Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging Pre-trained Models for Robust Federated Learning for Kidney Stone Type Recognition

Ivan Reyes-Amezcua, Michael Rojas-Ruiz, Gilberto Ochoa-Ruiz, Andres Mendez-Vazquez, Christian Daul

TL;DR

This research suggests a strong FL framework to improve kidney stone diagnosis and highlights the potential of merging pre-trained models with FL to address privacy and performance concerns in medical diagnostics, and guarantees improved patient care and enhanced trust in FL-based medical systems.

Abstract

Deep learning developments have improved medical imaging diagnoses dramatically, increasing accuracy in several domains. Nonetheless, obstacles continue to exist because of the requirement for huge datasets and legal limitations on data exchange. A solution is provided by Federated Learning (FL), which permits decentralized model training while maintaining data privacy. However, FL models are susceptible to data corruption, which may result in performance degradation. Using pre-trained models, this research suggests a strong FL framework to improve kidney stone diagnosis. Two different kidney stone datasets, each with six different categories of images, are used in our experimental setting. Our method involves two stages: Learning Parameter Optimization (LPO) and Federated Robustness Validation (FRV). We achieved a peak accuracy of 84.1% with seven epochs and 10 rounds during LPO stage, and 77.2% during FRV stage, showing enhanced diagnostic accuracy and robustness against image corruption. This highlights the potential of merging pre-trained models with FL to address privacy and performance concerns in medical diagnostics, and guarantees improved patient care and enhanced trust in FL-based medical systems.

Leveraging Pre-trained Models for Robust Federated Learning for Kidney Stone Type Recognition

TL;DR

This research suggests a strong FL framework to improve kidney stone diagnosis and highlights the potential of merging pre-trained models with FL to address privacy and performance concerns in medical diagnostics, and guarantees improved patient care and enhanced trust in FL-based medical systems.

Abstract

Deep learning developments have improved medical imaging diagnoses dramatically, increasing accuracy in several domains. Nonetheless, obstacles continue to exist because of the requirement for huge datasets and legal limitations on data exchange. A solution is provided by Federated Learning (FL), which permits decentralized model training while maintaining data privacy. However, FL models are susceptible to data corruption, which may result in performance degradation. Using pre-trained models, this research suggests a strong FL framework to improve kidney stone diagnosis. Two different kidney stone datasets, each with six different categories of images, are used in our experimental setting. Our method involves two stages: Learning Parameter Optimization (LPO) and Federated Robustness Validation (FRV). We achieved a peak accuracy of 84.1% with seven epochs and 10 rounds during LPO stage, and 77.2% during FRV stage, showing enhanced diagnostic accuracy and robustness against image corruption. This highlights the potential of merging pre-trained models with FL to address privacy and performance concerns in medical diagnostics, and guarantees improved patient care and enhanced trust in FL-based medical systems.
Paper Structure (13 sections, 7 figures, 1 table)

This paper contains 13 sections, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. State of the Art
  3. Federated Learning
  4. Robustness on Image Corruptions
  5. Method
  6. Preliminaries
  7. Federated Learning for Kidney Stone Identification
  8. Datasets
  9. Robustness Transfer
  10. Experimental Validation
  11. Experimental Setting
  12. Results
  13. Conclusions

Figures (7)

  • Figure 1: Example of image corruptions on Kidney stone dataset subjected to Section view (SEC) corruption at a third severity level. These image corruption simulates scenarios with structured environmental changes that can affect image quality.
  • Figure 2: The schematic representation illustrates two stages: Learning Parameters Optimization (LPO), where two clients optimize the $n_e$ and $n_r$ parameters through model training, and Federated Robustness Validation (FRV), which splits datasets into “good” and “corrupted” subsets. The “corrupted” images undergo diverse corruptions to evaluate the model's robustness using the optimized parameters.
  • Figure 3: Samples of ex-vivo kidney stone images captured using (a) a CCD camera and (b) an endoscope. "SEC" refers to section views, and "SUR" refers to surface views.
  • Figure 4: Sizes of the train and test subsets from the Jonathan El-Beze (Dataset A) and Michel Daudon (Dataset B) used in the Learning Parameter Optimization (LPO).
  • Figure 5: Kidney stone image affected by different types of corruption at the third severity level. This image depicts the MIX subset, showcasing the impact of various image corruptions on kidney stone images from our datasets.
  • ...and 2 more figures