Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Denoising OCT Images Using Steered Mixture of Experts with Multi-Model Inference

Aytaç Özkan, Elena Stoykova, Thomas Sikora, Violeta Madjarova

TL;DR

The paper tackles speckle noise in OCT that hinders diagnostic utility by introducing BM-SMoE-AE, a denoising framework that fuses Block-Matching with a Steered-Mixture of Experts and an enhanced autoencoder gating network. The method integrates an edge-aware SMoE, BM patch grouping, a gating network for SMoE parameter estimation, and a composite loss with multi-model fusion to achieve high edge fidelity and efficient computation. Quantitative results show BM-SMoE-AE surpassing BM3D and competing methods in PSNR and SSIM, with NR metrics corroborating perceptual quality, demonstrating potential for improved OCT usability in clinical diagnostics. The work advances OCT denoising by combining structured patch-based grouping, learned parameter estimation, and multi-expert fusion to balance artifact suppression with edge preservation, though it notes the model’s substantial parameter count and the need for broader validation.

Abstract

In Optical Coherence Tomography (OCT), speckle noise significantly hampers image quality, affecting diagnostic accuracy. Current methods, including traditional filtering and deep learning techniques, have limitations in noise reduction and detail preservation. Addressing these challenges, this study introduces a novel denoising algorithm, Block-Matching Steered-Mixture of Experts with Multi-Model Inference and Autoencoder (BM-SMoE-AE). This method combines block-matched implementation of the SMoE algorithm with an enhanced autoencoder architecture, offering efficient speckle noise reduction while retaining critical image details. Our method stands out by providing improved edge definition and reduced processing time. Comparative analysis with existing denoising techniques demonstrates the superior performance of BM-SMoE-AE in maintaining image integrity and enhancing OCT image usability for medical diagnostics.

Denoising OCT Images Using Steered Mixture of Experts with Multi-Model Inference

TL;DR

The paper tackles speckle noise in OCT that hinders diagnostic utility by introducing BM-SMoE-AE, a denoising framework that fuses Block-Matching with a Steered-Mixture of Experts and an enhanced autoencoder gating network. The method integrates an edge-aware SMoE, BM patch grouping, a gating network for SMoE parameter estimation, and a composite loss with multi-model fusion to achieve high edge fidelity and efficient computation. Quantitative results show BM-SMoE-AE surpassing BM3D and competing methods in PSNR and SSIM, with NR metrics corroborating perceptual quality, demonstrating potential for improved OCT usability in clinical diagnostics. The work advances OCT denoising by combining structured patch-based grouping, learned parameter estimation, and multi-expert fusion to balance artifact suppression with edge preservation, though it notes the model’s substantial parameter count and the need for broader validation.

Abstract

In Optical Coherence Tomography (OCT), speckle noise significantly hampers image quality, affecting diagnostic accuracy. Current methods, including traditional filtering and deep learning techniques, have limitations in noise reduction and detail preservation. Addressing these challenges, this study introduces a novel denoising algorithm, Block-Matching Steered-Mixture of Experts with Multi-Model Inference and Autoencoder (BM-SMoE-AE). This method combines block-matched implementation of the SMoE algorithm with an enhanced autoencoder architecture, offering efficient speckle noise reduction while retaining critical image details. Our method stands out by providing improved edge definition and reduced processing time. Comparative analysis with existing denoising techniques demonstrates the superior performance of BM-SMoE-AE in maintaining image integrity and enhancing OCT image usability for medical diagnostics.
Paper Structure (11 sections, 10 equations, 4 figures, 3 tables)

This paper contains 11 sections, 10 equations, 4 figures, 3 tables.

Table of Contents

  1. INTRODUCTION
  2. METHODS
  3. The Edge-Aware SMoE Model
  4. The Block-Matching Algorithm
  5. The Autoencoder and SMoE Gating Network
  6. Training SMoE Gating Network
  7. RESULTS
  8. Perceptual Quality Metrics for Reconstruction
  9. Dataset
  10. Experimental Results and Analysis
  11. CONCLUSION

Figures (4)

  • Figure 1: SMoE’s kernels, gates, and regression function respectively.
  • Figure 2: Encoder and Decoder (Gating) Network.
  • Figure 3: Visual denoising comparisons against the SotA methods for SDOCT DatasetRN88.
  • Figure 4: In house dataset, denoising performance comparison between PBML-SMoE and BM3D.