Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Resource-efficient medical image classification for edge devices

Mahsa Lavaei, Zahra Abadi, Salar Beigzad, Alireza Maleki

TL;DR

The paper tackles the challenge of deploying medical image classifiers on resource-constrained edge devices by combining quantization-aware training (QAT) with Parameterized Clipping Activation (PACT) and Saliency-Guided Training (SGT) to maintain diagnostic accuracy and robust interpretability. Using the Kvasir GI endoscopy dataset, it demonstrates substantial reductions in model size and inference cost while preserving clinically acceptable accuracy and stable saliency maps. The approach shows improvements over a saliency-guided baseline and validates the practicality of edge deployment for real-time AI-based GI diagnostics in remote settings. Overall, it offers a practical, interpretable pathway for deploying efficient AI-assisted healthcare tools on low-resource hardware.

Abstract

Medical image classification is a critical task in healthcare, enabling accurate and timely diagnosis. However, deploying deep learning models on resource-constrained edge devices presents significant challenges due to computational and memory limitations. This research investigates a resource-efficient approach to medical image classification by employing model quantization techniques. Quantization reduces the precision of model parameters and activations, significantly lowering computational overhead and memory requirements without sacrificing classification accuracy. The study focuses on the optimization of quantization-aware training (QAT) and post-training quantization (PTQ) methods tailored for edge devices, analyzing their impact on model performance across medical imaging datasets. Experimental results demonstrate that quantized models achieve substantial reductions in model size and inference latency, enabling real-time processing on edge hardware while maintaining clinically acceptable diagnostic accuracy. This work provides a practical pathway for deploying AI-driven medical diagnostics in remote and resource-limited settings, enhancing the accessibility and scalability of healthcare technologies.

Resource-efficient medical image classification for edge devices

TL;DR

The paper tackles the challenge of deploying medical image classifiers on resource-constrained edge devices by combining quantization-aware training (QAT) with Parameterized Clipping Activation (PACT) and Saliency-Guided Training (SGT) to maintain diagnostic accuracy and robust interpretability. Using the Kvasir GI endoscopy dataset, it demonstrates substantial reductions in model size and inference cost while preserving clinically acceptable accuracy and stable saliency maps. The approach shows improvements over a saliency-guided baseline and validates the practicality of edge deployment for real-time AI-based GI diagnostics in remote settings. Overall, it offers a practical, interpretable pathway for deploying efficient AI-assisted healthcare tools on low-resource hardware.

Abstract

Medical image classification is a critical task in healthcare, enabling accurate and timely diagnosis. However, deploying deep learning models on resource-constrained edge devices presents significant challenges due to computational and memory limitations. This research investigates a resource-efficient approach to medical image classification by employing model quantization techniques. Quantization reduces the precision of model parameters and activations, significantly lowering computational overhead and memory requirements without sacrificing classification accuracy. The study focuses on the optimization of quantization-aware training (QAT) and post-training quantization (PTQ) methods tailored for edge devices, analyzing their impact on model performance across medical imaging datasets. Experimental results demonstrate that quantized models achieve substantial reductions in model size and inference latency, enabling real-time processing on edge hardware while maintaining clinically acceptable diagnostic accuracy. This work provides a practical pathway for deploying AI-driven medical diagnostics in remote and resource-limited settings, enhancing the accessibility and scalability of healthcare technologies.

Paper Structure

This paper contains 13 sections, 2 equations, 2 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Dataset
  5. Saliency with PACT-Based Quantization
  6. Evaluation Metrics
  7. Proposed Training Algorithm
  8. Training Configuration
  9. Experiments and Results
  10. Performance Evaluation
  11. Saliency Map Analysis
  12. Discussion
  13. Conclusion

Figures (2)

  • Figure 1: Samples of gastrointestinal conditions and normal findings from endoscopic images in the Kvasir dataset Pogorelov:2017:KMI:3083187.3083212. Shown categories include Dyed Lifted Polyps, Normal Cecum, Normal Pylorus, Polyps, and Ulcerative Colitis, highlighting the diverse visual characteristics and diagnostic challenges in gastrointestinal imaging.
  • Figure 2: Saliency maps highlighting regions of importance in gastrointestinal images from the Kvasir dataset for our proposed method. The maps emphasize areas critical for model predictions.