Deep learning-based 3D quantitative total tumor burden predicts early recurrence of BCLC A and B HCC after resection

Hong Wei ^1,2, Tianying Zheng ¹, Xiaolan Zhang ³

¹Department of Radiology, Functional, and Molecular Imaging Key Laboratory of Sichuan Province, West China Hospital, Sichuan University, Chengdu, Sichuan, 610041, China.
²Department of Radiology, Seoul National University Hospital, Seoul, 03080, Republic of Korea.
³Shukun Technology Co., Ltd, Beijing, 100102, China.
⁴Big Data Research Center, University of Electronic Science and Technology of China, Chengdu, Sichuan, 610000, China.
⁵Department of Radiology, Seoul National University College of Medicine, Seoul, 03080, Republic of Korea.
⁶Department of Radiology, Duke University Medical Center, Durham, NC, 27710, USA.
⁷Center for Advanced Magnetic Resonance in Medicine, Duke University Medical Center, Durham, NC, 27705, USA.
⁸Division of Gastroenterology, Department of Medicine, Duke University Medical Center, Durham, NC, 27710, USA.
⁹Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, 100102, China.
¹⁰Department of Radiology, Functional, and Molecular Imaging Key Laboratory of Sichuan Province, West China Hospital, Sichuan University, Chengdu, Sichuan, 610041, China. hanyu_jiang@foxmail.com.
¹¹Department of Radiology, Functional, and Molecular Imaging Key Laboratory of Sichuan Province, West China Hospital, Sichuan University, Chengdu, Sichuan, 610041, China. songlab_radiology@163.com.
¹²Department of Radiology, Sanya People's Hospital, Sanya, Hainan, 572000, China. songlab_radiology@163.com.

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

European Radiology +

|

July 19, 2024

View abstract on PubMed

Summary

Deep learning-assisted MRI accurately predicts early recurrence of hepatocellular carcinoma (HCC). Total tumor burden (TTB) derived from automated segmentation improves Barcelona Clinic Liver Cancer (BCLC) staging for patients with HCC.

Area Of Science

Oncology
Radiology
Artificial Intelligence

Background

Hepatocellular carcinoma (HCC) recurrence after surgery poses a significant clinical challenge.
Accurate prediction of early recurrence (ER) is crucial for optimizing patient management and treatment strategies.
Current staging systems may not fully capture individual recurrence risk.

Purpose Of The Study

To evaluate deep learning (DL)-assisted automated three-dimensional quantitative tumor burden (TTB) on MRI for predicting postoperative ER in HCC.
To assess the prognostic value of DL-derived TTB compared to existing clinicopathological variables.
To determine if DL-derived TTB can refine the Barcelona Clinic Liver Cancer (BCLC) staging system.

Main Methods

Retrospective analysis of 592 patients with BCLC A and B HCC who underwent resection and preoperative contrast-enhanced MRI.
Automated segmentation using a DL tool to quantify total tumor volume and TTB (%).
Cox regression analyses to determine the prognostic value of clinicopathological and TTB parameters for ER.

Main Results

Total tumor burden (TTB) was the most significant predictor of ER (HR=2.2, p<0.001).
A TTB threshold of 6.84% stratified patients into distinct ER risk groups within BCLC A and B stages (p<0.001).
Refined staging (BCLC A<sub>n</sub> and B<sub>n</sub>) based on TTB demonstrated significantly different 2-year ER rates (30.5% vs. 58.1%, HR=2.8, p<0.001).

Conclusions

DL-assisted automated MRI assessment of TTB is a potent biomarker for predicting postoperative ER in HCC.
Incorporating TTB into the BCLC staging system allows for refined subcategorization of BCLC A and B patients.
This approach enhances risk stratification and may guide personalized treatment decisions for HCC patients.

Keywords:

Carcinoma (hepatocellular) Hepatectomy Magnetic resonance imaging Recurrence Tumor burden