Adaptive Wavelet Filters as Practical Texture Feature Amplifiers for Parkinson's Disease Screening in OCT

TL;DR

The paper tackles PD screening via retinal texture analysis in OCT, addressing the underexplored role of texture features. It introduces AWFNet, which uses Adaptive Wavelet Filters to amplify texture representations in high-level DNN features, guided by frequency-domain decomposition and a residual reconstruction, and couples this with a Balanced Confidence loss to improve trustworthiness. The approach yields superior PD screening performance compared with state-of-the-art DNNs and loss methods across multiple OCT datasets, while also improving calibration. This combination of texture-feature emphasis and calibrated predictions has practical potential for reliable, non-invasive PD screening using OCT. The work demonstrates both improved accuracy and clinical trustworthiness, suggesting feasibility for real-world deployment with further device-oriented refinement.

Abstract

Parkinson's disease (PD) is a prevalent neurodegenerative disorder globally. The eye's retina is an extension of the brain and has great potential in PD screening. Recent studies have suggested that texture features extracted from retinal layers can be adopted as biomarkers for PD diagnosis under optical coherence tomography (OCT) images. Frequency domain learning techniques can enhance the feature representations of deep neural networks (DNNs) by decomposing frequency components involving rich texture features. Additionally, previous works have not exploited texture features for automated PD screening in OCT. Motivated by the above analysis, we propose a novel Adaptive Wavelet Filter (AWF) that serves as the Practical Texture Feature Amplifier to fully leverage the merits of texture features to boost the PD screening performance of DNNs with the aid of frequency domain learning. Specifically, AWF first enhances texture feature representation diversities via channel mixer, then emphasizes informative texture feature representations with the well-designed adaptive wavelet filtering token mixer. By combining the AWFs with the DNN stem, AWFNet is constructed for automated PD screening. Additionally, we introduce a novel Balanced Confidence (BC) Loss by mining the potential of sample-wise predicted probabilities of all classes and class frequency prior, to further boost the PD screening performance and trustworthiness of AWFNet. The extensive experiments manifest the superiority of our AWFNet and BC over state-of-the-art methods in terms of PD screening performance and trustworthiness.

Figures (4)

• Figure 1: (a) An OCT image with annotated retinal nerve fiber layer (RNFL), ganglion cell layer and inner plexiform layer (GCL+IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer nuclear layer (ONL), photoreceptors inner segment (PRL-IS), photoreceptors outer segment (PRL-OS), retinal pigment epithelium (RPE), choroid layer (CHL). (b) The texture feature value distribution differences of first-order mean and gray level size zone matrix (GLSZM) feature-run entropy (RE) between the PD group and the healthy control group. These two texture features are extracted from RNFL, OPL, and ONL. (c) The visualizations of the correlation coefficient distributions among texture features extracted from different retinal layers based on OCT images and the frequency components extracted from high-level feature maps, which are obtained from ResNet (Left) and AWFNet (Right) accordingly. Here, the vertical axis denotes correlation coefficient distribution, and the horizontal axis denotes different retinal layers.
• Figure 2: The general architecture of Adaptive Wavelet Filtering Network (AWFNet) with the Balanced Confidence Loss (BC) $L_{BC}$ for trustworthy PD screening under OCT images, which emphasizes important texture feature representations in high-level feature maps by the well-designed AWF. Particularly, we design a novel BC loss to boost the trustworthiness of AWFNet and further improve PD screening performance. Therefore, our AWFNet can produce trustworthy PD screening results in the inference stage, which has great potential to be deployed on real medical devices.
• Figure 3: The simple implementation of adaptive group linear weighting operator.
• Figure 4: The visualizations of the correlation coefficients among texture features extracted from different retinal layers based on OCT images and the frequency components extracted from high-level feature maps, which are obtained from different DNNs ((ResNet18, ViT, SRMNet, and AWFNet, left-to-right order) accordingly. Here, the vertical axis denotes correlation coefficient distribution, and the horizontal axis denotes nine different retinal layers.