Multimodal deep learning for liver cancer applications: a scoping review

Area of Science:

• Medical imaging and Artificial Intelligence
• Oncology and Bioinformatics
• Data Science in Healthcare

Background:

• Hepatocellular carcinoma (HCC) is a major global cause of cancer mortality.
• Multimodal data, including medical images and electronic health records (EHR), aids in liver cancer diagnosis.
• Deep learning models integrating multimodal data can improve diagnostic accuracy and clinical decision-making for liver cancer patients.

Purpose of the Study:

• To explore the application of multimodal deep learning techniques in diagnosing and predicting hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA).
• To review the combination of medical imaging and EHR data for liver cancer diagnosis and prognosis.

Main Methods:

• A comprehensive literature search was performed across six databases.
• PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) extension for scoping reviews guided study selection.
• Thematic analysis was used for data extraction and synthesis from included studies.

Main Results:

• Ten studies utilized multimodal deep learning for HCC prediction and diagnosis; none focused on CCA.
• Commonly used imaging modalities included CT and MRI, alongside 51 EHR variables like age, gender, AFP, albumin, and bilirubin.
• Ten distinct deep learning techniques were applied to both imaging and EHR data for prediction and diagnosis.

Conclusions:

• Multimodal data combined with deep learning effectively aids in HCC diagnosis and prediction.
• Limited research and datasets for liver cancer hinder AI advancements in this field.
• Further investigation into multimodal deep learning for liver cancer applications is recommended.