence of only HA (green), surrounded by tissue (gray). FFPE, formalin-fixed, paraffin-embedded; H&E, hematoxylin and eosin.

Figure 1.

Experimental workflow and components. Overview of the experimental workflow (left) and representative Raman spectra of components detected in each Raman map and considered in this study (right). The Raman map reported as example was obtained from the analysis of a B5a sample and reports the presence of only HA (green), surrounded by tissue (gray). FFPE, formalin-fixed, paraffin-embedded; H&E, hematoxylin and eosin.

Close modal

Raman mapping measurements of each microcalcification identified by the pathologist on the contiguous hematoxylin and eosin slice were performed. All microcalcifications with diameter between 15 and 1,200 μm were selected. A squared region was centered on each microcalcifications defining a step size (mapping size) between 1 and 15 μm (on average 6 μm), depending on microcalcifications size. The focus was preset for each microcalcification in the middle of the calcified part. For each step the acquisition time was of 3 seconds with a single repetition.

Data processing of Raman data

Data analyses were performed using the commercial software WiRe (Renishaw plc) or MATLAB (MathWorks) or OriginPro2019 (Originlab Corporation). The data analysis included (i) preprocessing; (ii) spectral classification, and (iii) statistical/multivariate analysis.

Data preprocessing started with cosmic rays removal performed by WiRe (nearest neighbor and width of features algorithms). Next step was the background correction used to reduce fluorescence signals that may compromise the study of Raman spectra. Background signal was removed in MATLAB using baseline correction by fitting and subtracting a polynomial function of the 10th order to each spectrum. Then, data were normalized in MATLAB by unit vector function to neutralize spectra intensity differences. The normalized data were then filtered with a moving average filter with a window size equal to five to reduce spectral noise.

Spectral classification was carried out to assign specific biochemical features (i.e., to identify specific components) to each map data point (i.e., for each single spectrum acquired). For each spectrum, three different indices (i.e., monotonicity of the spectra, correlation with reference spectra, and specific peaks positions) were evaluated and used to perform spectral classification (24). Spectra described by a monotonic function are those for which the functions preserve the given order. Concerning Raman spectra, it is known that the fluorescence background is a monotonically decreasing (or increasing) function (25). Monotonicity was here mainly used to support the identification of spectra with high fluorescence (typical of necrotic regions, see below) and for the identification of spectra related to optical support (not monotonic functions). The reference spectra of inorganic components [HA, whitlockite (WIT), amorphous calcium carbonate (aCaCa), crystalline calcium carbonate (calcite), CaO] are reported in Fig. 1. Reference spectra were produced by collecting Raman signals from real samples and by confirming their nature by X-ray scattering analysis and/or by data reported in literature. In details, at least 25 spectra from at least two samples were recorded and averaged to produce each reference spectrum. For calcium phosphate (both HA and WIT), after confirming their nature by data in literature (8, 26), 50 μm contiguous slices, from the same samples, were studied by X-ray analysis that finally confirmed their nature (Supplementary Figs. S1 and S2). For CaO, calcite, and aCaCa reference spectra were confirmed thanks to the available data from literature [from refs. 8 and 27 for CaO; from refs. 28 and 29 (RRUFF id X050034) for calcite; and from refs. 29 and 30 for aCaCa). For these three cases, the scarce number and the small size of microcalcifications presenting these components did not permit to perform accurate X-ray scattering analysis. Together with mineralized components, also references of tissue surrounding microcalcifications, necrotic regions (presenting typical fluorescence spectra) and stainless steel microscope slide were prepared and used to isolate pure signals from the calcified area only. Peak positions, used as third criterion, were extracted, when possible, from above described reference spectra. The three mentioned indices were fused with the majority vote system to produce a thematic (false color) map, which contains a classification based only on spectral information. Spatial classification was taken into account adopting the K-means clustering algorithm, which is specific to detecting the borders (31). The algorithm is based on a random initialization of the groups; different executions of this algorithm produce different clusters, depending from the initial centroids. Therefore, to produce reproducible results, the clustering is repeated 10 times with different initializations. The results of 10 clustering phases are then fused together to obtain a classification false-color map. The classification map produced by the K-means clustering has not a biochemical meaning, because K-means is a nonsupervised classification method. The biochemical information was obtained by fusing the K-means clustering with the spectral classification map produced before. For each K-means cluster, the correspondent area was classified by the spectral classification and averaged. As result, for each map dataset (for each microcalcification), each map point was assigned to a specific preselected component and, for each component, an average Raman signal was produced and extracted.

Statistical analysis and multivariate analysis were then conducted on the calcified components only. If not specified, variables were reported as means (± SDs) or median with range of values or as absolute numbers and percentages. Continuous variables were compared using nonparametric Wilcoxon–Mann–Whitney/Kruskal–Wallis test for variables with nonnormal distribution. Statistical significance level was set at P < 0.05 (two-tailed).

For multivariate analysis, for each microcalcification, a single Raman spectrum, calculated by averaging only signals from Type II microcalcification (i.e., HA and WIT components) was than considered. Principal component analysis (PCA) was performed obtaining 380 principal components (PC). A linear discriminant analysis (LDA) classification model was built using the first 14 PCs as training data, using pure benign (i.e., B1 and B2 microcalcifications) and pure malignant (i.e., representative B5a and B5b microcalcifications) and setting prior probabilities proportional to the group size. The number of PCs to be utilized was selected to represent about 90% of dataset variability and to exclude PCs associated to noise and/or small artefacts. The PCA–LDA classification model was validated by the leave-one-out cross-validation. B3 microcalcifications (uncertain malignancy), B5a microcalcifications detected outside in situ carcinoma, or B5a microcalcifications detected outside invasive carcinoma were used as test data. ROC curve, with relative AUC, was automatically calculated by OriginLab using as input the canonical variable 1 emerging from the PCA–LDA classification. From the ROC curve the optimal threshold (cut-off point) was obtained. This was used to produce confusion matrices and relative diagnostic performances.

Small- and wide-angle X-ray scattering and X-ray radiography

The X-ray transmission microscopy (radiography, XTM) maps more (darker) or less (lighter) X-ray absorbing portions of the biopsy; small-angle X-ray scattering (SAXS) microscopies display the abundance and D-spacing of collagen; wide-angle X-ray scattering (WAXS) microscopies map the abundance and lattice parameters of the HA structure of microcalcifications. To confirm the crystal phase composition of microcalcification, 50-μm slices, contiguous to samples used for RS, were deparaffinized using the protocol reported above and analyzed using XTM, SAXS, and WAXS microscopy at the cSAXS beamline of the Swiss Light Source (32). The monochromatic X-ray beam had a wavelength of 0.091216 nm corresponding to an energy of 13.6 keV. It was focused down to about 20 μm (vertical) and 40 μm (horizontal) by a bent monochromator crystal and a bent mirror. Sample to detector distances were set to 308 mm (WAXS) and 7,089 mm (SAXS). Collection time per point was 0.3 seconds (WAXS and SAXS). A Pilatus 2 M detector (33) was used for recording SAXS/WAXS; the intensity of the transmitted direct beam was measured by means of a point detector in the beam stop, to obtain the XTM. The samples were raster-scanned through the focused X-ray beam in a continuous line-scan mode, that is, with the sample moving at constant velocity while the detectors record data at constant frequency. The resulting spatial distance between data points of the raster scan is 40 μm. Depending on the investigated area, data have been recorded from 6′731 to 61′875 positions on the sample in each raster scan.

Data processing—SAXS and WAXS

The SAXS and WAXS data have been azimuthally integrated within 16 angular sectors. The resulting integrated data were screened through a statistical signal classification approach (34) to reduce the large number of WAXS and SAXS data to few characteristic profiles. The characteristic WAXS profiles were fitted by a whole-profile Rietveld approach (35) implemented in the FullProf program (36). The crystallographic unit cell parameters (a and c), the unit cell volume (V), and the crystalline domain size along several crystallographic directions were determined, gaining the trend of these parameters for increasing malignancy. For validating the representativeness of the characteristic WAXS profiles selected using signal classification, we explored also the entire WAXS dataset across all investigated sample areas. For each pixel of the WAXS microscopy, the peak position of the [002] reflection as an isolated single diffraction peak was fitted with a Gaussian to extract peak position, amplitude, and width, and in this way their variations have been mapped across the whole area. Then, we evaluated the mean and dispersion of: the c parameter, by means of the Bragg law applied to the peak position; the crystalline domain size along the [002] crystallographic direction, extracted with the Scherrer formula from the peak width. The statistical evaluation of c value and domain size along the [002] direction mapped across the whole area, were found to be nicely in agreement with the same values extracted from the Rietveld fitting of the characteristic WAXS profiles determined in a signal classification approach. Both the c value and the crystalline domain size along the [002] direction were found to increase with the malignancy level.

In case of the SAXS data, the characteristic profiles have been used to define regions of interest for an analysis of characteristic peaks in the scattering signal at each point of each raster scan (37). In particular, two main peaks were identified and interpreted as the 6th order of collagen fibers with 63.0 nm D-spacing and as the 7th order of collagen fibers with 61.8 nm D-spacing. We monitored their variation across the investigated area and related them to WAXS and XTM microscopies.

A total of 474 microcalcifications from 56 patients were mapped by Raman spectroscopy as described above. Of these, 65 microcalcifications were detected in normal tissue (B1), 67 in samples with benign lesion (B2), 61 in samples with uncertain malignant features (B3), 97 in samples with in situ carcinoma (B5a), and 184 in samples with invasive carcinoma (B5b; Table 1; Supplementary Table S1). Only 59 and 63 microcalcifications from B5a and B5b, respectively, were observed inside the lesion and were primarily considered as representative microcalcifications. The remaining were locally surrounded by tissue categorized within a lower diagnostic category and studied separately.

Benign microcalcifications contain specific components and are more heterogeneous than malignant ones

A Raman mapping approach was optimized to characterize and automatically identify all inorganic components contained in microcalcification at micrometric scale (Fig. 1). Of 315 representative microcalcifications, 273 (86%) contains HA, the most common form of calcium phosphate defining Type II microcalcifications (Fig. 2A). Overall, HA also represents the most abundant component (>74%) of microcalcifications (Fig. 2B). Considering different diagnostic categories, benign calcifications (B1 and B2) are more heterogeneous than uncertain malignant (B3) or malignant microcalcifications (B5a and B5b). In particular, 33 and 40 microcalcifications found in B1 and B2 samples, respectively, reported spectral features of WIT (Fig. 2A and B; ref. 38), a particular magnesium-containing crystal phase of calcium phosphate, as also confirmed by WAXS performed on selected tissue specimens (Supplementary Fig. S2). In RS data, WIT was not found as unique inorganic component but always colocalized with HA and it constitutes, overall, 2.7% and 6.2% of B1 and B2 microcalcifications, respectively. Some benign microcalcifications (i.e., 7 and 25 microcalcifications in B1 and B2 lesion, respectively) also showed the presence of a few signals (<2%) of aCaCa. In addition, among benign microcalcifications, only 14 microcalcifications from B1 samples contain CaO as sole component (near 100%) and these calcifications correspond to Type I microcalcifications.

Figure 2.

Detailed composition of representative microcalcifications (n = 315). A, Number of microcalcifications exhibiting at least one pixel of the components described by the legend, for each diagnostic category. B, Overall composition of microcalcifications, calculated, considering altogether all microcalcifications belonging to each diagnostic category. Components: HA, calcium phosphate in the hydroxyapatite form; aCaCa, calcium phosphate in the WIT; CaO, calcite; MC, microcalcifications not showing specific Raman spectra associated to mineralized components (none).

Figure 2.

Detailed composition of representative microcalcifications (n = 315). A, Number of microcalcifications exhibiting at least one pixel of the components described by the legend, for each diagnostic category. B, Overall composition of microcalcifications, calculated, considering altogether all microcalcifications belonging to each diagnostic category. Components: HA, calcium phosphate in the hydroxyapatite form; aCaCa, calcium phosphate in the WIT; CaO, calcite; MC, microcalcifications not showing specific Raman spectra associated to mineralized components (none).

Close modal

Conversely, malignant (B5a and B5b) and B3 lesions, generally exhibit homogeneous microcalcifications, containing almost only HA, representing ≥97% of their overall composition (Fig. 2A and B). A single B3 sample microcalcification presented Type I microcalcification (n = 3) containing pure CaO.

Among the whole dataset, a few microcalcifications from both benign and malignant samples reported a highly crystalline form of calcium carbonate (calcite), never reported before in breast microcalcifications and apparently not associated with pathology from our data. Finally, 17 microcalcifications identified by the pathologists did not show any specific signal of mineralized components and were likely amorphous material attributable to necrotic regions.

The chemical features of microcalcifications are associated with the diagnostic classification

Type II microcalcifications were mainly investigated here considering that Type I microcalcifications were uncommon presentations, rarely observed in our samples. The mean Raman spectra of Type II microcalcifications from different diagnostic categories revealed some differences around the major signals; in particular, around the 960 cm−1 band, related to calcium phosphate vibrational modes, around the 1,070–1,090 cm−1 band, related to calcium carbonate content, and around the 1,450 cm−1 band, mainly related to protein and lipids (here referred as “organic matrix”) content (Fig. 3AF; Supplementary Fig. S3). In details, broadening of the phosphate band was observed in benign samples (Fig. 3A and D) and this was mainly associated with a more disordered and more substituted phosphate crystal lattice (39, 40). The broadening extent was similar in normal (B1) and benign (B2) microcalcifications (P = 0.43) but significantly higher (P = 7.90 × 10−10) if compared with microcalcifications found in lesion of uncertain malignancy (B3), in situ (B5a), and invasive lesions (B5b). In turn, B3, B5a, and B5b exhibit only minor variation of broadening values (P = 0.032), with main contribution by differences between B3 and B5b samples (P = 0.011; Supplementary Table S2). In parallel, observing the phosphate band, a statistically significant shift toward higher wavenumbers (here referred as “red-shift”) was also observed in benign microcalcifications (P = 0.025). This is mainly derived by the presence of WIT in benign microcalcifications, characterized by maximum intensity of the phosphate band, around 970 cm−1 (Fig. 3, Supplementary Fig. S4; Supplementary Table S3; ref. 38).

Figure 3.

Vibrational features of representative Type II microcalcifications (n = 264). A–C, Average phosphate band (A), average carbonate band (B), and average organic matrix band (C) of each of the diagnostic categories, including SD (shaded area). In A, the intensity of spectra is shifted for clarity. D, Box plot reporting the broadening of the phosphate band using the full width at half maximum (FWHM) of the band. E, Box plot reporting the intensity (area) of the carbonate band, between 1,070–1,090 cm−1. F, Box plot reporting the intensity (area) of organic matrix band, between 1,425–1,475 cm−1. Data are shown as box and whiskers. Each data point represents a single microcalcification. Each box represents the 25th to 75th percentiles [interquartile range (IQR)]. Dots inside the box are the mean; lines inside the boxes represent the median. The whiskers represent the lowest and highest values within the boxes ± 1.5× the IQR. All P values are reported in Supplementary Table S2. n.s., no significant difference was observed (P > 0.05; two tailed).

Figure 3.

Vibrational features of representative Type II microcalcifications (n = 264). A–C, Average phosphate band (A), average carbonate band (B), and average organic matrix band (C) of each of the diagnostic categories, including SD (shaded area). In A, the intensity of spectra is shifted for clarity. D, Box plot reporting the broadening of the phosphate band using the full width at half maximum (FWHM) of the band. E, Box plot reporting the intensity (area) of the carbonate band, between 1,070–1,090 cm−1. F, Box plot reporting the intensity (area) of organic matrix band, between 1,425–1,475 cm−1. Data are shown as box and whiskers. Each data point represents a single microcalcification. Each box represents the 25th to 75th percentiles [interquartile range (IQR)]. Dots inside the box are the mean; lines inside the boxes represent the median. The whiskers represent the lowest and highest values within the boxes ± 1.5× the IQR. All P values are reported in Supplementary Table S2. n.s., no significant difference was observed (P > 0.05; two tailed).

Close modal

Also, the carbonate band reveals higher intensity and broadening in benign samples in the spectral region between 1,070–1,090 cm−1, showing significant differences when comparing benign (B1 and B2) and nonbenign (B3, B5a, and B5b) subtypes (P = 2.86 × 10−9; Fig. 3B and E). No significant differences can be observed when comparing B1 and B2 (P = 0.51) and a certain variance characterize B3, B5a, and B5b (P = 0.001), especially due to differences between B5a and B5b (Fig. 3B and E). Finally, the intensity of the 1,450 cm−1 band, corresponding to the organic matrix, also changes among diagnostic categories, but differences do not correlate with the pathologic status (Fig. 3C and F).

The crystalline lattice of microcalcifications is more ordered and expands with increasing malignancy

To better understand some biochemical and structural features emerging from Raman data, crystallographic evaluations by WAXS were performed on randomly selected samples, focusing on Type II microcalcifications, and on HA structures. Figure 4A shows the HA crystalline domain parameter and the HA crystallographic unit cell “c” parameter increasing when passing from benign (B1 and B2) to carcinoma samples (B5a and B5b; overall diffractograms in Supplementary Fig. S1). This behavior was coherent with the broadening of phosphate Raman peak (Fig. 4A and B) passing from malignant to benign samples. Both these data indicate that the crystallinity and homogeneity of HA produced by malignant microcalcification was significantly higher if compared with benign samples. In particular, the increasing of the “c” parameter indicates that the unit cell of HA in microcalcifications in malignant samples was generally elongated in comparison with the one in benign samples.

Figure 4.

Crystalline feature of HA from selected representative samples. A, Variation of the crystalline domain size along the [002] direction. B, Variation of the crystallographic unit cell parameter c along the [002] direction. Filled circles, results of Rietveld fits of the few patterns extracted from the entire dataset by a statistical signal classification approach (32); empty circles, the main value resulting from a single-peak fit of the [002] reflection at 2θ = 15.29 deg for all the patterns collected in the entire explored sample area of a sample (the analysis reported in Supplementary Fig. S2 is not included in these trends as the sample was a mixture of two crystalline phases, i.e., HA and WIT).

Figure 4.

Crystalline feature of HA from selected representative samples. A, Variation of the crystalline domain size along the [002] direction. B, Variation of the crystallographic unit cell parameter c along the [002] direction. Filled circles, results of Rietveld fits of the few patterns extracted from the entire dataset by a statistical signal classification approach (32); empty circles, the main value resulting from a single-peak fit of the [002] reflection at 2θ = 15.29 deg for all the patterns collected in the entire explored sample area of a sample (the analysis reported in Supplementary Fig. S2 is not included in these trends as the sample was a mixture of two crystalline phases, i.e., HA and WIT).

Close modal

Finally, SAXS data show that collagen was colocated with microcalcifications and that it has longer intermolecular D-spacing of 63.0 nm in comparison with 61.8 nm of collagen found outside the microcalcification (Supplementary Figs. S5 and S6). These findings reveal that HA nanocrystals were spatially correlated to the larger D-spacing of the collagen fibers, which can be interpreted as a hint toward crystallization nucleating in collagen fibers. A weak tendency toward an overall increase of the intramolecular spacing of collagen was found passing from benign to malignant microcalcifications (Supplementary Fig. S6) and this was interpreted speculatively as a tighter binding of collagen to malignant microcalcifications.

The multivariate analysis of microcalcification Raman maps allows for the accurate classification of different histologic subtypes

Multivariate analysis was applied to extract all potential pathologic biomarkers from the complexity of Raman data thus improving the limited and subjective analysis of single spectrometric variables. First, PCA was performed and the first 14 PCs (Supplementary Fig. S7), representing 90% of the variability of the entire dataset, were extracted and used as variables for further classification. A classification model was made by LDA first using only representative microcalcifications from pure benign samples (i.e., B1 and B2) and from pure malignant samples (B5a and B5b), thus producing a Raman canonical correlation coefficient (canonical variable 1) further used as unique classification variable (Fig. 5AC). After excluding a correlation between age and the biochemical composition of microcalcifications (R2 = 0.020, slope P = 0.33; Supplementary Fig. S8), a ROC curve (Fig. 6) was calculated and used to find the optimal threshold point (i.e., −0.445 of canonical variable) for the classification. Of the 108 malignant microcalcifications, 103 were correctly classified giving 95.4% sensitivity and 94.6% negative predictive value (NPV; Supplementary Table S4). Among microcalcifications incorrectly assigned as benign, four were from carcinoma in situ and one from invasive carcinoma. Of the 103 benign microcalcifications, 15 were wrongly assigned as malignant giving 85.4% of specificity and 87.3% PPV. Of the benign microcalcifications wrongly classified as malignant, 10 were from normal tissue, one from fibroadenoma (FAD), two from fibrocystic change (FIB), and two from usual ductal hyperplasia (UDH). Overall, the model gives 90.5% accuracy. These results were validated through leave-one-out cross-validation, which virtually assigns all microcalcifications with an unknown diagnosis before to be classified. This gave 93.5% sensitivity and 80.6% specificity with 87.2% of overall accuracy (Supplementary Table S5).

Figure 5.

Results from the LDA classification model. A, Box plot showing the canonical variable score emerging from the LDA-based classification of microcalcifications from pure benign (B1 and B2) and pure malignant (B5a and B5b) samples (n = 211). B, Box plot reporting the canonical variable score of microcalcifications found in B3 samples and used as test dataset (n = 53). C, Box plot reporting the canonical variable score of microcalcifications found outside carcinoma lesions in B5a (n = 28) and B5b samples (n = 88), used as test dataset. Data are shown as box and whiskers plot. Each data point represents a single microcalcification. Each box represents the 25th to 75th percentiles (interquartile range, IQR). Dots inside the box are the mean; lines inside the boxes represent the median. The whiskers represent the lowest and highest values within the boxes ± 1.5× the IQR. The dashed lines refer to the optimal cut-off value (−0.445) defined by the ROC.

Figure 5.

Results from the LDA classification model. A, Box plot showing the canonical variable score emerging from the LDA-based classification of microcalcifications from pure benign (B1 and B2) and pure malignant (B5a and B5b) samples (n = 211). B, Box plot reporting the canonical variable score of microcalcifications found in B3 samples and used as test dataset (n = 53). C, Box plot reporting the canonical variable score of microcalcifications found outside carcinoma lesions in B5a (n = 28) and B5b samples (n = 88), used as test dataset. Data are shown as box and whiskers plot. Each data point represents a single microcalcification. Each box represents the 25th to 75th percentiles (interquartile range, IQR). Dots inside the box are the mean; lines inside the boxes represent the median. The whiskers represent the lowest and highest values within the boxes ± 1.5× the IQR. The dashed lines refer to the optimal cut-off value (−0.445) defined by the ROC.

Close modal
Figure 6.

ROC curve. The curve was obtained from the post probability assignment produced by the LDA-based classification model and considering only pure benign (B1 and B2) and pure malignant microcalcifications (B5a and B5b; n = 211).

Figure 6.

ROC curve. The curve was obtained from the post probability assignment produced by the LDA-based classification model and considering only pure benign (B1 and B2) and pure malignant microcalcifications (B5a and B5b; n = 211).

Close modal

The classification model was then used to investigate the pathological fingerprint of microcalcifications found in lesion of uncertain malignancy (B3). Of the 53 B3 microcalcifications, 44 [29 from flat epithelial atypia (FEA) tissue and 15 from atypical ductal hyperplasia (ADH)] were assigned as malignant and nine microcalcifications (seven ADH, once FEA, and one from papillary lesion) were recognized as having benign features (Fig. 5B; Supplementary Table S6). Moreover, also a correlation between the Raman data and mammographic evaluation according to the Breast Imaging-Reporting and Data System (BI-RADS; ref. 26) classification was observed (Supplementary Fig. S9). Finally, considering that microcalcification size can be very heterogeneous, as shown in Supplementary Fig. S10 [mean: 14,511 μm2; SD: 27,229 μm2; interquartile range (IQR) 25%–75%: 11,700 μm2], the correlation between microcalcification size and malignant features, as revealed by Raman mapping (i.e., canonical variable 1), has been investigated, showing a weak correlation [−0,1894; P = 0.0002 (Spearman correlation)], suggesting that larger microcalcifications are slightly associated with lower malignancy.

Most of microcalcifications found in locally healthy tissue neighboring carcinoma lesions show malignant features

A total of 116 microcalcifications (28 and 88 for B5a and B5b, respectively) were excluded from the main data analysis because they were locally surrounded by noncancerous tissue in B5a samples, or because surrounded by noncancerous tissue or in situ carcinoma in B5b samples. When data from B5a samples were used as test dataset, of the 28 microcalcifications found locally benign, 18 showed malignant features and were classified as malignant, the remaining 10 microcalcifications were classified as benign (Fig. 5C; Supplementary Table S6). Of the 88 microcalcifications from B5b detected outside the invasive carcinoma, only two microcalcifications (locally B1) were classified as benign but the remaining 80 were classified as malignant (Fig. 5C; Supplementary Table S6). These include 29 microcalcifications situated in in situ carcinoma (B5a), 23 microcalcifications surrounded by B3 tissue, but also 34 microcalcifications surrounded by B1 tissue.

The rapid and accurate characterization of breast microcalcifications is currently an unmet clinical need. Microcalcifications are considered as suspicious signs of a breast lesion, and histopathology is mandatory to assess the nature of such lesions. Up to now, microcalcifications have been traditionally seen by mammography as simple bystanders of cancer and only characterized by descriptive criteria (i.e., morphology and spatial distribution), according to BI-RADS (41). Recently, research teams have begun to show the potential of further spectral analysis using Raman and IR spectroscopy (8, 17, 21, 40, 42), but this article advances the concept significantly.

Here, we investigated breast microcalcifications from different diagnostic categories by studying the whole area of each microcalcification using a Raman mapping approach on a relevant number of microcalcifications. This allowed for the first time to describe all inorganic components contained in microcalcifications and to correlate it with the pathologic status. Previously reported Raman characterizations have been made on a significantly lower number of microcalcifications and by performing single acquisitions at selected sites inside the calcified lesion (8) or single spectra on fresh biopsies (40), thus obtaining partial information. Only recently, the spatial composition of microcalcifications has been investigated by Raman mapping on the whole-microcalcifications surface, although on a few specimens, with explorative aims (26).

The first evidence emerging from this study is that microcalcifications from pure benign lesions (i.e., B1 and B2) are largely heterogeneous if compared with both lesions of uncertain malignant potential (B3) and with carcinoma lesions (B5a and B5b). First, in benign samples both Type I (CaO) and Type II (HA) microcalcifications were observed, confirming what previously reported; also confirming that Type I microcalcifications are uncommon findings (8). Second, in most of Type II microcalcifications identified in benign samples, HA is not the only component but also WIT and aCaCa were observed, (in 55% and 24% of benign microcalcifications, respectively). WIT is a crystal phase of calcium phosphate, sporadically detected in human tissues where magnesium partially substitutes calcium if compared with HA (43). In this study, investigating 56 sample and 315 representative microcalcifications, WIT overall represents 4.5% of benign calcifications (B1 and B2), 0.32% of B3 samples, and the 0.22% of cancerous calcification (B5a and B5b). These results were further confirmed when the mean spectrum of each microcalcifications was considered. Recently, WIT was identified by X-ray in a previous study reporting more WIT in malignant (1.18%) then in benign (0.46%) samples (44). This discrepancy could be explained by a lower number of samples (57 microcalcifications from 15 biopsies) included in the mentioned study and by the lower spatial resolution of X-ray measures (10 μm), if compared with an average mapping size of 6 μm used here for Raman mapping.

In parallel, the finding of both amorphous (aCaCa) and calcite as isolated components in microcalcification is a new evidence. aCaCa was mostly found in benign samples, always localized with WIT, and this could be due to the fact that magnesium, contained in WIT, is known to stabilize aCaCa (45). Calcite is the stable and crystalline form of CaCa that can naturally form if not stabilized by active processes (46) and has been find in both benign and malignant samples only as single very small crystals of pure calcite. The amount of data is not sufficient to explain an association with pathology.

When the overall Raman signals originating from single Type II microcalcifications were extracted to verify their correlation with diagnosis, the first evident feature was the broadening and red-shift of the 960 cm−1 phosphate band of calcium phosphate in B1 and B2 samples. The broadening of the phosphate band is normally explained by the increase of carbonate content into HA, resulting in the alteration of the symmetry of the crystal structure (39). The red-shift of the phosphate band is due to the change of the crystal composition as in the case of WIT, due to the presence of magnesium (38). In our data, both contributions have been seen and this is explained by the copresence of (carbonated) apatite and of WIT in benign microcalcifications. Previous RS studies on real samples, only mentioned peak broadening and this could be due to single point acquisition, not able to recapitulate the whole microcalcification composition (8, 40). Noteworthy, when 960 cm−1 peak broadening (full width at half maximum) is compared among diagnostic categories, B3, B5a, and B5b samples cluster together with only minor differences (P = 0.032). In parallel, also the band of carbonate changes among diagnostic categories, decreasing in its overall intensity passing from benign to malignant samples. Also in this case, there is a possible contribution derived by the presence of WIT in benign samples, producing band broadening and a weak shoulder at higher Raman shift, and this was not revealed in previous studies describing carbonate as indicator of benignity (8, 21, 40). As a final point, organic matrix (i.e., proteins and lipids) has been described as microcalcification component but its contribution did not show any statistically significant correlation with pathology as reported previously (8, 40).

Spectral data were also confirmed by WAXS acquisitions on randomly selected samples, showing that malignant samples (B5a and B5b) exhibit an increase of crystallinity, in accordance with previous findings (44). In connection to the observation in RS that WIT is more abundant in benign than in malignant microcalcification, one could speculate that malignant microcalcification contains less Mg, thus exhibiting a more well-ordered HA crystalline lattice. In healthy conditions, biomineralization involves HA with partial substitution of Ca ions (47). Smaller cations, such as Mg ions, lead to a cell contraction compared with the stoichiometric crystal (48). Therefore, the observed trend of the “c” parameter can be correlated to an altered metabolism of Mg ions, which changes with the malignancy level. Indeed, neoplastic cells, that is, for B5 samples, are avid of Mg cations (49), which therefore might reduce the Mg quantity available for microcalcifications, justifying the lattice expansion, compared with what was expected in benign biomineralization. In addition, the observed trend of the “c” parameter, which increases passing from benign to malignant microcalcifications, may be also associated with carbonate substitution in HA structures (50). This is coherent with relatively higher levels of carbonate Raman signals in benign samples, reported in this and in previous studies (21, 40). In particular, lowering of the “c” parameter occurs when carbonate mainly substitutes the hydroxyl site of HA (called “A-Type substitution”) and this has also been observed in a very recent X-ray diffraction study on microcalcifications (51). Finally, SAXS data exhibit slightly larger D-spacing of collagen in microcalcifications found in malignant samples compared with benign ones, in agreement with SAXS data collected from tissue (52). This weak trend can be speculatively attributed to a tighter binding of microcalcification to collagen with increasing malignancy.

A general overview of results coming from the detailed characterization of breast microcalcifications suggests that malignant microcalcifications are more homogeneous, more crystalline, less substituted (by Mg and carbonate), and more tightly bound to collagen, if compared with benign ones. In addition, chemistry and structure of microcalcifications from lesion of uncertain malignancy (B3) are generally similar to pure malignant ones (B5a and B5b). Studies performed on tissue samples and in vitro on cultured cells demonstrated that microcalcification formation is a cell active process influenced by the microenvironment and by the overexpression of bone matrix proteins (i.e., osteonectin and osteopontin; refs. 53, 54). Both these studies reported that active processes of microcalcification formation are significantly more represented in case of malignancy. In parallel, a retrospective study on patients referred for needle-guided biopsy, reported that the formation of new microcalcifications significantly correlates with high probability of ductal invasive carcinoma (55). These data suggest that malignant lesions are more active in the formation of microcalcifications. In this regard, we have seen a weak correlation between microcalcification size and malignancy features, indicating that smaller microcalcifications are lightly associated with higher malignancy. This could be speculatively associated with the smaller size of new forming microcalcifications during carcinoma growth. Starting from these assumptions, the crystallinity and homogeneity of malignant microcalcifications could originate from a faster and active process stimulated by cancer and by its microenvironment. On the contrary, the heterogeneity and low crystallinity of benign microcalcifications could be explained by a slower and less regulated mechanism of mineralization, thus permitting both loss of crystal stability and/or the intercalation of external components (i.e., carbonate and magnesium) from the surrounding tissue.

To verify the accuracy of the presented Raman-based microcalcification characterization, a multivariate approach was applied, including the use of a LDA-based model to automatically verify the potential diagnostic performance of the proposed approach. The model was built using pure benign (B1 and B2) and pure malignant (B5a and B5b) categories; including only microcalcifications detected inside carcinoma region in case of malignant microcalcification. The results are promising, showing 93.5% sensitivity and 80.6% specificity with an NPV of 92.2% after cross-validation. The same classification model was used to investigate the grade of malignancy of B3, according to the biochemical composition, revealing that 39 (83%) of 53 B3 microcalcifications detected show malignant features, mostly by FEA subtype. These data are in agreement with the fact that FEA represents a direct precancerous lesion possibly leading to ductal carcinoma in situ in up to 18.6% of cases (56).

In addition, we investigated whether all microcalcifications detected in malignant biopsy samples, but outsides the specific cancer lesion, exhibit the features of the local “benign” (or B3 or in situ carcinoma) surrounding tissue, or the malignant features of the carcinoma nearby. As described above, 64% of locally benign (or B3) microcalcifications found around in situ carcinoma samples showed malignant features. In invasive carcinoma samples, 86 (98%) of 88 microcalcifications found outside the invasive region [including 34 (94%) of 36 locally classified as B1] were classified as malignant. This is interesting data, showing that the biochemical composition and structural features of microcalcifications are influenced by the tumor even if microcalcifications are not directly surrounded or in close contact with cancer cells. As a consequence, we can assume that the tumor environment, and probably the metabolism of breast tissue in the presence of malignancy (especially if invasive carcinoma), influence a relatively extended tissue region around the malignancy, as also recently suggested (57). This evidence may also help the translation of Raman-based optical probes or Raman-based noninvasive tools, potentially compatible with in vivo approaches (17–19). In particular, these data suggest that the investigation of a relatively large region (500–2,000 μm) of suspected tissue containing microcalcifications could correctly inform about the malignancy even if some locally benign microcalcification regions surrounding the lesion are inevitably probed because of low resolution of in vivo configuration.

In conclusion, this study reports new detailed information about microcalcification composition and demonstrates that Raman-based approaches can provide a direct and reliable description of breast lesions, thanks to the study of microcalcifications. These evidences may play an important role in the development of new tools for the assessment of suspected lesions presenting microcalcifications. For example, SORS and transmission Raman approaches are able to illuminate a circumscribed portion of sample at a certain depth in a noninvasive manner and they were already tested on phantom samples containing calcified components (17, 18). Similarly, Raman needle probes are able to perform accurate measurements into the tissue in a semi-invasive modality (15, 16) and a proof-of-concept was also tested to study breast microcalcifications ex vivo (19, 42). Moreover, preliminary experiments using phase-contrast X-ray imaging approaches demonstrated the possibility to study the crystalline structure of microcalcifications (10). The mentioned tools have not yet been tested or transferred in clinics also due to the lack of detailed information about microcalcification composition. The data reported here may be step forward to this direction.

No potential conflicts of interest were disclosed.

Conception and design: R. Vanna, C. Morasso, O. Bunk, C. Giannini, F. Corsi

Development of methodology: R. Vanna, B. Marcinnò, E. Torti, C. Giannini

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Vanna, F. Piccotti, D. Altamura, M. Agozzino, L. Villani, L. Sorrentino, O. Bunk, C. Giannini, F. Corsi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Vanna, C. Morasso, E. Torti, D. Altamura, S. Albasini, O. Bunk, F. Leporati, C. Giannini

Writing, review, and/or revision of the manuscript: R. Vanna, E. Torti, L. Sorrentino, O. Bunk, F. Leporati, C. Giannini, F. Corsi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Piccotti

Study supervision: R. Vanna, F. Corsi

Other (designed and optimized preprocessing and extraction Raman mapping data): B. Marcinnò, E. Torti, F. Leporati

Other (X-ray data collection and analysis): D. Altamura

Other (contributions to designing the X-ray SAXS/WAXS experiment, acquiring the X-ray data, analyzing and interpreting them, with a focus on the SAXS data, contributions to the article): O. Bunk

Other (contribution about X-ray data section): C. Giannini

The proposal “Correlative Imaging of scanning micro Raman and SAXS/WAXS Scanning Microscopies […]” (E181100214) and thereby access to the cSAXS beamline of the Swiss Light Source has been supported by the European Union's Horizon 2020 research and innovation program under grant agreement No. 731019 (EUSMI; to C.Giannini). We thank Andreas Menzel and Xavier Donath for providing assistance for the X‐ray measurements. We also thank Luciana Russo and Chiara Guerra for technical support in samples' preparation and Daniele Baldassarra for his help in collecting Raman data.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Gajdos
C
,
Ian Tartter
P
,
Bleiweiss
IJ
,
Hermann
G
,
de Csepel
J
,
Estabrook
A
, et al
Mammographic appearance of nonpalpable breast cancer reflects pathologic characteristics
.
Ann Surg
2002
;
235
:
246
51
.
2.
Lakhani
SR
,
Ellis
IO
,
Schnitt
SJ
,
Tan
PH
,
van de Vijver
MJ
.
WHO classification of tumours of the breast
.
Lyon, France
:
International Agency for Research on Cancer
; 
2012
.
3.
Wilkinson
L
,
Thomas
V
,
Sharma
N
. 
Microcalcification on mammography: approaches to interpretation and biopsy
.
Br J Radiol
2016
;
90
:
20160594
.
4.
Ong
M-S
,
Mandl
KD
. 
National expenditure for false-positive mammograms and breast cancer overdiagnoses estimated at $4 Billion a year
.
Health Aff
2015
;
34
:
576
83
.
5.
Vlahiotis
A
,
Griffin
B
,
Stavros
AT
,
Margolis
J
. 
Analysis of utilization patterns and associated costs of the breast imaging and diagnostic procedures after screening mammography
.
Clinicoecon Outcomes Res
2018
;
10
:
157
67
.
6.
Frappart
L
,
Remy
I
,
Lin
HC
,
Bremond
A
,
Raudrant
D
,
Grousson
B
, et al
Different types of microcalcifications observed in breast pathology
.
Vichows Archiv A Pathol Anat
1987
;
410
:
179
87
.
7.
Morgan
MP
,
Cooke
MM
,
McCarthy
GM
. 
Microcalcifications associated with breast cancer: an epiphenomenon or biologically significant feature of selected tumors?
J Mammary Gland Biol Neoplasia
2005
;
10
:
181
7
.
8.
Haka
AS
,
Shafer-Peltier
KE
,
Fitzmaurice
M
,
Crowe
J
,
Dasari
RR
,
Feld
MS
. 
Identifying microcalcifications in benign and malignant breast lesions by probing differences in their chemical composition using Raman spectroscopy
.
Cancer Res
2002
;
62
:
5375
80
.
9.
Yassin
NIR
,
Omran
S
,
El Houby
EMF
,
Allam
H
. 
Machine learning techniques for breast cancer computer aided diagnosis using different image modalities: a systematic review
.
Comput Methods Programs Biomed
2018
;
156
:
25
45
.
10.
Wang
Z
,
Hauser
N
,
Singer
G
,
Trippel
M
,
Kubik-Huch
RA
,
Schneider
CW
, et al
Non-invasive classification of microcalcifications with phase-contrast X-ray mammography
.
Nat Commun
2014
;
5
:
3797
.
11.
Ellis
DI
,
Cowcher
DP
,
Ashton
L
,
O'Hagan
S
,
Goodacre
R
. 
Illuminating disease and enlightening biomedicine: Raman spectroscopy as a diagnostic tool
.
Analyst
2013
;
138
:
3871
84
.
12.
Mandair
GS
,
Morris
MD
. 
Contributions of Raman spectroscopy to the understanding of bone strength
.
BoneKEy Rep
2015
;
4
:
620
.
13.
Jermyn
M
,
Mok
K
,
Mercier
J
,
Desroches
J
,
Pichette
J
,
Saint-Arnaud
K
, et al
Intraoperative brain cancer detection with Raman spectroscopy in humans
.
Sci Transl Med
2015
;
7
:
274ra19
.
14.
Matousek
P
,
Draper
ERC
,
Goodship
AE
,
Clark
IP
,
Ronayne
KL
,
Parker
AW
. 
Noninvasive Raman spectroscopy of human tissue in vivo
.
Appl Spectrosc
2006
;
60
:
758
63
.
15.
Day
JCC
,
Bennett
R
,
Smith
B
,
Kendall
C
,
Hutchings
J
,
Meaden
GM
, et al
A miniature confocal Raman probe for endoscopic use
.
Phys Med Biol
2009
;
54
:
7077
87
.
16.
Hattori
Y
,
Komachi
Y
,
Asakura
T
,
Shimosegawa
T
,
Kanai
G-I
,
Tashiro
H
, et al
In vivo Raman study of the living rat esophagus and stomach using a micro-Raman probe under an endoscope
.
Appl Spectrosc
2007
;
61
:
579
84
.
17.
Stone
N
,
Matousek
P
. 
Advanced transmission Raman spectroscopy: a promising tool for breast disease diagnosis
.
Cancer Res
2008
;
68
:
4424
30
.
18.
Stone
N
,
Baker
R
,
Rogers
K
,
Parker
AW
,
Matousek
P
. 
Subsurface probing of calcifications with spatially offset Raman spectroscopy (SORS): future possibilities for the diagnosis of breast cancer
.
Analyst
2007
;
132
:
899
905
.
19.
Saha
A
,
Barman
I
,
Dingari
NC
,
McGee
S
,
Volynskaya
Z
,
Galindo
LH
, et al
Raman spectroscopy: a real-time tool for identifying microcalcifications during stereotactic breast core needle biopsies
.
Biomed Opt Express
2011
;
2
:
2792
803
.
20.
Saha
A
,
Barman
I
,
Dingari
NC
,
Galindo
LH
,
Sattar
A
,
Liu
W
, et al
Precision of Raman spectroscopy measurements in detection of microcalcifications in breast needle biopsies
.
Anal Chem
2012
;
84
:
6715
22
.
21.
Baker
R
,
Rogers
KD
,
Shepherd
N
,
Stone
N
. 
New relationships between breast microcalcifications and cancer
.
Br J Cancer
2010
;
103
:
1034
9
.
22.
Ellis
IO
,
Humphreys
S
,
Michell
M
,
Pinder
SE
,
Wells
CA
,
Zakhour
HD
, et al
Best Practice No 179. Guidelines for breast needle core biopsy handling and reporting in breast screening assessment
.
J Clin Pathol
2004
;
57
:
897
902
.
23.
Faoláin
,
Hunter
MB
,
Byrne
JM
,
Kelehan
P
,
Lambkin
HA
,
Byrne
HJ
, et al
Raman spectroscopic evaluation of efficacy of current paraffin wax section dewaxing agents
.
J Histochem Cytochem
2005
;
53
:
121
9
.
24.
Torti
E
,
Marcinnò
B
,
Vanna
R
,
Morasso
C
,
Piccotti
F
,
Villani
L
, et al
Automatic and unsupervised identification of specific biochemical features from Raman mapping data. 2019 22nd Euromicro Conference on Digital System Design (DSD), Kallithea, Greece. IEEE Computer Society Press ed. Los Alamitos, California 2019, 464-9
.
DOI: 10.1109/DSD.2019.00073.
25.
Bigio
IJ
,
Fantini
S
.
Quantitative biomedical optics: theory, methods, and applications
.
Cambridge, United Kingdom:
Cambridge University Press
; 
2016
.
26.
Kunitake
JAMR
,
Choi
S
,
Nguyen
KX
,
Lee
MM
,
He
F
,
Sudilovsky
D
, et al
Correlative imaging reveals physiochemical heterogeneity of microcalcifications in human breast carcinomas
.
J Struct Biol
2018
;
202
:
25
34
.
27.
Kodati
VR
,
Tomasi
GE
,
Turumin
JL
,
Tu
AT
. 
Raman spectroscopic identification of calcium-oxalate-type kidney stone
.
Appl Spectrosc
1990
;
44
:
1408
11
.
28.
Lafuente
B
,
Downs
RT
,
Yang
H
,
Stone
N
. 
The power of databases: the RRUFF project
. In:
Highlights Mineral Crystallography. Berlin Germany: De Gruyter
2015
;
1
30
.
29.
Wehrmeister
U
,
Jacob
DE
,
Soldati
AL
,
Loges
N
,
Häger
T
,
Hofmeister
W
. 
Amorphous, nanocrystalline and crystalline calcium carbonates in biological materials
.
J Raman Spectrosc
2011
;
42
:
926
35
.
30.
Tlili
MM
,
Amor
MB
,
Gabrielli
C
,
Joiret
S
,
Maurin
G
,
Rousseau
P
. 
Characterization of CaCO3 hydrates by micro-Raman spectroscopy
.
J Raman Spectrosc
2002
;
33
:
10
6
.
31.
Torti
E
,
Florimbi
G
,
Castelli
F
,
Ortega
S
,
Fabelo
H
,
Callicó
GM
, et al
Parallel K-means clustering for brain cancer detection using hyperspectral images
.
Electronics
2018
;
7
:
283
.
32.
Bunk
O
,
Bech
M
,
Jensen
TH
,
Feidenhans'I
R
,
Binderup
T
,
Menzel
A
, et al
Multimodal x-ray scatter imaging
.
N J Phys
2009
;
11
:
123016
.
33.
Kraft
P
,
Bergamaschi
A
,
Bronnimann
C
,
Dinapoli
R
,
Eikenberry
EF
,
Graafsma
H
, et al
Characterization and calibration of PILATUS detectors
.
IEEE Trans Nucl Sci
2009
;
56
:
758
64
.
34.
Lutz-Bueno
V
,
Arboleda
C
,
Leu
L
,
Blunt
MJ
,
Busch
A
,
Georgiadis
A
, et al
Model-free classification of X-ray scattering signals applied to image segmentation
.
J Appl Crystallogr
2018
;
51
:
1378
86
.
35.
Rietveld
HM
. 
A profile refinement method for nuclear and magnetic structures
.
J Appl Crystallogr
1969
;
2
:
65
71
.
36.
Rodríguez-Carvajal
J
. 
Recent advances in magnetic structure determination by neutron powder diffraction
.
Physica B
1993
;
192
:
55
69
.
37.
Giannini
C
,
Ladisa
M
,
Lutz-Bueno
V
,
Terzi
A
,
Ramella
M
,
Fusaro
L
, et al
X-ray scanning microscopies of microcalcifications in abdominal aortic and popliteal artery aneurysms
.
IUCrJ
2019
;
6
:
267
76
.
38.
Jolliff
BL
,
Hughes
JM
,
Freeman
JJ
,
Zeigler
RA
. 
Crystal chemistry of lunar merrillite and comparison to other meteoritic and planetary suites of whitlockite and merrillite
.
Am Mineral
2006
;
91
:
1583
95
.
39.
de Mul
FFM
,
Otto
C
,
Greve
J
,
Arends
J
,
Bosch
JJT
. 
Calculation of the Raman line broadening on carbonation in synthetic hydroxyapatite
.
J Raman Spectrosc
1988
;
19
:
13
21
.
40.
Sathyavathi
R
,
Saha
A
,
Soares
JS
,
Spegazzini
N
,
McGee
S
,
Rao Dasari
R
, et al
Raman spectroscopic sensing of carbonate intercalation in breast microcalcifications at stereotactic biopsy
.
Sci Rep
2015
;
5
:
9907
.
41.
D'Orsi
C
,
Sickles
E
,
Mendelson
E
,
Morris
E
.
ACR BI-RADS atlas, breast imaging reporting and data system
.
Reston, VA
:
American College of Radiology
; 
2013
.
42.
Barman
I
,
Dingari
NC
,
Saha
A
,
McGee
S
,
Galindo
LH
,
Liu
W
, et al
Application of Raman spectroscopy to identify microcalcifications and underlying breast lesions at stereotactic core needle biopsy
.
Cancer Res
2013
;
73
:
3206
15
.
43.
Lagier
R
,
Baud
C-A
. 
Magnesium whitlockite, a calcium phosphate crystal of special interest in pathology
.
Pathol Res Pract
2003
;
199
:
329
35
.
44.
Scott
R
,
Stone
N
,
Kendall
C
,
Geraki
K
,
Rogers
K
. 
Relationships between pathology and crystal structure in breast calcifications: an in situ X-ray diffraction study in histological sections
.
NPJ Breast Cancer
2016
;
2
:
16029
.
45.
Loste
E
,
Wilson
RM
,
Seshadri
R
,
Meldrum
FC
. 
The role of magnesium in stabilising amorphous calcium carbonate and controlling calcite morphologies
.
J Cryst Growth
2003
;
254
:
206
18
.
46.
Beniash
E
,
Aizenberg
J
,
Addadi
L
,
Weiner
S
. 
Amorphous calcium carbonate transforms into calcite during sea urchin larval spicule growth
.
Proc Biol Sci
1997
;
264
:
461
5
.
47.
Boanini
E
,
Gazzano
M
,
Bigi
A
. 
Ionic substitutions in calcium phosphates synthesized at low temperature
.
Acta Biomater
2010
;
6
:
1882
94
.
48.
Laurencin
D
,
Almora-Barrios
N
,
de Leeuw
NH
,
Gervais
C
,
Bonhomme
C
,
Mauri
F
, et al
Magnesium incorporation into hydroxyapatite
.
Biomaterials
2011
;
32
:
1826
37
.
49.
Castiglioni
S
,
Maier
JAM
. 
Magnesium and cancer: a dangerous liason
.
Magnes Res
2011
;
24
:
S92
100
.
50.
Šupová
M
. 
Substituted hydroxyapatites for biomedical applications: a review
.
Ceram Int
2015
;
41
:
9203
31
.
51.
Gosling
S
,
Scott
R
,
Greenwood
C
,
Bouzy
P
,
Nallala
J
,
Lyburn
ID
, et al
Calcification microstructure reflects breast tissue microenvironment
.
J Mammary Gland Biol Neoplasia
2019
;
24
:
333
42
.
52.
Fernández
M
,
Keyriläinen
J
,
Serimaa
R
,
Torkkeli
M
,
Karjalainen-Lindsberg
M-L
,
Tenhunen
M
, et al
Small-angle x-ray scattering studies of human breast tissue samples
.
Phys Med Biol
2002
;
47
:
577
92
.
53.
Cox
RF
,
Hernandez-Santana
A
,
Ramdass
S
,
McMahon
G
,
Harmey
JH
,
Morgan
MP
. 
Microcalcifications in breast cancer: novel insights into the molecular mechanism and functional consequence of mammary mineralisation
.
Br J Cancer
2012
;
106
:
525
37
.
54.
Bellahcène
A
,
Castronovo
V
. 
Increased expression of osteonectin and osteopontin, two bone matrix proteins, in human breast cancer
.
Am J Pathol
1995
;
146
:
95
100
.
55.
Lev-Toaff
AS
,
Feig
SA
,
Saitas
VL
,
Finkel
GC
,
Schwartz
GF
. 
Stability of malignant breast microcalcifications
.
Radiology
1994
;
192
:
153
6
.
56.
Rudin
AV
,
Hoskin
TL
,
Fahy
A
,
Farrell
AM
,
Nassar
A
,
Ghosh
K
, et al
Flat epithelial atypia on core biopsy and upgrade to cancer: a systematic review and meta-analysis
.
Ann Surg Oncol
2017
;
24
:
3549
58
.
57.
Vidavsky
N
,
Kunitake
JAMR
,
Chiou
AE
,
Northrup
PA
,
Porri
TJ
,
Ling
L
, et al
Studying biomineralization pathways in a 3D culture model of breast cancer microcalcifications
.
Biomaterials
2018
;
179
:
71
82
.