HCC-MPDS Multimodal Deep Learning Prognostic Model for Resectable HCC

Multimodal Hybrid Deep Learning-based Prognostic Decision System to Guide Adjuvant Therapy in Hepatocellular Carcinoma After Surgical Resection

This is a completed multi-center retrospective observational study focusing on patients who received curative surgical resection for hepatocellular carcinoma (HCC). We collected anonymized historical clinical data, preoperative gadoxetic acid-enhanced MRI scans and digital pathological whole-slide images from three Chinese medical centers and the public TCGA-LIHC database. We developed a multimodal hybrid deep learning prognostic model (named HCC-MPDS) integrating multi-source medical data to stratify HCC patients into high-risk and low-risk recurrence groups after surgery. The primary goal of this study is to compare postoperative recurrence-free survival (RFS) and overall survival (OS) between the two risk subgroups, and evaluate whether this artificial intelligence prediction system can help clinicians identify patients who may benefit from postoperative adjuvant therapy. No new experimental drugs or interventional treatments were provided to participants; all data were retrospectively extracted from archived medical records under institutional ethics approval, with waived written informed consent for de-identified historical samples and images.

Inclusion Criteria:

1. patients with pathologically confirmed hepatocellular carcinoma (HCC);
2. patients accepting curative surgical resection (SR) without preoperative adjuvant treatment;
3. MRI were performed two weeks before surgery.

Exclusion Criteria:

1. incomplete clinical information;
2. incomplete or suboptimal quality of preoperative MRI imaging data;
3. unavailable or poor-quality pathological whole slide images (WSIs);
4. incomplete follow-up data.