TREAT-Net: Tabular-Referenced Echocardiography Analysis for Acute Coronary Syndrome Treatment Prediction

2025-09-28

AI Summary

To address the high resource consumption, procedural risks, and diagnostic delays associated with invasive coronary angiography in acute coronary syndrome (ACS) diagnosis, this study proposes a non-invasive multimodal predictive framework. The method fuses echocardiographic video sequences with structured clinical records, leveraging a tabular-guided cross-modal attention mechanism and a late-fusion architecture to predict therapeutic interventions non-invasively. Its key innovations include (i) using structured clinical tabular data to guide visual–textual cross-modal interaction, and (ii) designing a multi-source medical data alignment strategy to harmonize heterogeneous inputs. Evaluated on over 9,000 real-world ACS cases, the model achieves a balanced accuracy of 67.6%, an AUROC of 71.1%, and 88.6% consistency in cross-modal intervention prediction—demonstrating substantial improvements in triage efficiency and clinical decision reliability, particularly in resource-constrained settings.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Coronary angiography remains the gold standard for diagnosing Acute Coronary Syndrome (ACS). However, its resource-intensive and invasive nature can expose patients to procedural risks and diagnostic delays, leading to postponed treatment initiation. In this work, we introduce TREAT-Net, a multimodal deep learning framework for ACS treatment prediction that leverages non-invasive modalities, including echocardiography videos and structured clinical records. TREAT-Net integrates tabular-guided cross-attention to enhance video interpretation, along with a late fusion mechanism to align predictions across modalities. Trained on a dataset of over 9000 ACS cases, the model outperforms unimodal and non-fused baselines, achieving a balanced accuracy of 67.6% and an AUROC of 71.1%. Cross-modality agreement analysis demonstrates 88.6% accuracy for intervention prediction. These findings highlight the potential of TREAT-Net as a non-invasive tool for timely and accurate patient triage, particularly in underserved populations with limited access to coronary angiography.

Problem

• Predicting ACS treatment using non-invasive echocardiography and clinical data
• Reducing invasive angiography risks through multimodal deep learning
• Improving patient triage accuracy in resource-limited healthcare settings

Innovation

• Multimodal deep learning framework using echocardiography and clinical records
• Tabular-guided cross-attention enhances echocardiography video interpretation
• Late fusion mechanism aligns predictions across different clinical modalities