HD3C: Efficient Medical Data Classification for Embedded Devices

2025-09-18

AI Summary

Deploying energy-intensive, GPU-dependent deep learning models on resource-constrained embedded devices—such as home- or point-of-care medical terminals—remains highly challenging. Method: This paper proposes Hyperdimensional Computing with Class-Wise Clustering (HD³C), the first framework to integrate class-aware clustering into hyperdimensional computing. HD³C constructs intra-class cluster prototypes, enables Hamming-space similarity retrieval, and employs binary hypervector representations for efficient, robust classification. Contribution/Results: Theoretical analysis and experiments demonstrate that, on phonocardiogram classification, HD³C achieves >350× energy reduction over Bayesian ResNet with <1% accuracy degradation. It exhibits strong robustness against sensor noise and hardware bit-flip errors, and supports few-shot learning. Collectively, HD³C significantly enhances practicality and deployment feasibility for edge-based healthcare applications.

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: 3D Computer Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text

Abstract

Energy-efficient medical data classification is essential for modern disease screening, particularly in home and field healthcare where embedded devices are prevalent. While deep learning models achieve state-of-the-art accuracy, their substantial energy consumption and reliance on GPUs limit deployment on such platforms. We present Hyperdimensional Computing with Class-Wise Clustering (HD3C), a lightweight classification framework designed for low-power environments. HD3C encodes data into high-dimensional hypervectors, aggregates them into multiple cluster-specific prototypes, and performs classification through similarity search in hyperspace. We evaluate HD3C across three medical classification tasks; on heart sound classification, HD3C is $350 imes$ more energy-efficient than Bayesian ResNet with less than 1% accuracy difference. Moreover, HD3C demonstrates exceptional robustness to noise, limited training data, and hardware error, supported by both theoretical analysis and empirical results, highlighting its potential for reliable deployment in real-world settings. Code is available at https://github.com/jianglanwei/HD3C.

Problem

• Develops energy-efficient medical data classification for embedded devices
• Addresses deep learning models' high energy consumption on low-power platforms
• Enables robust disease screening in home and field healthcare settings

Innovation

• Hyperdimensional computing with class-wise clustering
• Encodes data into high-dimensional hypervectors
• Performs classification through similarity search