Weakly-supervised deep learning for ultrasound diagnosis of breast cancer

• Published in: Scientific reports Vol. 11; no. 1; p. 24382
• Main Authors: Kim, Jaeil, Kim, Hye Jung, Kim, Chanho, Lee, Jin Hwa, Kim, Keum Won, Park, Young Mi, Kim, Hye Won, Ki, So Yeon, Kim, You Me, Kim, Won Hwa
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 21.12.2021; Nature Publishing Group; Nature Portfolio
• Subjects: 631/67; 631/67/1347; 692/4028; 692/700; 692/700/1421/1860; Adolescent; Adult; Aged; Aged, 80 and over; Algorithms; Annotations; Automation; Breast - diagnostic imaging; Breast - pathology; Breast cancer; Breast Neoplasms - diagnosis; Breast Neoplasms - diagnostic imaging; Case-Control Studies; Deep Learning; Diagnosis, Differential; Differential diagnosis; Female; Follow-Up Studies; Humanities and Social Sciences; Humans; Image Interpretation, Computer-Assisted - methods; Localization; Middle Aged; multidisciplinary; Neural Networks, Computer; Prognosis; Retrospective Studies; ROC Curve; Science; Science (multidisciplinary); Ultrasonic imaging; Ultrasonography, Mammary - methods; Ultrasound; Young Adult; United States > US
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-021-03806-7

Conventional deep learning (DL) algorithm requires full supervision of annotating the region of interest (ROI) that is laborious and often biased. We aimed to develop a weakly-supervised DL algorithm that diagnosis breast cancer at ultrasound without image annotation. Weakly-supervised DL algorithms were implemented with three networks (VGG16, ResNet34, and GoogLeNet) and trained using 1000 unannotated US images (500 benign and 500 malignant masses). Two sets of 200 images (100 benign and 100 malignant masses) were used for internal and external validation sets. For comparison with fully-supervised algorithms, ROI annotation was performed manually and automatically. Diagnostic performances were calculated as the area under the receiver operating characteristic curve (AUC). Using the class activation map, we determined how accurately the weakly-supervised DL algorithms localized the breast masses. For internal validation sets, the weakly-supervised DL algorithms achieved excellent diagnostic performances, with AUC values of 0.92–0.96, which were not statistically different (all P s > 0.05) from those of fully-supervised DL algorithms with either manual or automated ROI annotation (AUC, 0.92–0.96). For external validation sets, the weakly-supervised DL algorithms achieved AUC values of 0.86–0.90, which were not statistically different ( P s > 0.05) or higher ( P  = 0.04, VGG16 with automated ROI annotation) from those of fully-supervised DL algorithms (AUC, 0.84–0.92). In internal and external validation sets, weakly-supervised algorithms could localize 100% of malignant masses, except for ResNet34 (98%). The weakly-supervised DL algorithms developed in the present study were feasible for US diagnosis of breast cancer with well-performing localization and differential diagnosis.