A novel multimodal deep learning model for preoperative prediction of microvascular invasion and outcome in hepatocellular carcinoma. Issue 1 (January 2023)

• Record Type: Journal Article
• Authors: Wang, Fang; Chen, Qingqing; Chen, Yinan; Zhu, Yajing; Zhang, Yuanyuan; Cao, Dan; Zhou, Wei; Liang, Xiao; Yang, Yunjun; Lin, Lanfen; Hu, Hongjie
• Is Part Of: European journal of surgical oncology. Volume 49:Issue 1(2023)
• Page Start: 156
• Page End: 164
• Publication Date: 2023-01
• Subjects: Hepatocellular carcinoma -- Microvascular invasion -- Multimodal -- Deep learning; Oncology -- Periodicals; Cancer -- Surgery -- Periodicals; Medical Oncology -- Periodicals; Neoplasms -- surgery -- Periodicals; Cancer -- Chirurgie -- Périodiques; Cancérologie -- Périodiques; Oncologie; Chirurgie (geneeskunde); Electronic journals; Electronic journals -- Sciences; Electronic journals -- Medicine; Electronic journals; 616.994059005
• DOI: 10.1016/j.ejso.2022.08.036
• Languages: English
• ISSNs: 0748-7983

Abstract

Abstract: Background: Accurate preoperative identification of the microvascular invasion (MVI) can relieve the pressure from personalized treatment adaptation and improve the poor prognosis for hepatocellular carcinoma (HCC). This study aimed to develop and validate a novel multimodal deep learning (DL) model for predicting MVI based on multi-parameter magnetic resonance imaging (MRI) and contrast-enhanced computed tomography (CT). Methods: A total of 397 HCC patients underwent both CT and MRI examinations before surgery. We established the radiological models (R CT, R MRI ) by support vector machine (SVM), DL models (DL CT_ALL, DL MRI_ALL, DL CT + MRI ) by ResNet18. The comprehensive model (C ALL ) involving multi-modality DL features and clinical and radiological features was constructed using SVM. Model performance was quantified by the area under the receiver operating characteristic curve (AUC) and compared by net reclassification index (NRI) and integrated discrimination improvement (IDI). Results: The DL CT + MRI model exhibited superior predicted efficiency over single-modality models, especially over the DL CT_ALL model (AUC: 0.819 vs. 0.742, NRI > 0, IDI > 0). The DL MRI_ALL model improved the performance over the R MRI model (AUC: 0.794 vs. 0.766, NRI > 0, IDI < 0), but no such difference was found between the DL CT_ALL model and R CT model (AUC: 0.742 vs. 0.710, NRI < 0, IDI < 0). Furthermore, both the DL CT + MRI and C ALL models revealed the prognostic power in recurrence-free survival stratification ( P < 0.001). Conclusion: The proposed DL CT + MRI model showed robust capability in predicting MVI and outcomes for HCC. Besides, the identification ability of the multi-modality DL model was better than any single modality, especially for CT.