InfoMotion: A Graph-Based Approach to Video Dataset Distillation for Echocardiography

2025-12-10

AI Summary

The exponential growth of echocardiography video data poses severe storage and computational bottlenecks for model training. Method: We propose a novel video data distillation framework integrating temporal motion feature encoding—via optical flow and temporal CNNs—with intra-class Infomap graph clustering. Specifically, we construct class-level adjacency graphs, apply community detection to identify representative sample clusters, synthesize compact video sets, and employ reweighted fine-tuning to enhance generalization. Results: On EchoNet-Dynamic, our method achieves 69.38% test accuracy using only 25 distilled videos—outperforming state-of-the-art distillation approaches. This work is the first to jointly leverage dynamic motion modeling and graph-structured clustering for medical video distillation, effectively preserving clinically critical motion patterns while maintaining inter-sample diversity. It significantly improves training efficiency and scalability without compromising diagnostic relevance.

Technology Category

• Computer Vision: Video Understanding & Activity Analysis
• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Mining of Visual, Multimedia & Multimodal Data

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Echocardiography playing a critical role in the diagnosis and monitoring of cardiovascular diseases as a non-invasive real-time assessment of cardiac structure and function. However, the growing scale of echocardiographic video data presents significant challenges in terms of storage, computation, and model training efficiency. Dataset distillation offers a promising solution by synthesizing a compact, informative subset of data that retains the key clinical features of the original dataset. In this work, we propose a novel approach for distilling a compact synthetic echocardiographic video dataset. Our method leverages motion feature extraction to capture temporal dynamics, followed by class-wise graph construction and representative sample selection using the Infomap algorithm. This enables us to select a diverse and informative subset of synthetic videos that preserves the essential characteristics of the original dataset. We evaluate our approach on the EchoNet-Dynamic datasets and achieve a test accuracy of (69.38%) using only (25) synthetic videos. These results demonstrate the effectiveness and scalability of our method for medical video dataset distillation.

Problem

• Develops a method to compress echocardiographic video datasets for efficient storage and training.
• Selects a small, diverse subset of synthetic videos that retain key clinical motion features.
• Evaluates the approach on EchoNet-Dynamic to maintain accuracy with minimal synthetic data.

Innovation

• Motion feature extraction captures temporal dynamics
• Class-wise graph construction organizes data structure
• Infomap algorithm selects representative synthetic videos