MRI radiomics based on deep learning automated segmentation to predict early recurrence of hepatocellular carcinoma

Hong Wei ¹, Tianying Zheng ¹, Xiaolan Zhang ²

⁰Department of Radiology, Functional and Molecular Imaging Key Laboratory of Sichuan Province, West China Hospital, Sichuan University, Chengdu, Sichuan, 610041, China.

Insights Into Imaging +

|

May 19, 2024

View abstract on PubMed

Summary

Deep learning (DL) models using MRI radiomics can predict early hepatocellular carcinoma (HCC) recurrence after surgery. This hybrid approach outperforms current staging systems for risk stratification.

Area Of Science

Radiology
Artificial Intelligence
Oncology

Background

Hepatocellular carcinoma (HCC) recurrence after surgical resection remains a significant clinical challenge.
Accurate prediction of early recurrence is crucial for personalized treatment strategies and improved patient outcomes.

Purpose Of The Study

To evaluate the efficacy of deep learning (DL) automated segmentation-based MRI radiomic features combined with clinical-radiological data in predicting early recurrence of single HCC after curative resection.
To compare the predictive performance of this hybrid model against established staging systems like BCLC and CNLC.

Main Methods

A retrospective study of 434 patients with single HCC who underwent contrast-enhanced MRI before curative hepatectomy.
Automated segmentation of liver and HCC using 3D U-net DL algorithms on six MRI sequences.
Extraction of radiomic features from tumor, tumor border extensions, and liver; construction of a hybrid Cox regression model with optimal radiomic signature and preoperative characteristics.

Main Results

The optimal radiomic signature included HCC with a 5 mm tumor border extension and liver features (training C-index, 0.696).
The hybrid model, incorporating this signature with rim arterial phase hyperenhancement (APHE) and incomplete tumor capsule, achieved a validation C-index of 0.706.
The hybrid model demonstrated superior 2-year time-dependent AUC (0.743) compared to BCLC (0.550) and CNLC (0.635) systems, effectively stratifying patients into distinct risk groups.

Conclusions

A preoperative imaging model integrating DL-based radiomics, rim APHE, and incomplete tumor capsule accurately predicts early postsurgical recurrence of single HCC.
This advanced model offers potential for individualized risk assessment, guiding postoperative management for HCC patients.

Keywords:

Artificial intelligence Carcinoma (hepatocellular) Machine learning Magnetic resonance imaging Recurrence

Related Concept Videos

Magnetic Resonance Imaging

Magnetic resonance imaging (MRI) is a noninvasive medical imaging technique based on a phenomenon of nuclear physics discovered in the 1930s, in which matter exposed to magnetic fields and radio waves was found to emit radio signals. In 1970, a physician and researcher named Raymond Damadian noticed that malignant (cancerous) tissue gave off different signals than normal body tissue. He applied for a patent for the first MRI scanning device in clinical use by the early 1980s. The early MRI...