The role of mammography in breast cancer radiomics

Cover Page


Mammography is still the only screening method for breast cancer. Although digital mammography is the most common and widely used method for detecting breast cancer, it is ineffective at detecting and assessing intratumoral heterogeneity. Due to the small size of the tissue sample or tumor, biopsies often fail to represent the entire tumor. For this reason, selecting a treatment and determining a patient’s prognosis becomes difficult. In this case, medical imaging is a noninvasive approach that can provide a more comprehensive view of the entire tumor, act as a “virtual biopsy,” and be useful for monitoring disease progression and response to therapy.

Radiomics with texture analysis allows you to look at an image as a group of numerical data, moving beyond the usual visual perception and into a deeper analysis of digital, pixel data to improve the accuracy of differential diagnosis. Radiogenomics is a natural extension of radiomics that focuses on determining gene expression based on radiologic tumor phenotype. The purpose of this review is to evaluate the role of mammography in breast cancer radiomics and radiogenomics.

The article presents a literature review of relevant Russian scientific articles found in databases such as PubMed, Medline, Springer, eLibrary, and Google Scholar. The information obtained was then pooled, structured, and analyzed to examine the role of mammography in breast cancer screening radiomics.

Full Text


Breast cancer (BC) is a pressing issue in modern oncology since it ranks first in terms of prevalence among all malignant neoplasms in women [1]. In Russia, the incidence of BC was 89.8 cases per 100,000 female population in 2018 [1]; in 2019, 73,366 breast cancer cases were detected, with 27.7% of patients in stages III and IV [2].

BC is a heterogeneous disease, which means that tumor morphology and expression subtypes differ depending on the receptor status of BC [3, 4]. Further, the expression of the estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) determines BC receptor status. The proliferation marker Ki-67 and the epidermal growth factor receptor are also immunochemically stained to determine the molecular subtype of BC [4].

The following are the five molecular subtypes of BC:

1) Luminal A [ER+, PR+ high (≥20%), HER2-, Ki-67 low (≤20%)]: estrogen-dependent low-aggressive tumors with no overexpression of HER2 protein receptors; characterized by high expression of the ER gene

2) Luminal B [ER+, PR+ low (≤20%), HER2-, Ki-67 high]: estrogen-dependent tumors with no overexpression of HER2 protein receptors

3) Luminal B [ER+, HER2+, any Ki-67 level, any PR]: estrogen-dependent aggressive tumors; expressed amplification of the HER2 oncogene; apparent expression of the ER gene

4) HER2 positive [ER- and PR-, any Ki-67, HER2+]: estrogen-independent aggressive tumors; expressed amplification of the HER2 oncogene

5) Triple negative (basal-like): estrogen-independent aggressive tumors with the worst survival rates (ER-, PR-, HER2-) [3–5]

Tumor biology is known to influence the selection of therapy as well as the outcome prognosis, with ER+ and PR+ patients having a longer relapse-free survival ability, while triple-negative BC (TNBC) (ER-, PR-, HER2-) has the most aggressive course and the worst survival rates [3, 6]. The use of biological markers to identify BC subtypes improves patient survival by allowing for more accurate disease diagnosis. For example, patients with ER and PR expression in their tumors should receive endocrine therapy, while patients with HER2 expression should receive anti-HER2 therapy [7].

Intratumoral heterogeneity is defined as the heterogeneity of the morphological structure and the variability in the expression of various markers by individual groups of cells within the same tumor [8, 9]. On the other hand, morphological intratumoral heterogeneity can be defined as diversity in different areas of the tumor, i.e., spatial heterogeneity, or as tumor progression in time, i.e., heterogeneity in time [8]. Due to such heterogeneity of neoplasms and the small size of the puncture tissue sample, the biopsy cannot reflect the histological presentation of the tumor as a whole. Therefore, choosing the appropriate treatment and determining the prognosis becomes difficult. When the tumor is small, biopsies can be difficult. In this case, a noninvasive approach such as medical imaging provides a more consistent view of the tumor and holds promise for “virtual biopsy,” as well as monitoring disease progression and response to therapy [7, 10–12].


Cancer detection at an early stage is an effective method to reduce patient mortality [13]. Mammography is still the only method for screening and diagnosing BC [10]. Although digital mammography is the most commonly used method for early detection of BC, its efficiency in detecting findings is limited, and mammography has a lower sensitivity in patients with high mammary gland density (ACR-C and D) [14], since the pathological lesion can be overlapped by fibroglandular structures in the image [15, 16]. Despite the reduced sensitivity in one of the groups of patients, digital mammography currently has the best combination of sensitivity and specificity in diagnostics of BC, but these two indicators vary between 75%–90% and 80%–90%, respectively, depending on the country [15]. In their recent study, O. Demircioglu et al. [17] showed that the interpretation of low-quality images by radiologists with limited experience leads to overdiagnosis and unnecessary painful invasive procedures in roughly half of clinical cases [6, 15, 17].

Recent advances in artificial intelligence (AI) technologies used for image analysis hold promise for detecting tumors and reducing the burden on doctors, evaluating treatment, and monitoring disease progression [6]. However, in BC, the primary tasks of clinical practice and research are early detection of the disease prognosis and prediction of the response to therapy. From this point of view, other applications of AI are possible, such as using texture analysis to determine the cancer subtypes and predict treatment response [6, 18, 19].

The interpretation of images by a radiologist with an assessment of the tumor structure, its relationship to the surrounding tissues, special aspects of the location, and structure of microcalcifications are all part of the mammographic study analysis. To create truly personalized therapy, a quantitative (objective) assessment of the lesion is also required [6].

Intratumoral heterogeneity is important for accurate diagnosis, clinical prognosis (response to treatment, survival rate, disease progression, etc.), and treatment of oncological diseases [20, 21]. Early detection of tumor resistance to therapy is critical for improving outcomes, allowing for timely treatment regimen changes [6].

Thus, there is a need to improve the efficiency of detection, prediction of outcome, and response to treatment of BC. A unique set of techniques, combined in radiomics and radiogenomics, is gaining traction as a tool for maximizing the information that can be extracted from virtually any modality of digital medical imaging [15].


Radiation diagnostic images contain information that indicates pathophysiological processes, and this relationship can be identified using quantitative image analysis [22]. To put it another way, tumor characteristics at the cellular and genetic levels are reflected in the phenotypic patterns of the tumor, which can be manifested and detected in images [23].

Radiomics is a process that includes the stages of preparation and subsequent quantitative analysis of multidimensional data obtained from digital medical images (the “omic” suffix appears in the names of molecular biology fields that deal with large amounts of data [24]). Radiomics is defined as image analysis that uses specific algorithms to extract numerical characteristics of images in order to create classification models to improve medical decision-making support, as well as to determine the disease prognosis [25, 26] and treatment [27], which is especially significant for personalized therapy. In radiomics, one area of interest in an image is used to obtain a set (sometimes tens or hundreds) of numerical characteristics, each of which can hold a certain information and theoretical aspect (often referred to as a “radiomic sign”), which is not available in normal viewing of images [15]. Radiomics transforms medical imaging data into a dataset of order statistics by using automatic texture sign extraction algorithms for digital medical images [28]. In other words, radiomics with the use of texture analysis allows you to think of an image as a collection of numerical characteristics, go beyond the usual visual perception, and analyze multidimensional data.

Radiogenomics is a technology that connects a patient’s genotype to an imaging phenotype. It should be noted that the term “radiogenomics” can also refer to genetic variability and its relationship with response to radiation therapy [29, 30], but it is more often used to assess the relationship between the image characteristics of a tumor or any other disease and its gene expression patterns and gene mutations [25, 26].

Radiogenomics is a method for determining gene expression in a tumor based on its radiation phenotype. This is important because tumors are heterogeneous, and radiomics data are extracted from the region of interest (tumor) as a whole rather than from a separate sample [22]. Radiogenomics also allows for the assessment of treatment response that is not solely based on the traditional measurement of tumor size over time [25]. The combination of radiomics and radiogenomics can detect gene abnormalities in images [6]. Radiomics and radiogenomics improve the accuracy of clinical diagnosis and have prognostic value by identifying relationships between various types of clinical data [22].


When considering radiomics as a process, several major stages can be distinguished, namely, image acquisition, highlighting the area of interest, extraction of radiomic signs from the area of interest (texture analysis of images), analysis of textural signs, and construction of various prediction and classification models using the obtained radiomic data with the option of including additional information (e.g., clinical, demographic, or genomic data; the presence of comorbid conditions) [22, 23, 31]. The stages of radiomics are depicted in Fig. 1, and their more detailed characteristics are discussed further below:


Fig. 1. The diagram illustrates the typical stages in radiomics. After obtaining medical images (1), they are manually or automatically segmented (2). Using special software or programming language modules, radiomic signs of the first and higher orders are extracted from segmented regions of interest (3). Next, the analysis and selection of the most significant textural signs obtained are carried out. Finally, based on the analyzed radiomic data, various clinical and diagnostic models of classification or prediction are constructed (4)


  1. Determination of the clinical problem and acquisition of digital medical images, excluding low-quality studies.
  2. Segmentation of images to the main analyzed areas of interest [32], such as a malignant neoplasm, to assess intratumoral heterogeneity. Many tumors have indistinct boundaries, which complicates the reproducibility of their segmentation [33]. Although it is preferable to use semiautomatic or fully automatic selection of the area of interest using special software, in some cases, expert specification and manual selection are required [23, 34]. The selection process of the region of interest is not standardized, and the region of interest may contain the entire tumor or some of its parts [35, 36]. Manually determining the region of interest is time-consuming and variable due to differences in image interpretation by different radiologists [33], which ultimately affects the accuracy of the radiomic models constructed; however, modern deep learning technologies using big data are capable of mitigating this effect [37].
  3. Extraction of a variety of radiomic signs from a segmented region of interest using mathematical operations involving numerical values of intensities and relative positions of pixels or voxels in images. The extracted quantitative signs are classified into two categories: morphological signs (volume and shape) and histogram signs (description of the intensity of gray tone levels) of the first, second, and higher orders [26, 34].

Morphological signs reflect the shape of the region of interest. For planar images, 2D signs of shape are relevant, such as the perimeter-to-surface ratio and roundness as a measure of the approximation of the shape of the region of interest to the shape of a circle. For example, a stellate tumor will have a higher surface-to-volume ratio than a round tumor [31].

First-order histogram signs [38] indicate the distribution of gray-level intensities for pixels in the region of interest. The most common signs (mean and median) in this category indicate the width of the range of intensities; entropy is a measure of irregularity in the distribution of intensities (higher values indicate a more heterogeneous region) [39]. However, first-order statistics do not account for the spatial arrangement of pixels.

Second-order histogram signs [38], also known as texture signs, indicate the spatial relationship between two adjacent pixels with the same or different brightness values. The most common technique for extracting texture signs is based on a gray-level co-occurrence matrix, which is a matrix whose rows and columns represent gray intensity-level values; the matrix cells indicate the number of times the corresponding gray values are in a certain relationship (angle and distance between the pixels analyzed). For example, signs obtained by using such a matrix include second-order entropy, which indicates heterogeneity; energy, which describes image homogeneity; and contrast range, which determines the local change in intensities [10]. In radiomics, texture analysis provides information on the measure of intratumoral heterogeneity [22, 40].


Fig. 2. Comparison of histogram signs of the first and second orders. The two different initial regions of interest of the segmented image (a) comprise an equal number of pixels in light gray, dark gray, and black shades. Brightness histograms based on the number of pixels of certain shades (histogram signs of the first order) are the same (b). These signs do not indicate the mutual arrangement of the pixels. Adjacency matrices (second-order histogram signs) reflect the heterogeneity of images (c). In the future, mathematical algorithms derived from the obtained histograms of intensities and adjacency matrices of the gray level will be used to calculate a variety of radiomic signs for analysis and modeling


Figure 2 shows a comparison of the histogram signs of the first and second orders, as well as the formation of the adjacency matrix of the gray tone level, where Fig. 2 (a) presents two original images, Fig. 2 (b) histograms of the first order, and Fig. 2 (c) grayscale adjacency matrices obtained for the original images. The row and column headings of these matrices contain the shade of gray numbers. Each cell of the table contains the number of horizontal pairs of pixels, in which pixels with the shade indicated in the header of the row and column of this cell are located relative to each other at an angle of 0º [41]. Subsequently, mathematical algorithms from the obtained histograms of intensities and adjacency matrices of the gray level are used to calculate a set of radiomics signs for analysis and modeling.

  1. Analysis and modeling: the radiomic signs obtained, depending on the question posed, can be analyzed in various ways, ranging from statistical models to machine learning methods.

Given the large amount of data extracted from the images, step 1 is selection or reduction of signs. Irreproducible signs should be excluded, since they most probably lead to false results of the models constructed [42, 43].

Step 2 is multivariate data analysis [31] and the construction of models classified into three main groups: predictive, explanatory, and descriptive [15]. Descriptive models are used to obtain a broad representation of each sign, summarizing its key characteristics. Thus, explanatory methods often used for biomedical data frequently focus on the ability of the model to establish a relationship between a sign and an outcome, such as the relationship between the texture characteristics of the gray-level coincidence matrix and the morphological type of BC within the region of interest. Further, machine learning methods are used to create predictive models, which analyze the probability of certain outcomes based on the input data obtained [15], such as a radiomic model for predicting the lack of response to neoadjuvant BC chemotherapy. Before using the models in clinical settings, the quality and reproducibility of the results of operation obtained should be assessed [31].


Recognition of a malignant neoplasm

The most difficult and crucial step in mammography is classifying mammogram findings as benign or malignant [44]. In their recent study, N. Mao et al. [45] demonstrated that using quantitative signs in conjunction with AI can provide greater diagnostic efficiency when using mammography compared to the efficiency of diagnostics performed by experienced radiologists [15].

The process of classifying microcalcifications as benign or malignant based on images is still a difficult task for radiologists [46]. When suspicious calcifications are detected, texture analysis of images can be performed in conjunction with AI methods, potentially reducing the number of unnecessary biopsies [47, 48].

Specific features of the mammary gland parenchyma may reflect biological risk factors for BC. H. Li et al. [49] showed that using textural signs extracted from mammograms of the affected and contralateral (with normal parenchyma) glands improves the accuracy of digital mammography in the diagnosis of BC. Studies reveal that radiomics with high sensitivity and specificity can distinguish between malignant and benign mammary gland neoplasms [50].

Definition of BC subtypes

Recent radiogenomics studies have confirmed the relationship between MR signs of BC imaging and molecular subtypes, namely, luminal A, luminal B, HER2, and TNBC [51]. Although mammography images provide less information than magnetic resonance imaging (MRI), several studies are currently underway to demonstrate the potential of mammography in radiomics and radiogenomics of BC. In their study, W. Ma et al. [10] demonstrated the possibility of predicting the molecular subtype of BC by extracting radiomic characteristics from mammographic images. The most significant signs were roundness, concavity, mean gray value, and correlation. The results revealed that luminal subtypes and TNBC have distinct textural signs, in contrast to other subtypes, which allow them to be quantitatively distinguished using radiomics.

In some BC patients, the use of neoadjuvant chemotherapy does not provide an effective therapeutic response, resulting in delayed surgery, poor prognosis, and increase in treatment costs. Moreover, the use of radiomics in conjunction with independent clinical risk factors (e.g., Ki-67 index, HER2 status) has been shown to improve the predictive model of nonresponse to neoadjuvant chemotherapy [52].

Early detection of a more aggressive subtype of BC, namely, TNBC, using medical imaging will allow clinicians to prescribe treatment prior to definitive biopsy confirmation [53]. In a study by H.X. Zhang et al. [53], TNBC had greater roundness and concavity compared to other subtypes; the area under the ROC curve (receiver operating characteristic curve; classic ROC curve, a graph of sensitivity versus specificity [54]) was used to assess the accuracy of these two signs in differentiating TNBC from other BC subtypes and was greater than 0.70 [53, 55]. In this study, the skewness coefficient (a histogram attribute reflecting the skewness of the distribution of values relative to the mean) of all subtypes was less than 0 (negative or left-sided skewness). Further, the asymmetry coefficient of TNBC was found to be lower than the coefficients of the other subtypes under study. Therefore, the above radiomic signs can be considered as potential markers of differences between TNBC and other subtypes of BC in the future [53].

Predicting the development of BC and the possibility of personalized screening

Radiomics-based technologies can help advance personalized screening by developing tools for individual risk assessment and including them in decision-making support tools for mammographic screening, as well as individual screening intervals [56–58]. A higher density of mammary glands has been linked to an increased risk of BC development [59]. The term “density” refers to the degree of attenuation of X-ray radiation as it passes through the gland and reflects the distribution of fibroglandular tissue. However, the definition of density alone does not represent the entire complexity of the gland structure. Image-derived textural signs have been proposed as markers of changes in the parenchyma, indicating a link to the development of BC [57, 59]. In their study, D. Kontos et al. [59] (2019) identified radiomic phenotypes on mammograms that reflect the complexity of the parenchyma (in addition to density) and are independently associated with BC. In contrast to the conventional definition of density, textural signs indicated a subtler and more localized complexity of the parenchymal pattern. The density of the mammary glands differed between the phenotypes of low and medium complexity of the parenchyma but was similar for the other phenotypes. There are interesting data on the phenotype with the least complexity (parenchyma complexity) in women with high mammary gland density due to their greater homogeneity, whereas the phenotype with low and medium parenchyma complexity included a small number of high-density images [59].

Preoperative detection of axillary lymph node metastases

BC metastases are most commonly found in the axillary lymph nodes. Axillary lymph node status is an important factor in determining overall and relapse-free survival in BC patients [60]. An accurate preoperative determination of the status of the axillary lymph nodes can provide doctors with information that allows them to decide whether or not to perform lymphadenectomy and prescribe adjuvant therapy. Currently, the status is determined by biopsy of the sentinel lymph node, which can lead to complications, such as damage to blood vessels and nerves, as well as the development of lymphedema; and diagnostics using imaging methods has a low sensitivity [60]. J. Yang et al. [60] developed a model that includes radiomic signs extracted from mammograms, which can be used as a noninvasive method for determining metastases in the axillary lymph nodes prior to surgery when combined with additional clinical and pathological information.


Although radiomics and radiogenomics hold great promise for the advancement of personalized medicine, they must be validated on an independent dataset to confirm their diagnostic and predictive value. It will take time for these technologies to gain significant practical value in cancer research and even more time before they can be applied in clinical practice. These limitations are due to the fact that the available large amounts of data do not currently contain the full characteristics of patients [6]. The complexity of the reproducibility of radiomics results is associated with disadvantages at each stage, namely, different textural signs are obtained on different equipment and visualization protocols [61, 62]; the gold standard for manual tumor segmentation is time-consuming and operator-dependent [63]; semiautomatic and automatic segmentations, which reduce variability [64, 65], are not standardized; there is obvious repeatability between texture signs, necessitating the reduction of the size of the data [66, 67]; and there is no clear explanation of the relationship between the unit of radiomics (the basic unit of the texture) and human tissues. Furthermore, any “meaningful” research results obtained should be reviewed when the underlying theory is unclear and technical methods are not standardized [68].


Convincing evidence have been accumulated that MRI of the mammary glands is superior in diagnostic accuracy to traditional diagnostic methods such as mammography [69]. Aside from detecting more cases of duct carcinoma in situ, MRI of the mammary glands often changes the stage of the oncological process, which helps to optimize the treatment process.

It has been established that radiomic signs extracted from MR images of mammary glands indicate tumor heterogeneity and vascularization [70], as well as enable to differentiate duct carcinoma from a benign focus [71]. Existing radiomic models continue to lag behind expert mammologists in terms of area under the curve for differentiating benign from malignant lesions [72]. However, promising results in identifying suspicious (BI-RADS 4 and 5) lesions using diffusion-weighted imaging radiomics have been obtained [73].

Radiomics appears to be capable of assisting in clinical decision-making while avoiding potentially invasive interventions in the armpit. Two different studies have shown that the radiomic model can predict sentinel lymph node metastases [74, 75], which is extremely important in clinical practice.

Another application of radiomics is associated with the Ki-67 proliferation index, which is used as a prognostic marker in BC [72]. Recent studies have investigated the possibility of predicting the expression of the Ki-67 proliferation marker using radiomics of a series of dynamic contrast enhancements [76–79].


One of the key concepts in radiomics is that ray diagnostic images contain data that can provide more information about the region of interest than previously believed. Mammography is the most effective method for early detection of BC. Mammographic images can be used for radiomic analysis, which can be used to identify malignant neoplasms, BC subtypes, disease progression, and response to treatment.

Radiomics-based technologies, such as in the field of mammography, may be incorporated into medical decision-making support tools in the future to determine strategies for individual screening, follow-up, and possibly preventive therapy. However, it should be noted that radiomics is still in its early stages of development, with much more research needed before clinical application.

Radiomics mammography provides important diagnostic and prognostic information about BC, which has the potential to reduce the need for invasive and often complicated procedures. Although radiomics-based techniques will not replace biopsy in the near future, pending support with new research, the integration of radiomics and radiogenomics into clinical practice will be one of the most important and promising tasks in achieving the goal of reducing BC mortality.


Funding. The authors declare that there is no external funding for the exploration and analysis work.

Conflict of interest. The authors declare no obvious and potential conflicts of interest related to the publication of this article.

Authors’ contribution. V.G. Govorukhina, S.S. Semenov ― search for publications on the topic, analysis of literature, writing a text; P.B. Gelezhe, V.V. Didenko ― search for publications on the topic, text writing, expert assessment of literature, S.P. Morozov, A.E. Andreychenko ― defining the main focus of the review, expert assessment of the literature review, systematization and editing of the review. All authors made a substantial contribution to the conception of the work, acquisition, analysis, interpretation of data for the work, drafting and revising the work, final approval of the version to be published and agree to be accountable for all aspects of the work.


About the authors

Veronika G. Govorukhina

Lomonosov Moscow State University

ORCID iD: 0000-0002-1611-9618
SPIN-code: 7038-8580


Russian Federation, 24-1 Petrovka street, 127051 Moscow

Serafim S. Semenov

Research and Practical Clinical Center of Diagnostics and Telemedicine Technologies, Department of Healthcare of Moscow; Moscow Clinical Scientific Center n.a. A.S. Loginov, Moscow Healthcare Department

Author for correspondence.
ORCID iD: 0000-0003-2585-0864
SPIN-code: 4790-0416

Engineer of Medical Informatics, Radiomics & Radiogenomics group, Radiology Clinical Resident, MD

Russian Federation, Moscow; Moscow

Pavel B. Gelezhe

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of Moscow Health Care

ORCID iD: 0000-0003-1072-2202
SPIN-code: 4841-3234

MD, Cand. Sci. (Med.)

Russian Federation, Moscow

Vera V. Didenko

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of Moscow Health Care; City Clinical Oncology Hospital № 1

ORCID iD: 0000-0001-9068-1273
SPIN-code: 5033-8376


Russian Federation, Moscow; Moscow

Sergey P. Morozov

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of Moscow Health Care

ORCID iD: 0000-0001-6545-6170
SPIN-code: 8542-1720

MD, Dr. Sci. (Med.), Professor

Russian Federation, Moscow

Anna E. Andreychenko

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of Moscow Health Care

ORCID iD: 0000-0001-6359-0763
SPIN-code: 6625-4186

Cand. Sci. (Phys.-Math.)

Russian Federation, Moscow


  1. Kaprin AD, Starinsky VV, Petrova GV. Malignant neoplasms in Russia in 2018 (morbidity and mortality). Moscow : P.A. Herzen Moscow State Research Institute; 2019. 250 р. (In Russ).
  2. Kaprin AD, Starinsky VV. The state of oncological care for the population of Russia in 2019. Moscow : P.A. Herzen Moscow State Research Institute; 2020. 239 p. (In Russ).
  3. Grishina KA, Muzaffarova TA, Khailenko VA, Karpukhin AV. Molecular genetic markers of breast cancer. Tumors of the female reproductive system. 2016;12(3):36–42. (In Russ). doi: 10.17650/1994-4098-2016-12-3-36-42
  4. Wu M, Ma J. Association between imaging characteristics and different molecular subtypes of breast cancer. Acad Radiol. 2017;24(4):426–434. doi: 10.1016/j.acra.2016.11.012
  5. Stenina MB, Zhukova LG, Koroleva IA, et al. Practical recommendations for the drug treatment of breast cancer. Malig tumours. 2021;10(3). (In Russ).
  6. Pesapane F, Suter MB, Rotili A, et al. Will traditional biopsy be substituted by radiomics and liquid biopsy for breast cancer diagnosis and characterisation? Med Oncol. 2020;37(4):29. doi: 10.1007/s12032-020-01353-1
  7. Januškevičienė I, Petrikaitė V. Heterogeneity of breast cancer: the importance of interaction between different tumor cell populations. Life Sci. 2019;(239):117009. doi: 10.1016/j.lfs.2019.117009
  8. Turashvili G, Brogi E. Tumor heterogeneity in breast cancer. Front Med. 2017;4:227. doi: 10.3389/fmed.2017.00227
  9. Zavyalova MV, Vtorushin SV, Tsyganov MM. Intra-tumor heterogeneity: nature and biological significance. Review. Biochemistry. 2013;78(11):1531–1549. (In Russ).
  10. Ma W, Zhao Y, Ji Y, et al. Breast cancer molecular subtype prediction by mammographic radiomic features. Acad Radiol. 2019;26(2):196–201. doi: 10.1016/j.acra.2018.01.023
  11. Aerts HJ, Velazquez ER, Leijenaar RT, et al. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nat Commun. 2014;5:4006. doi: 10.1038/ncomms5006
  12. Lin F, Wang Z, Zhang K, et al. Contrast-Enhanced spectral mammography-based radiomics nomogram for identifying benign and malignant breast lesions of Sub-1 cm. Front Oncol. 2020;10:573630. doi: 10.3389/fonc.2020.573630
  13. Li X, Qin G, He Q, et al. Digital breast tomosynthesis versus digital mammography: integration of image modalities enhances deep learning-based breast mass classification. Eur Radiol. 2020;30(2):778–788. doi: 10.1007/s00330-019-06457-5
  14. Nazari SS, Mukherjee P. An overview of mammographic density and its association with breast cancer. Breast Cancer. 2018;25(3):259–267. doi: 10.1007/s12282-018-0857-5
  15. Conti A, Duggento A, Indovina I, et al. Radiomics in breast cancer classification and prediction. Semin Cancer Biol. 2021;72:238–250. doi: 10.1016/j.semcancer.2020.04.002
  16. Hogg P, Kelly J, Mercer C. Digital mammography: A holistic approach. Springer; 2015. 309 p.
  17. Demircioglu O, Uluer M, Aribal E. How many of the biopsy decisions taken at inexperienced breast radiology units were correct? J Breast Heal. 2017;13(1):23–26. doi: 10.5152/tjbh.2016.2962
  18. Valdora F, Houssami N, Rossi F, et al. Rapid review: radiomics and breast cancer. Breast Cancer Res Treat. 2018;169(2):217–229. doi: 10.1007/s10549-018-4675-4
  19. Mayerhoefer ME, Materka A, Langset G, et al. Introduction to radiomics. J Nucl Med. 2020;61(4):488–495. doi: 10.2967/jnumed.118.222893
  20. Lambin P, Zindler J, Vanneste BG, et al. Decision support systems for personalized and participative radiation oncology. Adv Drug Deliv Rev. 2017;109:131–153. doi: 10.1016/j.addr.2016.01.006
  21. Wen YL, Leech M. Review of the role of radiomics in tumour risk classification and prognosis of cancer. Anticancer Research. 2020;40(7):3605–3618. doi: 10.21873/anticanres.14350
  22. Gillies RJ, Kinahan PE, Hricak H. Radiomics: Images are more than pictures, they are data. Radiology. 2016;278(2):563–577. doi: 10.1148/radiol.2015151169
  23. Lambin P, Leijenaar RT, Deist TM, et al. Radiomics: The bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14(12):749–762. doi: 10.1038/nrclinonc.2017.141
  24. Vailati-Riboni M, Palombo V, Loor JJ. What are omics sciences? Periparturient diseases of dairy cows: a systems biology approach. 2017. Р. 1–7. doi: 10.1007/978-3-319-43033-1_1
  25. Porcu M, Solinas C, Mannelli L, et al. Radiomics and “radi-omics” in cancer immunotherapy: a guide for clinicians. Crit Rev Oncol Hematol. 2020;154:103068. doi: 10.1016/j.critrevonc.2020.103068
  26. Ognerubov NA, Shatov IA, Shatov AV. Radiogenomics and radiomics in the diagnostics of malignant tumours: a literary review. Tambov Univ Reports Ser Nat Tech Sci. 2017;22(6-2):1453–1460. doi: 10.20310/1810-0198-2017-22-6-1453-1460
  27. Nasief H, Hall W, Zhenget C, et al. Improving treatment response prediction for chemoradiation therapy of pancreatic cancer using a combination of delta-radiomics and the clinical biomarker CA19-9. Front Oncol. 2020;9:1464. doi: 10.3389/fonc.2019.01464
  28. Lambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics: Extracting more information from medical images using advanced feature analysis. Eur J Cancer. 2012;48(4):441–446. doi: 10.1016/j.ejca.2011.11.036
  29. Kerns SL, Ostrer H, Rosenstein BS. Radiogenomics: Using genetics to identify cancer patients at risk for development of adverse effects following radiotherapy. Cancer Discovery. 2014;4(2):155–165. doi: 10.1158/2159-8290
  30. Neri E, Del Re M, Paiar F, et al. Radiomics and liquid biopsy in oncology: the holons of systems medicine. Insights Imaging. 2018;9(6):915–924. doi: 10.1007/s13244-018-0657-7
  31. Rizzo S, Botta F, Raimondi S, et al. Radiomics: the facts and the challenges of image analysis. Eur Radiol Exp. 2018;2(1):36. doi: 10.1186/s41747-018-0068-z
  32. Kumar V, Gu Y, Basu S, et al. Radiomics: The process and the challenges. Magn Reson Imaging. 2012;30(9):1234–1248. doi: 10.1016/j.mri.2012.06.010
  33. Sala E, Mema E, Himoto Y, et al. Unravelling tumour heterogeneity using next-generation imaging: radiomics, radiogenomics, and habitat imaging. Clinical Radiology. 2017;72(1):3–10. doi: 10.1016/j.crad.2016.09.013
  34. Lee SH, Park H, Ko ES. Radiomics in breast imaging from techniques to clinical applications: A review. Korean J Radiol. 2020;21(7):779–792. doi: 10.3348/kjr.2019.0855
  35. Tagliafico AS, Piana M, Schenone D, et al. Overview of radiomics in breast cancer diagnosis and prognostication. Breast. 2020;(49):74–80. doi: 10.1016/j.breast.2019.10.018
  36. Li H, Giger ML, Huo Z, et al. Computerized analysis of mammographic parenchymal patterns for assessing breast cancer risk: Effect of ROI size and location. Med Phys. 2004;31(3):549–555. doi: 10.1118/1.1644514
  37. Holbrook MD, Blocker SJ, Mowery YM, et al. Mri-based deep learning segmentation and radiomics of sarcoma in mice. Tomography. 2020;6(1):23–33. doi: 10.18383/j.tom.2019.00021
  38. Pratt W. Digital Image Processing. Moscow : World; 1982. 480 p.
  39. Santos JM, Oliveira BC, de Araujo-Filho J, et al. State-of-the-art in radiomics of hepatocellular carcinoma: a review of basic principles, applications, and limitations. Abdom Radiol. 2020;45(2):342–353. doi: 10.1007/s00261-019-02299-3
  40. Avanzo M, Stancanello J, El Naqa I. Beyond imaging: The promise of radiomics. Phys Medica. 2017;(38):122–139. doi: 10.1016/j.ejmp.2017.05.071
  41. Verma M, Raman B, Murala S. Local extrema co-occurrence pattern for color and texture image retrieval. Neurocomputing. 2015;165:255–269. doi: 10.1016/j.neucom.2015.03.015
  42. Tunali I, Hall LO, Napel S, et al. Stability and reproducibility of computed tomography radiomic features extracted from peritumoral regions of lung cancer lesions. Med Phys. 2019;46(11):5075–5085. doi: 10.1002/mp.13808
  43. Nasief H, Zheng C, Schott D, et al. A machine learning based delta-radiomics process for early prediction of treatment response of pancreatic cancer. NPJ Precis Oncol. 2019;3(1):25. doi: 10.1038/s41698-019-0096-z
  44. Cui Y, Li Y, Xing D, et al. Improving the prediction of benign or malignant breast masses using a combination of image biomarkers and clinical parameters. Front Oncol. 2021;11:629321. doi: 10.3389/fonc.2021.629321
  45. Mao N, Yin P, Wang Q, et al. Added value of radiomics on mammography for breast cancer diagnosis: a feasibility study. J Am Coll Radiol. 2019;16(4 Pt A):485–491. doi: 10.1016/j.jacr.2018.09.041
  46. Fanizzi A, TM, Losurdo L, et al. A machine learning approach on multiscale texture analysis for breast microcalcification diagnosis. BMC Bioinformatics. 2020;21(Suppl 2):91. doi: 10.1186/s12859-020-3358-4
  47. Stelzer PD, Steding O, Raudner MW, et al. Combined texture analysis and machine learning in suspicious calcifications detected by mammography: Potential to avoid unnecessary stereotactical biopsies. Eur J Radiol. 2020;132:109309. doi: 10.1016/j.ejrad.2020.109309
  48. Karahaliou A, Skiadopoulos S, Boniatis I, et al. Texture analysis of tissue surrounding microcalcifications on mammograms for breast cancer diagnosis. Br J Radiol. 2007;80(956):648–656. doi: 10.1259/bjr/30415751
  49. Li H, Mendel KR, Lan L, et al. Digital mammography in breast cancer: Additive value of radiomics of breast parenchyma. Radiology. 2019;291(1):15–20. doi: 10.1148/radiol.2019181113
  50. Parekh VS, Jacobs MA. MPRAD: A multiparametric radiomics framework. arXiv. 2018.
  51. Rozhkova NI, Bozhenko VK, Burdina II, et al. Radiogenomics of breast cancer as new vector of interdisciplinary integration of radiation and molecular biological technologies (literature review). Med Alph. 2020;(20):21–29. doi: 10.33667/2078-5631-2020-20-21-29
  52. Wang Z, Lin F, Ma H, et al. Contrast-Enhanced spectral mammography-based radiomics nomogram for the prediction of neoadjuvant chemotherapy-insensitive breast cancers. Front Oncol. 2021;11:605230. doi: 10.3389/fonc.2021.605230
  53. Zhang HX, Sun ZQ, Cheng YG, et al. A pilot study of radiomics technology based on X-ray mammography in patients with triple-negative breast cancer. J Xray Sci Technol. 2019;27(3):485–492. doi: 10.3233/XST-180488
  54. Morozov SP, Vladzimirsky AV, Klyashtornyy VG, et al. Clinical acceptance of software based on artificial intelligence technologies (Radiology). Moscow: Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies; 2019. 45 р.
  55. Mandrekar JN. Receiver operating characteristic curve in diagnostic test assessment. J Thorac Oncol. 2010;5(9):1315–1316. doi: 10.1097/JTO.0b013e3181ec173d
  56. Rahbar H, McDonald ES, Lee JM, et al. How can advanced imaging be used to mitigate potential breast cancer overdiagnosis? Academic Radiology. 2016;23(6):768–773. doi: 10.1016/j.acra.2016.02.008
  57. Pinker K. Beyond breast density: Radiomic phenotypes enhance assessment of breast cancer risk. Radiology. 2019;290(1):50–51. doi: 10.1148/radiol.2018182296
  58. Sun W, Tseng TL, Qian W, et al. Using multiscale texture and density features for near-term breast cancer risk analysis. Med Phys. 2015;42(6):2853–2862. doi: 10.1118/1.4919772
  59. Kontos D, Winham SJ, Oustimov А, et al. Radiomic phenotypes of mammographic parenchymal complexity: Toward augmenting breast density in breast cancer risk assessment. Radiology. 2019;290(1):41–49. doi: 10.1148/radiol.2018180179
  60. Yang J, Wang T, Yang L, et al. Preoperative prediction of axillary lymph node metastasis in breast cancer using mammography-based radiomics method. Sci Rep. 2019;9(1):4429. doi: 10.1038/s41598-019-40831-z
  61. Zhao B, Tan Y, Tsai WY, et al. Reproducibility of radiomics for deciphering tumor phenotype with imaging. Sci Rep. 2016;6:23428. doi: 10.1038/srep23428
  62. Lu L, Ehmke RC, Schwartz LH, Zhao B. et al. Assessing agreement between radiomic features computed for multiple CT imaging settings. PLoS One. 2016;11(12):e0166550. doi: 10.1371/journal.pone.0166550
  63. Velazquez ER, Parmar C, Jermoumi M, et al. Volumetric CT-based segmentation of NSCLC using 3D-Slicer. Sci Rep. 2013;3:3529. doi: 10.1038/srep03529
  64. Qiu Q, Duan J, Gong G, et al. Reproducibility of radiomic features with GrowCut and GraphCut semiautomatic tumor segmentation in hepatocellular carcinoma. Transl Cancer Res. 2017;6(5). doi: 10.21037/tcr.2017.09.47
  65. Qiu Q, Duan J, Duan Z, et al. Reproducibility and non-redundancy of radiomic features extracted from arterial phase CT scans in hepatocellular carcinoma patients: Impact of tumor segmentation variability. Quant Imaging Med Surg. 2019;9(3):453–464. doi: 10.21037/qims.2019.03.02
  66. Hunter LA, Krafft S, Stingo F, et al. High quality machine-robust image features: Identification in nonsmall cell lung cancer computed tomography images. Med Phys. 2013;40(12):121916. doi: 10.1118/1.4829514
  67. O’Connor JP, Aboagye EO, Adams JE, et al. Imaging biomarker roadmap for cancer studies. Nat Rev Clin Oncol. 2017;14(3):169–186. doi: 10.1038/nrclinonc.2016.162
  68. Chalkidou A, O’Doherty M, Marsden PK. False discovery rates in PET and CT studies with texture features: a systematic review. PLoS One. 2015;10(5):e0124165. doi: 10.1371/journal.pone.0124165
  69. Mann RM, Kuhl CK, Moy L. Contrast-enhanced MRI for breast cancer screening. J Magn Reson Imaging. 2019;50(2):377–390. doi: 10.1002/jmri.26654
  70. Parekh VS, Jacobs MA. Integrated radiomic framework for breast cancer and tumor biology using advanced machine learning and multiparametric MRI. NPJ Breast Cancer. 2017;3:43. doi: 10.1038/s41523-017-0045-3
  71. Whitney HM, Taylor NS, Drukker K, et al. Additive benefit of radiomics over size alone in the distinction between benign lesions and luminal a cancers on a large clinical breast MRI dataset. Acad Radiol. 2019;26(2):202–209. doi: 10.1016/j.acra.2018.04.019
  72. Crivelli P, Ledda RE, Parascandolo N, et al. A new challenge for radiologists: radiomics in breast cancer. BioMed Research International. 2018;2018:6120703. doi: 10.1155/2018/6120703
  73. Bickelhaupt S, Paech D, Kickingereder P, et al. Prediction of malignancy by a radiomic signature from contrast agent-free diffusion MRI in suspicious breast lesions found on screening mammography. J Magn Reson Imaging. 2017;46(2):604–616. doi: 10.1002/jmri.25606
  74. Han L, Zhu Y, Liu Z, et al. Radiomic nomogram for prediction of axillary lymph node metastasis in breast cancer. Eur Radiol. 2019;29(7):3820–3829. doi: 10.1007/s00330-018-5981-2
  75. Dong Y, Feng Q, Yang W, et al. Preoperative prediction of sentinel lymph node metastasis in breast cancer based on radiomics of T2-weighted fat-suppression and diffusion-weighted MRI. Eur Radiol. 2018;28(2):582–591. doi: 10.1007/s00330-017-5005-7
  76. Ma W, Ji Y, Qi L, et al. Breast cancer Ki67 expression prediction by DCE-MRI radiomics features. Clin Radiol. 2018;73(10):909.e1–909.e5. doi: 10.1016/j.crad.2018.05.027
  77. Jagadish K, Sheela GM, Naidu BP, et al. Big data analytics and radiomics to discover diagnostics and therapeutics for gastric cancer. In: Recent Advancements in Biomarkers and Early Detection of Gastrointestinal Cancers. 2020. Р. 213–219. doi: 10.1007/978-981-15-4431-6_12
  78. Scheckenbach K. Radiomics: Big data instead of biopsies in the future? Laryngorhinootologie. 2018;97(S 01):S114–S141. doi: 10.1055/s-0043-121964
  79. European Society of Radiology (ESR). Medical imaging in personalised medicine: a white paper of the research committee of the European Society of Radiology (ESR). Insights Imaging. 2015;6(2):141–155. doi: 10.1007/s13244-015-0394-0.

Supplementary files

Supplementary Files
1. Fig. 1. The diagram illustrates the typical stages in radiomics. After obtaining medical images (1), they are manually or automatically segmented (2). Using special software or programming language modules, radiomic signs of the first and higher orders are extracted from segmented regions of interest (3). Next, the analysis and selection of the most significant textural signs obtained are carried out. Finally, based on the analyzed radiomic data, various clinical and diagnostic models of classification or prediction are constructed (4)

Download (167KB)
2. Fig. 2. Comparison of histogram signs of the first and second orders. The two different initial regions of interest of the segmented image (a) comprise an equal number of pixels in light gray, dark gray, and black shades. Brightness histograms based on the number of pixels of certain shades (histogram signs of the first order) are the same (b). These signs do not indicate the mutual arrangement of the pixels. Adjacency matrices (second-order histogram signs) reflect the heterogeneity of images (c). In the future, mathematical algorithms derived from the obtained histograms of intensities and adjacency matrices of the gray level will be used to calculate a variety of radiomic signs for analysis and modeling

Download (231KB)



Abstract: 607

PDF (Russian): 123

PDF (English): 13

PDF (Chinese): 37


Article Metrics

Metrics Loading ...


Copyright (c) 2021 Govorukhina V.G., Semenov S.S., Gelezhe P.B., Didenko V.V., Morozov S.P., Andreychenko A.E.

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

This website uses cookies

You consent to our cookies if you continue to use our website.

About Cookies