Deep transfer learning for detection of breast arterial calcifications on mammograms: a comparative study

<p>Introduction Breast arterial calcifications (BAC) are common incidental findings on routine mammograms, which<br>have been suggested as a sex-specific biomarker of cardiovascular disease (CVD) risk. Previous work showed the efficacy<br>of a pretrained convolutional network (CNN), VCG16, for automatic BAC detection. In this study, we further tested<br>the method by a comparative analysis with other ten CNNs.<br>Material and methods Four-view standard mammography exams from 1,493 women were included in this retrospective<br>study and labeled as BAC or non-BAC by experts. The comparative study was conducted using eleven<br>pretrained convolutional networks (CNNs) with varying depths from five architectures including Xception, VGG,<br>ResNetV2, MobileNet, and DenseNet, fine-tuned for the binary BAC classification task. Performance evaluation<br>involved area under the receiver operating characteristics curve (AUC-ROC) analysis, F1-score (harmonic mean of precision<br>and recall), and generalized gradient-weighted class activation mapping (Grad-CAM++) for visual explanations.<br>Results The dataset exhibited a BAC prevalence of 194/1,493 women (13.0%) and 581/5,972 images (9.7%). Among<br>the retrained models, VGG, MobileNet, and DenseNet demonstrated the most promising results, achieving AUCROCs<br>&gt; 0.70 in both training and independent testing subsets. In terms of testing F1-score, VGG16 ranked first, higher<br>than MobileNet (0.51) and VGG19 (0.46). Qualitative analysis showed that the Grad-CAM++ heatmaps generated<br>by VGG16 consistently outperformed those produced by others, offering a finer-grained and discriminative localization<br>of calcified regions within images.<br>Conclusion Deep transfer learning showed promise in automated BAC detection on mammograms, where relatively<br>shallow networks demonstrated superior performances requiring shorter training times and reduced resources.<br>Relevance statement Deep transfer learning is a promising approach to enhance reporting BAC on mammograms<br>and facilitate developing efficient tools for cardiovascular risk stratification in women, leveraging large-scale mammographic<br>screening programs.<br>Key points<br>&bull; We tested different pretrained convolutional networks (CNNs) for BAC detection on mammograms.<br>&bull; VGG and MobileNet demonstrated promising performances, outperforming their deeper, more complex<br>counterparts.<br>&bull; Visual explanations using Grad-CAM++ highlighted VGG16&rsquo;s superior performance in localizing BAC.</p>