生醫電資所教師研究亮點-110年10月份「張瑞峰教授」

本所資訊組合聘教授「張瑞峰」帶領研究團隊

致力於醫學影像領域上應用深度學習技術

開發醫療影像自動化電腦輔助偵測與診斷系統

近年團隊學術研究亮點著重於

「全自動乳房超音波乳癌偵測與診斷系統」

本系統具有高檢測率、高診斷率、低偽陽性、並兼具強韌度(可偵測1公分以下的乳癌),並採用一次性檢查設計,1秒內即可完成一個全自動乳房超音波(300張影像)的乳癌檢閱程序,精確定位乳癌位置及顯示區域,同時進行診斷大幅加快閱片時間。此外,乳癌診斷準確度高達89.2%,在檢測上具有高度臨床價值。

In response to the Ministry of Science and Technology’s promotion about AI innovation research projects and focusing on AI core technologies.Professor “Ruey-Feng, Chang ” leads the research and development team (R&D team) and has dedicated to research the application of deep learning technology in the field of medical imaging.His team develops the automated computer-aided diagnosis and diagnosis system for medical imaging. The highlights of the team’s academic research in recent years have focused on:

ABUS Cancer Computer-aided Detection and Diagnosis SystemThe “ABUS Cancer Computer-aided Detection and Diagnosis System” is a breast cancer detection and diagnosis system designed with the artificial intelligence technology. is developed to locate the breast tumor in automatic whole breast ultrasound image (ABUS). Through the one-take and the one-stage detection architecture, our CADe can completely review an ABUS image within 1 second. Our system has properties of high detection rate, high diagnosis rate, low false positive (FP), robustness (tumor size < 1 cm), and high diagnosis accuracy. In detection, it has only 2 FPs in 95% sensitivity, and significantly reduces the detection time, which indicates the system is of great worth in clinical breast examination. Moreover, in breast cancer diagnosis, the accuracy rate of our system is up to 89.2%, and has high value in clinic.

「多疾病胸部X光電腦輔助診斷系統」

胸部X光電腦輔助診斷系統,能診斷出14種常見的胸腔疾病,包括肺不張、實變、浸潤、氣胸、水腫、氣腫、纖維化、積液、肺炎、胸膜增厚、心臟肥大、結節、腫塊和疝氣。高效圖像表徵提取器與增強的疾病關聯性資訊幫助X光影像分類不同胸腔疾病。本系統比最先進的 (SOTA) 技術擁有更好的效能(平均AUC 可達 0.8266),減少了閱片者間變異,客觀地提供診斷結果。

Multi-disease Thorax X-ray CADx system

The chest X-ray is one of the most common imaging examinations in clinical diagnosis on the lung. Our thorax X-ray CADx system is useful for diagnosing different thorax diseases, which has excellent diagnostic accuracy and reduces viewer variances.Four advantages of the Thorax X-ray CADx system:

  1. 14 thorax diseases diagnosis in one CADx system

Atelectasis, Consolidation, Infiltration, Pneumothorax, Edema, Emphysema, Fibrosis, Effusion, Pneumonia, Pleural thickening, Cardiomegaly, Nodule, Mass, and Hernia included.

  1. Efficient image representation feature extractor

Extract representative image information.

  1. Enhance the correlation between different thorax diseases

Fibrosis → Atelectasis, Pleural Effusion → Atelectasis.

  1. Accurate diagnostic performance

Better diagnostic performance than SOTA (mean AUC = 0.8266).