Deep Learning
Diagnosis of thyroid nodules on ultrasonography by a deep convolutional neural network
SERA (Severance Diagnostic Helper based on Deeplearning designed by Yonsei-CSE)
Training: 2004년부터 2019년까지 세브란스병원에서 수집한 13,560장의 갑상선 ROI 영상
Test – Multicenter study
Internal test: 634 세브란스 test set
External test: 781 삼성의료원, 200 분당차병원, 200 경희대병원
Two individual CNNs and two classification ensembles (CNNE1 and CNNE2) were tested to differentiate malignant and benign thyroid nodules. CNNs demonstrated high area under the curves (AUCs) to diagnose malignant thyroid nodules (0.898-0.937 for the internal test set and 0.821-0.885 for the external test sets).
* 모든 환자데이터는 각 의료기관의 IRB 승인을 받아 진행
Reference: Jieun Koh, Eunjung Lee, Kyunghwa Han, Eun‑Kyung Kim, Eun Ju Son, Yu‑Mee Sohn, Mirinae Seo, Mi‑ri Kwon, Jung Hyun Yoon, Jin Hwa Lee, Young Mi Park, Sungwon Kim, Jung Hee Shin, Jin Young Kwak, Diagnosis of thyroid nodules on ultrasonography by a deep convolutional neural network, Scientific Reports, 10(1) (2020 Sep) 15245
특허 : 등록특허 10-2209382 병변 진단에 대한 정보 제공 방법 및 이를 이용한 병변 진단에 대한 정보 제공용 디바이스
Feature Concatenation
The features extracted from deeper layer are compressive, so discriminative information may be missed. Examine the features extracted from different layers or CNNs in various combinations.
Classification Ensemble
For extracted features, several classifiers are considered such as SVM and random forests(RF). To obtain complementary result, apply ensemble method to each classifier result.
Reference: Eunjung Lee, Heonkyu Ha, Hye Jung Kim, Hee Jung Moon, Jung Hee Byon, Sun Huh, Jinwoo Son, Jiyoung Yoon, Kyunghwa Han, Jin Young Kwak, Differentiation of thyroid nodules on US using features learned and extracted from various convolutional neural networks, Scientific Reports, 9 (2019 December) 19854-1-19854-11