Conference Presentations

P02-08: Performance of AI models in detection microvascular invasion in hepatocellular carcinoma across multi-modal imaging: a systematic review and meta-analysis

Authors

Minh Huu Nhat Le1,2,3,4,5, Dang The Hung6, Thanh V. Kim4,5,7, Nghia Minh Tran8, Phat Tuan Nguyen9, Thi-My-Trang Luong10, Luu Ngoc Mai11, Nguyen Hai Nam12,13, Han Hong Huynh14, Dang Hai Nguyen15, Phat Kim Huynh16, Hien Quang Kha1,2,3, Doan Y Dao4,5,*, Nguyen Quoc Khanh Le2,3,17,18, *

Presenter: Minh Huu Nhat Le, MD, PhD student

Affiliations

1 International Ph.D. Program in Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.

2 Research Center for Artificial Intelligence in Medicine, Taipei Medical University, Taipei 110, Taiwan.

3 AIBioMed Research Group, Taipei Medical University, Taipei 110, Taiwan.

4 Vietnam Viral Hepatitis Alliance, Reston, VA, USA.

5 Center of Excellence for Liver Disease in Viet Nam, Johns Hopkins University School of Medicine, Baltimore, MD, USA.

6 School of Biomedical Engineering & Imaging Sciences, Faculty of Life Sciences & Medicine, King's College London, London, UK

7 Department of Epidemiology, Pham Ngoc Thach University of Medicine, Ho Chi Minh City, Viet Nam.

8 Faculty of Medicine, Can Tho University of Medicine and Pharmacy, Can Tho City, Vietnam.

9 Faculty of Medicine, Hue University of Medicine and Pharmacy, Hue University, Hue, Vietnam.

10 International Master Program in Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.

11 Department of Internal Medicine, University of Medicine and Pharmacy at Ho Chi Minh City, Ho Chi Minh City 70000, Vietnam

12 Department of Liver Tumor, Cancer Center, Cho Ray Hospital, Ho Chi Minh City, Vietnam

13 Liver Transplant Unit, Cancer Center, Cho Ray Hospital, Ho Chi Minh City, Vietnam

14 International Master Program for Translational Science, College of Medical Science and Technology, Taipei Medical University, Taipei 110, Taiwan.

15Department of Medical Engineering, University of South Florida, Tampa, FL, USA

16 Industrial and Systems Engineering, North Carolina Agricultural and Technical State University, Greensboro, NC 27411, US.

17 Professional Master Program in Artificial Intelligence in Medicine, College of Medicine, Taipei Medical University, Taipei 110, Taiwan.

18 Translational Imaging Research Center, Taipei Medical University Hospital, Taipei 110, Taiwan.

ABSTRACT

Background: Hepatocellular carcinoma (HCC) ranks as the sixth most common and third deadliest cancer globally. Microvascular invasion (MVI) is a crucial prognostic factor in HCC that impacts recurrence rates and survival. Radiomics-based AI models are proving effective in non-invasively predicting MVI using MRI and multiphase CT images to analyze radiomic features. These models not only help in identifying MVI but also enhance treatment planning by integrating radiomic data with clinical information. Further, the combination of multiple diagnostic modalities into a unified nomogram significantly improves predictive accuracy, offering critical insights for surgical strategies and patient outcomes. This AI-driven approach facilitates personalized treatment and improved prognostic assessments in HCC.

Objectives: This study evaluates the predictive accuracy of radiomics and artificial intelligence (AI) models, including deep learning (DL) and machine learning (ML) techniques, for predicting microvascular invasion (MVI) in hepatocellular carcinoma (HCC) using modalities like Computed Tomography (CT), Magnetic Resonance Imaging (MRI), ultrasound (US), Positron Emission Tomography/CT (PET/CT), and histopathology.

Methods: Following PRISMA guidelines, we conducted a comprehensive systematic review and database search to identify studies that applied AI to imaging data for detecting MVI in HCC. We extracted the basic information from these studies, as well as the performance of predictive models using multimodal imaging and multiple AI approaches. The studies were assessed for quality using the QUADAS-2 tool, and the methodological quality was evaluated using the Radiomics Quality Score (RQS).

Results: This meta-analysis assessed the diagnostic accuracy of AI models across 51 studies, involving 6,257 records, and segmented by various imaging modalities. For deep learning (DL) models, the overall pooled AUC was 0.85 (95% CI: 0.81-0.89). MRI-specific DL models achieved an AUC of 0.87 (95% CI: 0.83-0.91), CT-specific models had an AUC of 0.84 (95% CI: 0.76-0.91), and models using both MRI and CT reported an AUC of 0.82 (95% CI: 0.72-0.92). Machine learning (ML) models showed a slightly lower overall pooled AUC of 0.83 (95% CI: 0.80-0.85). In detailed findings, MRI-specific ML models reached an AUC of 0.86 (95% CI: 0.82-0.90), CT alone had an AUC of 0.81 (95% CI: 0.76-0.85), and PET/CT models demonstrated an AUC of 0.88 (95% CI: 0.85-0.92). Ultrasound-specific models had an AUC of 0.86 (95% CI: 0.77-0.95), while the combination of MRI, CT, and ultrasound in ML models resulted in an AUC of 0.80 (95% CI: 0.74-0.86). The inclusion of clinical variables significantly enhanced diagnostic accuracies. MRI models with added clinical data yielded an AUC of 0.91 (95% CI: 0.87-0.95), while CT and PET/CT models reached AUCs of 0.82 (95% CI: 0.75-0.90) and 0.89 (95% CI: 0.85-0.93), respectively.

Conclusions: AI models show high accuracy in predicting MVI in HCC, supporting their potential integration into preoperative diagnostic protocols to enhance treatment planning and outcomes. Despite promising results, the retrospective design and variability among studies necessitate further prospective studies to standardize and validate these AI applications in clinical practice.

Keywords: Hepatocellular carcinoma, microvascular invasion, artificial intelligence, systematic review, meta-analysis, deep learning, machine learning.

Corresponding author: Professor Nguyen Quoc Khanh Le (Professional Master Program in Artificial Intelligence in Medicine, College of Medicine, Taipei Medical University, Taipei 110, Taiwan., Email: khanhlee@tmu.edu.tw). 

Presenter author: Minh Huu Nhat Le, email: d142111009@tmu.edu.tw (ORCID: 0000-0002-7728-1539)

Conference information: APASL Single Topic Conference on Metabolic Associated Fatty Liver Disease (MAFLD), Link: https://www.apaslstc2024kaohsiung.org/

Poster Presentation

Conference Title

2024 APASL Single Topic Conference on MAFLD (2024 APASL STC on MAFLD)

Date

June 28-30, 2024

Venue

June 28

9F, Heart Reading Banquet, TAI Urban Resort

June 29-30

Kaohsiung Exhibition Center (KEC)

TAI Urban Resort, Kaohsiung Taiwan

Kaohsiung Exhibition Center (KEC), Kaohsiung Taiwan

International Conjunct Organization

The Asian Pacific Association for the Study of the Liver (APASL)

Organizer

Taiwan Association for the Study of the Liver (TASL)

Taiwan Association for the Study of the Liver Foundation (TASLF)

Supporting Associations

∙ The Gastroenterological Society of Taiwan

∙ Taiwan Surgical Society of Gastroenterology

∙ Taiwan Society of Pediatric Gastroenterology, Hepatology and Nutrition

∙ Taiwan Liver Cancer Association

∙ Taiwanese Association of Diabetes Educators

∙ Taiwan Society of Cardiology

∙ The Diabetes Association of the Republic of China (Taiwan)

∙ The Endocrine Society of the Republic of China (Taiwan)

Conference Language
English


About APASL (The Asian Pacific Association for the Study of the Liver)

The Asian Pacific Association for the Study of the Liver (APASL)
Since its inception in August 1978, The Asian Pacific Association for the Study of the Liver (APASL), it never looks back but sticks to reach a goal towards the progress of scientific and practical advancements of the field of Hepatology.

Today, it is one of the leading associations based on the investigations and treatments of liver diseases internationally. And it is the largest scientific body that upholds a standard of professional research, aiming to create improved treatment of methods for millions of liver patients particularly in the area of Asia Pacific Region.

The main objectives of APASL are aiming, firstly to promote the latest scientific advancement and education of hepatology science; secondly, to value the exchange of new information as well as to develop the consensus in order to encourage more practices of medicine to overcome liver diseases; lastly, to coordinate related studies of both various scientists and clinicians throughout the region.

Our members include all medical professionals dedicated to hepatology—its research, practice, and care. We have members covering the region of Manchuria in the North to Australia in the South, the Pacific Islands in the East, and, Iran in the West. Our members are elected by the evaluations of their publications. It has been an evident platform for mentoring and sharing of knowledge as well as the solid dedication towards professional progression and development among the core values of APASL and its members' commitments.

About TASL (Taiwan Association for the Study of the Liver)

Taiwan Association for the Study of the Liver (TASL) was founded in January 1990, by the same founder of The Gastroenterological Society of Taiwan (GEST), Professor Juei-Low Sung, named "the father of Taiwan's liver research".