Dr. Lawrence W.C. Chan is currently Associate Professor, Department of Health Technology and Informatics, Hong Kong Polytechnic University; and Guest Professor, Institute of Nuclear Medicine and Molecular Imaging, Zhejiang University. Dr. Chan received his PhD in Artificial Intelligence in 2001 from the University of Hong Kong. His services to professional bodies include but are not limited to (1) Founder, Biomedical Big Data Laboratory, PolyU, 2017-2022; (2) Editorial Board, Genomics, Proteomics & Bioinformatics, 2020-2022; (3) Expert Panel Member, HKSTP, 2019-2021; (4) Affiliate member, Hong Kong Society of Medical Informatics (HKSMI). Dr. Chan’s research interest covers bioinformatics, imaging informatics and clinical decision support. Dr. Chan received substantial amount of external research grants, including CRF, GRF, HMRF and ITF in PI, Co-PI or Co-I capacity.
Hepatocellular carcinoma (HCC) represents a major primary liver cancer type where the poor prognosis is often attributed by its metastasis. The prediction of lymph node metastasis (LNM) can assist the management and thus better treatment outcomes. Radiomics analysis of medical images, which quantifies, chooses, and identifies the radiomics features of the region of interest in high throughput manner, can be used for predicting the lymph node status.
This study hypothesizes that the HCC metastasis can be characterized and predicted by the radiomic features derived from the primary tumours in contrast enhanced computed tomography (CECT) images. A dataset of 400 abdominal CECT examinations covering the whole liver was obtained from the image repository hosted by Department of Diagnostic Radiology, University of Hong Kong.
According to the de-identified patient records, the HCC tumours were located and the cases with LNM (positive) were confirmed. The dataset was split into training and test sets, which were used for training the U-Net for liver segmentation, tumour detection and deep radiomic feature extraction. Through statistical analysis, a radiomics model was built for precisely predicting the LNM and assisting the clinical diagnosis and management.