Mammography

Mammography

Transcript

1. Development and Validation of Deep Learning Algorithms for Diagnosing Breast Cancers in Mammograms. [1] Department of Diagnostic and Interventional Radiology, Osaka City University, Graduate School of Medicine [2] Department of Gastroenterology, Osaka City University, Graduate School of Medicine [3] Department of Premier Preventive Medicine, Osaka City University, Graduate School of Medicine [4] Department of Surgical Oncology, Osaka City University, Graduate School of Medicine [5] YEBIS.XYZ Inc., [6] tech vein Inc., [7] Kaeru Inc. Daiju Ueda, Akira Yamamoto, Masatsugu Shiba, Shinya Fukumoto, Satoru Noda, Naoyoshi Onoda, Akitoshi Shimazaki, Hiroyuki Tatekawa, Shinichi Tsutsumi, Yoshikazu Hashimoto, Mitsuhiro Inomata, Hiroko Osaki, Yukio Miki [1] [1] [2] [3] [4] [1] [1] [7] [6] [5] [1] [4] [1]
2. Deep learning is inspired by structures and functions of a brain. Simple layer perceptron All-or-Nothing Law x x y 1 2 1 2 w w y = x w + x w 2 2 1 1 h(x) = 0 (y ≦ 0) 1 (y > 0) {
3. Deep learning is inspired by structures and functions of a brain. Multi layers perceptron Deep learning
4. Accuracy [%] 0 20 40 60 80 100 Year 2010 2011 2012 2013 2014 2015 ImageNet: 1000 categories recognition State-of-art technique Deep Learning Traditional Method
5. Results: Accuracies Architecture Epochs Top train-data accuracy [%] Top validation-data accuracy [%] InceptionV3 211 94 92 ResNet152 186 90 91 VGG16 173 88 89
6. Results: ROCs Sensitivity : 95% Specificity : 90.1% True positive rate False positive rate 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 AUC: 0.95
7. Results: Head map (Grad-CAM) Mass Calcification [1] [1] Selvaraju RR, Cogswell M, Das A, Vedantam R, Parikh D, Batra D. Grad-cam: Visual explanations from deep networks via gradient- based localization. See https://arxiv org/abs/161002391 v3. 2016;7(8).
8. [1] Lévy D, Jain A. Breast mass classification from mammograms using deep convolutional neural networks. arXiv preprint arXiv:161200542. 2016. [2] Ribli D, Horváth A, Unger Z, Pollner P, Csabai I. Detecting and classifying lesions in mammograms with Deep Learning. Sci Rep. 2018;8(1):4165. [3] de Lima SM, da Silva-Filho AG, dos Santos WP. Detection and classification of masses in mammographic images in a multi-kernel approach. Comput Methods Programs Biomed. 2016;134:11-29. [4] Lotter W, Sorensen G, Cox D. A multi-scale cnn and curriculum learning strategy for mammogram classification. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: Springer, 2017; p. 169-77. [5] Jadoon MM, Zhang Q, Haq IU, Butt S, Jadoon A. Three-class mammogram classification based on descriptive CNN features. BioMed research international. 2017;2017. No Accuracy [%] AUC Comment [1] 0.93 Mass images only [2] 0.95, 0.85 [3] 0.94 Mass images only [4] 0.92 [5] 0.84 This study 0.94 0.95 Discussion Comparison to prior Research