A Deep Learning Mammography-based Model for Improved Breast Cancer Risk Prediction

Abstract

Background: Mammographic density improves the accuracy of breast cancer risk models. However, the use of breast density is limited by subjective assessment, variation across radiologists, and restricted data. A mammography-based deep learning (DL) model may provide more accurate risk prediction. Purpose:To develop a mammography-based DL breast cancer risk model that is more accurate than established clinical breast cancer risk models. Materials and Methods:This retrospective study included 88 994 consecutive screening mammograms in 39 571 women between January 1, 2009, and December 31, 2012. For each patient, all examinations were assigned to either training, validation, or test sets, resulting in 71 689, 8554, and 8751 examinations, respectively. Cancer outcomes were obtained through linkage to a regional tumor registry. By using risk factor information from patient questionnaires and electronic medical records review, three models were developed to assess breast cancer risk within 5 years: a risk-factor-based logistic regression model (RF-LR) that used traditional risk factors, a DL model (image-only DL) that used mammograms alone, and a hybrid DL model that used both traditional risk factors and mammograms. Comparisons were made to an established breast cancer risk model that included breast density version 8 [TC]). Model performance was compared by using areas under the receiver operating characteristic curve (AUCs) with DeLong test (P , .05). Results:The test set included 3937 women, aged 56.20 years 6 10.04. Hybrid DL and image-only DL showed AUCs of 0.70 (95% confidence interval [CI]: 0.66, 0.75) and 0.68 (95% CI: 0.64, 0.73), respectively. RF-LR and TC showed AUCs of 0.67 (95% CI: 0.62, 0.72) and 0.62 (95% CI: 0.57, 0.66), respectively. Hybrid DL showed a significantly higher AUC (0.70) than TC (0.62; P , .001) and RF-LR (0.67; P = .01). Conclusion:Deep learning models that use full-field mammograms yield substantially improved risk discrimination compared with the Tyrer-Cuzick (version 8) model.

Going Beyond Conventional Mammographic Density to Discover Novel Mammogram-Based Predictors of Breast Cancer Risk

“…Another attempt to achieve this aim using 'deep learning' applied to a large data set [32] failed to use cases and controls of the same age, so they in effect learnt how to measure age from a mammogram, derived inflated estimates of risk discrimination based on naïve use of AUC statistics, and might not have discovered anything new. We know age predicts breast cancer risk, and is easily measured, so we want to know what it is about a mammogram that predicts risk for women of the same age.…”

Cancer Prevention Using Machine Learning, Nudge Theory and Social Impact Bond

“…Machine learning models are being intensively developed for the diagnosis phase to deduce and detect cancer from various available sources of medical images. In particular, the following examples were highly successful: A better prediction of breast cancer with a deep learning model combined with a logistic regression model (area under the curve (AUC) = 0.70 vs. 0.62, respectively) in which the image data and assessment records of over 39,571 subjects were combined [12]; the detection of early stage stomach cancer using a convolutional neural network (CNN, detection rate = 82.8%) [13]; the deduction of the risk level of lung cancer using a CNN model (AUC = 0.94) [14]. Regarding the treatment phase, machine learning is most used in cancer therapeutics.…”

“…Health data can be broadly organized into four categories: Image, examination data including medical insurance receipts, genetic information, and text, such as articles/reports, dialogues, and medical interviews. As an example, image data is currently utilized to perform imaging-based diagnosis [12],genetic data is used to prescribe therapeutic regimens for cancer treatment, and text data is used for chat-based automated diagnosis [10]. The examination data that the present study focuses on is being proprietarily used to predict the worsening of conditions and the onset of diseases [28].…”

Assessing Cancer Risk from Mammograms: Deep Learning Is Superior to Conventional Risk Models

“…he last few years have produced enormous growth in the radiologic application of computer vision deep learning (DL) algorithms and machine learning, often referred to as artificial intelligence (1). In this issue of Radiology, Yala et al (2) demonstrated the effectiveness of DL methods in assessing breast cancer risk by using clinical data, breast density scores, and mammograms. On March 28, 2019, the U.S. Food and Drug Administration announced a proposed rule (3) to update the landmark policy passed by Congress in 1992 to ensure quality of mammography for early breast cancer detection (known as the Mammography Quality Standards Act).…”

“…Yala et al (2) found that DL risk models (ie, the third and fourth models) consistently performed better than the first and second models for prediction of breast cancer within 5 years after mammography. The hybrid DL model was the best and the clinical data model was the worst.…”