Accuracy of Distinguishing Atypical Ductal Hyperplasia From Ductal Carcinoma In Situ With Convolutional Neural Network-Based Machine Learning Approach Using Mammographic Image Data

Radiomic and clinical model for predicting atypical ductal hyperplasia upgrades and potentially reduce unnecessary surgical treatments

OBJECTIVE

To identify patients with atypical ductal hyperplasia (ADH) at low risk of upgrading to carcinoma. This study aims to assess the performance of radiomics combined with clinical factors to predict occult breast cancer among women diagnosed with ADH.

METHODS

This study retrospectively included microcalcification clusters of patients who underwent Mx and VABB with a diagnosis of ADH at a tertiary center from January 2015 to May 2023. Clinical and radiological data (age, cluster size, BI-RADS classification, mammographic density, breast cancer history, residual microcalcifications) were collected. Surgical outcomes were used to determine upgrade. Four logistic regression models were developed to predict the risk of upgrade. The performance was evaluated using the area under the receiver operating characteristic curve (AUC) and performance scores.

RESULTS

A total of 143 patients with 153 clusters were included. Twelve radiomic features and six clinical factors were selected for model development. The sample was split into 107 training and 46 test cases. Clinical features achieved an AUC of 0.72 (0.60-0.84), radiomic features an AUC of 0.73 (0.61-0.85). Radiomic features with "cluster size" and "age" improved the AUC to 0.79 (0.67-0.91). The best model, incorporating all data, achieved an AUC of 0.82 (0.71-0.92), a specificity of 0.89 (0.75, 0.97), and NPV of 0.92 (0.78-0.98).

CONCLUSION

This study demonstrates the potential of radiomic as a valuable tool for reducing unnecessary treatments for patient classified as "low risk of ADH upgrade". Combining radiomic information with clinical data improved the accuracy of risk prediction.

Breast cancer risk prediction using machine learning: a systematic review

Background

Breast cancer is the leading cause of cancer-related fatalities among women worldwide. Conventional screening and risk prediction models primarily rely on demographic and patient clinical history to devise policies and estimate likelihood. However, recent advancements in artificial intelligence (AI) techniques, particularly deep learning (DL), have shown promise in the development of personalized risk models. These models leverage individual patient information obtained from medical imaging and associated reports. In this systematic review, we thoroughly investigated the existing literature on the application of DL to digital mammography, radiomics, genomics, and clinical information for breast cancer risk assessment. We critically analyzed these studies and discussed their findings, highlighting the promising prospects of DL techniques for breast cancer risk prediction. Additionally, we explored ongoing research initiatives and potential future applications of AI-driven approaches to further improve breast cancer risk prediction, thereby facilitating more effective screening and personalized risk management strategies.

Objective and methods

This study presents a comprehensive overview of imaging and non-imaging features used in breast cancer risk prediction using traditional and AI models. The features reviewed in this study included imaging, radiomics, genomics, and clinical features. Furthermore, this survey systematically presented DL methods developed for breast cancer risk prediction, aiming to be useful for both beginners and advanced-level researchers.

Results

A total of 600 articles were identified, 20 of which met the set criteria and were selected. Parallel benchmarking of DL models, along with natural language processing (NLP) applied to imaging and non-imaging features, could allow clinicians and researchers to gain greater awareness as they consider the clinical deployment or development of new models. This review provides a comprehensive guide for understanding the current status of breast cancer risk assessment using AI.

Conclusion

This study offers investigators a different perspective on the use of AI for breast cancer risk prediction, incorporating numerous imaging and non-imaging features.

Active Surveillance for Atypical Ductal Hyperplasia and Ductal Carcinoma In Situ

Atypical ductal hyperplasia (ADH) and ductal carcinoma in situ (DCIS) are relatively common breast lesions on the same spectrum of disease. Atypical ductal hyperblasia is a nonmalignant, high-risk lesion, and DCIS is a noninvasive malignancy. While a benefit of screening mammography is early cancer detection, it also leads to increased biopsy diagnosis of noninvasive lesions. Previously, treatment guidelines for both entities included surgical excision because of the risk of upgrade to invasive cancer after surgery and risk of progression to invasive cancer for DCIS. However, this universal management approach is not optimal for all patients because most lesions are not upgraded after surgery. Furthermore, some DCIS lesions do not progress to clinically significant invasive cancer. Overtreatment of high-risk lesions and DCIS is considered a burden on patients and clinicians and is a strain on the health care system. Extensive research has identified many potential histologic, clinical, and imaging factors that may predict ADH and DCIS upgrade and thereby help clinicians select which patients should undergo surgery and which may be appropriate for active surveillance (AS) with imaging. Additionally, multiple clinical trials are currently underway to evaluate whether AS for DCIS is feasible for a select group of patients. Recent advances in MRI, artificial intelligence, and molecular markers may also have an important role to play in stratifying patients and delineating best management guidelines. This review article discusses the available evidence regarding the feasibility and limitations of AS for ADH and DCIS, as well as recent advances in patient risk stratification.

Deep Learning Approaches with Digital Mammography for Evaluating Breast Cancer Risk, a Narrative Review

Breast cancer remains the leading cause of cancer-related deaths in women worldwide. Current screening regimens and clinical breast cancer risk assessment models use risk factors such as demographics and patient history to guide policy and assess risk. Applications of artificial intelligence methods (AI) such as deep learning (DL) and convolutional neural networks (CNNs) to evaluate individual patient information and imaging showed promise as personalized risk models. We reviewed the current literature for studies related to deep learning and convolutional neural networks with digital mammography for assessing breast cancer risk. We discussed the literature and examined the ongoing and future applications of deep learning techniques in breast cancer risk modeling.

Deep learning prediction of axillary lymph node status using ultrasound images

OBJECTIVE

To investigate the ability of our convolutional neural network (CNN) to predict axillary lymph node metastasis using primary breast cancer ultrasound (US) images.

METHODS

In this IRB-approved study, 338 US images (two orthogonal images) from 169 patients from 1/2014-12/2016 were used. Suspicious lymph nodes were seen on US and patients subsequently underwent core-biopsy. 64 patients had metastatic lymph nodes. A custom CNN was utilized on 248 US images from 124 patients in the training dataset and tested on 90 US images from 45 patients. The CNN was implemented entirely of 3 × 3 convolutional kernels and linear layers. The 9 convolutional kernels consisted of 6 residual layers, totaling 12 convolutional layers. Feature maps were down-sampled using strided convolutions. Dropout with a 0.5 keep probability and L2 normalization was utilized. Training was implemented by using the Adam optimizer and a final SoftMax score threshold of 0.5 from the average of raw logits from each pixel was used for two class classification (metastasis or not).

RESULTS

Our CNN achieved an AUC of 0.72 (SD ± 0.08) in predicting axillary lymph node metastasis from US images in the testing dataset. The model had an accuracy of 72.6% (SD ± 8.4) with a sensitivity and specificity of 65.5% (SD ± 28.6) and 78.9% (SD ± 15.1) respectively. Our algorithm is available to be shared for research use. (https://github.com/stmutasa/MetUS).

CONCLUSION

It's feasible to predict axillary lymph node metastasis from US images using a deep learning technique. This can potentially aid nodal staging in patients with breast cancer.

Understanding artificial intelligence based radiology studies: CNN architecture

Artificial intelligence (AI) in radiology has gained wide interest due to the development of neural network architectures with high performance in computer vision related tasks. As AI based software programs become more integrated into the clinical workflow, radiologists can benefit from better understanding the principles of artificial intelligence. This series aims to explain basic concepts of AI and its applications in medical imaging. In this article, we will review the background of neural network architecture and its application in imaging analysis.

Diagnostic value of radiomics and machine learning with dynamic contrast-enhanced magnetic resonance imaging for patients with atypical ductal hyperplasia in predicting malignant upgrade

Purpose

To investigate whether radiomics features extracted from magnetic resonance imaging (MRI) of patients with biopsy-proven atypical ductal hyperplasia (ADH) coupled with machine learning can differentiate high-risk lesions that will upgrade to malignancy at surgery from those that will not, and to determine if qualitatively and semi-quantitatively assessed imaging features, clinical factors, and image-guided biopsy technical factors are associated with upgrade rate.

Methods

This retrospective study included 127 patients with 139 breast lesions yielding ADH at biopsy who were assessed with multiparametric MRI prior to biopsy. Two radiologists assessed all lesions independently and with a third reader in consensus according to the BI-RADS lexicon. Univariate analysis and multivariate modeling were performed to identify significant radiomic features to be included in a machine learning model to discriminate between lesions that upgraded to malignancy on surgery from those that did not.

Results

Of 139 lesions, 28 were upgraded to malignancy at surgery, while 111 were not upgraded. Diagnostic accuracy was 53.6%, specificity 79.2%, and sensitivity 15.3% for the model developed from pre-contrast features, and 60.7%, 86%, and 22.8% for the model developed from delta radiomics datasets. No significant associations were found between any radiologist-assessed lesion parameters and upgrade status. There was a significant correlation between the number of specimens sampled during biopsy and upgrade status (p = 0.003).

Conclusion

Radiomics analysis coupled with machine learning did not predict upgrade status of ADH. The only significant result from this analysis is between the number of specimens sampled during biopsy procedure and upgrade status at surgery.

Dynamic Changes of Convolutional Neural Network-based Mammographic Breast Cancer Risk Score Among Women Undergoing Chemoprevention Treatment

INTRODUCTION

We investigated whether our convolutional neural network (CNN)-based breast cancer risk model is modifiable by testing it on women who had undergone risk-reducing chemoprevention treatment.

MATERIALS AND METHODS

We conducted a retrospective cohort study of patients diagnosed with atypical hyperplasia, lobular carcinoma in situ, or ductal carcinoma in situ at our institution from 2007 to 2015. The clinical characteristics, chemoprevention use, and mammography images were extracted from the electronic health records. We classified two groups according to chemoprevention use. Mammograms were performed at baseline and subsequent follow-up evaluations for input to our CNN risk model. The 2 chemoprevention groups were compared for the risk score change from baseline to follow-up. The change categories included stayed high risk, stayed low risk, increased from low to high risk, and decreased from high to low risk. Unordered polytomous regression models were used for statistical analysis, with P < .05 considered statistically significant.

RESULTS

Of 541 patients, 184 (34%) had undergone chemoprevention treatment (group 1) and 357 (66%) had not (group 2). Using our CNN breast cancer risk score, significantly more women in group 1 had shown a decrease in breast cancer risk compared with group 2 (33.7% vs. 22.9%; P < .01). Significantly fewer women in group 1 had an increase in breast cancer risk compared with group 2 (11.4% vs. 20.2%; P < .01). On multivariate analysis, an increase in breast cancer risk predicted by our model correlated negatively with the use of chemoprevention treatment (P = .02).

CONCLUSIONS

Our CNN-based breast cancer risk score is modifiable with potential utility in assessing the efficacy of known chemoprevention agents and testing new chemoprevention strategies.

Understanding artificial intelligence based radiology studies: What is overfitting?

Artificial intelligence (AI) is a broad umbrella term used to encompass a wide variety of subfields dedicated to creating algorithms to perform tasks that mimic human intelligence. As AI development grows closer to clinical integration, radiologists will need to become familiar with the principles of artificial intelligence to properly evaluate and use this powerful tool. This series aims to explain certain basic concepts of artificial intelligence, and their applications in medical imaging starting with a concept of overfitting.

Prospective Analysis Using a Novel CNN Algorithm to Distinguish Atypical Ductal Hyperplasia From Ductal Carcinoma in Situ in Breast

INTRODUCTION

We previously developed a convolutional neural networks (CNN)-based algorithm to distinguish atypical ductal hyperplasia (ADH) from ductal carcinoma in situ (DCIS) using a mammographic dataset. The purpose of this study is to further validate our CNN algorithm by prospectively analyzing an unseen new dataset to evaluate the diagnostic performance of our algorithm.

MATERIALS AND METHODS

In this institutional review board-approved study, a new dataset composed of 280 unique mammographic images from 140 patients was used to test our CNN algorithm. All patients underwent stereotactic-guided biopsy of calcifications and underwent surgical excision with available final pathology. The ADH group consisted of 122 images from 61 patients with the highest pathology diagnosis of ADH. The DCIS group consisted of 158 images from 79 patients with the highest pathology diagnosis of DCIS. Two standard mammographic magnification views (craniocaudal and mediolateral/lateromedial) of the calcifications were used for analysis. Calcifications were segmented using an open source software platform 3D slicer and resized to fit a 128 × 128 pixel bounding box. Our previously developed CNN algorithm was used. Briefly, a 15 hidden layer topology was used. The network architecture contained 5 residual layers and dropout of 0.25 after each convolution. Diagnostic performance metrics were analyzed including sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve. The "positive class" was defined as the pure ADH group in this study and thus specificity represents minimizing the amount of falsely labeled pure ADH cases.

RESULTS

Area under the receiver operating characteristic curve was 0.90 (95% confidence interval, ± 0.04). Diagnostic accuracy, sensitivity, and specificity was 80.7%, 63.9%, and 93.7%, respectively.

CONCLUSION

Prospectively tested on new unseen data, our CNN algorithm distinguished pure ADH from DCIS using mammographic images with high specificity.

Upgrade Rate of Percutaneously Diagnosed Pure Atypical Ductal Hyperplasia: Systematic Review and Meta-Analysis of 6458 Lesions

Background Management of percutaneously diagnosed pure atypical ductal hyperplasia (ADH) is an unresolved clinical issue. Purpose To calculate the pooled upgrade rate of percutaneously diagnosed pure ADH. Materials and Methods A search of MEDLINE and EMBASE databases was performed in October 2018. Preferred Reporting Items for Systematic Reviews and Meta-Analyses, or PRISMA, guidelines were followed. A fixed- or random-effects model was used, along with subgroup and meta-regression analyses. The Newcastle-Ottawa scale was used for study quality, and the Egger test was used for publication bias. Results Of 521 articles, 93 were analyzed, providing data for 6458 ADHs (5911 were managed with surgical excision and 547 with follow-up). Twenty-four studies used core-needle biopsy; 44, vacuum-assisted biopsy; 21, both core-needle and vacuum-assisted biopsy; and four, unspecified techniques. Biopsy was performed with stereotactic guidance in 29 studies; with US guidance in nine, with MRI guidance in nine, and with mixed guidance in eight. Overall heterogeneity was high (I² = 80%). Subgroup analysis according to management yielded a pooled upgrade rate of 29% (95% confidence interval [CI]: 26%, 32%) for surgically excised lesions and 5% (95% CI: 4%, 8%) for lesions managed with follow-up (P < .001). Heterogeneity was entirely associated with surgically excised lesions (I² = 78%) rather than those managed with follow-up (I² = 0%). Most variability was explained by guidance and needle caliper (P = .15). At subgroup analysis of surgically excised lesions, the pooled upgrade rate was 42% (95% CI: 31%, 53%) for US guidance, 23% (95% CI: 19%, 27%) for stereotactic biopsy, and 32% (95% CI: 22%, 43%) for MRI guidance, with heterogeneity (52%, 63%, and 56%, respectively) still showing the effect of needle caliper. When the authors considered patients with apparent complete lesion removal after biopsy (subgroups in 14 studies), the pooled upgrade rate was 14% (95% CI: 8%, 23%). Study quality was low to medium; the risk of publication bias was low (P = .10). Conclusion Because of a pooled upgrade rate higher than 2% (independent of biopsy technique, needle size, imaging guidance, and apparent complete lesion removal), atypical ductal hyperplasia diagnosed with percutaneous needle biopsy should be managed with surgical excision. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Brem in this issue.

Assessment of Machine Learning of Breast Pathology Structures for Automated Differentiation of Breast Cancer and High-Risk Proliferative Lesions

IMPORTANCE

Following recent US Food and Drug Administration approval, adoption of whole slide imaging in clinical settings may be imminent, and diagnostic accuracy, particularly among challenging breast biopsy specimens, may benefit from computerized diagnostic support tools.

OBJECTIVE

To develop and evaluate computer vision methods to assist pathologists in diagnosing the full spectrum of breast biopsy samples, from benign to invasive cancer.

DESIGN, SETTING, AND PARTICIPANTS

In this diagnostic study, 240 breast biopsies from Breast Cancer Surveillance Consortium registries that varied by breast density, diagnosis, patient age, and biopsy type were selected, reviewed, and categorized by 3 expert pathologists as benign, atypia, ductal carcinoma in situ (DCIS), and invasive cancer. The atypia and DCIS cases were oversampled to increase statistical power. High-resolution digital slide images were obtained, and 2 automated image features (tissue distribution feature and structure feature) were developed and evaluated according to the consensus diagnosis of the expert panel. The performance of the automated image analysis methods was compared with independent interpretations from 87 practicing US pathologists. Data analysis was performed between February 2017 and February 2019.

MAIN OUTCOMES AND MEASURES

Diagnostic accuracy defined by consensus reference standard of 3 experienced breast pathologists.

RESULTS

The accuracy of machine learning tissue distribution features, structure features, and pathologists for classification of invasive cancer vs noninvasive cancer was 0.94, 0.91, and 0.98, respectively; the accuracy of classification of atypia and DCIS vs benign tissue was 0.70, 0.70, and 0.81, respectively; and the accuracy of classification of DCIS vs atypia was 0.83, 0.85, and 0.80, respectively. The sensitivity of both machine learning features was lower than that of the pathologists for the invasive vs noninvasive classification (tissue distribution feature, 0.70; structure feature, 0.49; pathologists, 0.84) but higher for the classification of atypia and DCIS vs benign cases (tissue distribution feature, 0.79; structure feature, 0.85; pathologists, 0.72) and the classification of DCIS vs atypia (tissue distribution feature, 0.88; structure feature, 0.89; pathologists, 0.70). For the DCIS vs atypia classification, the specificity of the machine learning feature classification was similar to that of the pathologists (tissue distribution feature, 0.78; structure feature, 0.80; pathologists, 0.82).

CONCLUSION AND RELEVANCE

The computer-based automated approach to interpreting breast pathology showed promise, especially as a diagnostic aid in differentiating DCIS from atypical hyperplasia.