Prediction of reader estimates of mammographic density using convolutional neural networks

Abstract

Mammographic density is an important risk factor for breast cancer. In recent research, percentage density assessed visually using visual analogue scales (VAS) showed stronger risk prediction than existing automated density measures, suggesting readers may recognize relevant image features not yet captured by hand-crafted algorithms. With deep learning, it may be possible to encapsulate this knowledge in an automatic method. We have built convolutional neural networks (CNN) to predict density VAS scores from full-field digital mammograms. The CNNs are trained using whole-image mammograms, each labeled with the average VAS score of two independent readers. Each CNN learns a mapping between mammographic appearance and VAS score so that at test time, they can predict VAS score for an unseen image. Networks were trained using 67,520 mammographic images from 16,968 women and for model selection we used a dataset of 73,128 images. Two case-control sets of contralateral mammograms of screen detected cancers and prior images of women with cancers detected subsequently, matched to controls on age, menopausal status, parity, HRT and BMI, were used for evaluating performance on breast cancer prediction. In the case-control sets, odd ratios of cancer in the highest versus lowest quintile of percentage density were 2.49 (95% CI: 1.59 to 3.96) for screendetected cancers and 4.16 (2.53 to 6.82) for priors, with matched concordance indices of 0.587 (0.542 to 0.627) and 0.616 (0.578 to 0.655), respectively. There was no significant difference between reader VAS and predicted VAS for the prior test set (likelihood ratio chi square, p ¼ 0.134). Our fully automated method shows promising results for cancer risk prediction and is comparable with human performance.

Mammographic density change in a cohort of premenopausal women receiving tamoxifen for breast cancer prevention over 5 years

“…Stratus includes an algorithm to align images so that density change is more stable. NN-VAS was trained on visual assessment of raw unprocessed images from the PROCAS study [22], which was from the same Manchester setting as the present study. Densitas uses processed DICOM files.…”

“…Four fully objective measures of breast density were computed on each of the images. They were (i) NN-VAS (neural network fitted to percentage density on visual assessment scale) version 1.0 [22], (ii) Volpara volumetric density version 1.5.2 [23], (iii) Stratus [24] and (iv) Densitas version 2.0.0 [20]. The mean of each density measure over all views and both breasts was the single measure used at each time point.…”

An Intelligent Decision-Making Support System for the Detection and Staging of Prostate Cancer in Developing Countries

“…The weight of the decision-making stage is unreasonable. Georgia et al proposed a convolutional neural network- (CNN-) based intelligent diagnosis system for breast cancer diagnosis and related processing, through the related image processing, with images as input, diagnose breast cancer, and classify breast cancer histopathology images [ 25 – 27 ].…”

Artificial intelligence and convolution neural networks assessing mammographic images: a narrative literature review

“…Studies which used imaging modalities other than mammography were excluded. A total of 33 relevant peer‐reviewed articles were found . To compare the results among studies, relevant features were extracted: (1) author and publication year; (2) study aim; (3) utilised database; (4) number of images used; and (5) performance measures including sensitivity, specificity and area under the curve (AUC).…”