Enhancing Diagnostic Precision in Gastric Bleeding through Automated Lesion Segmentation: A Deep DuS-KFCM Approach

TL;DR

This study seeks to revolutionize this domain by introducing a novel deep learning model, the Dual Spatial Kernelized Constrained Fuzzy C-Means (Deep DuS-KFCM) clustering algorithm, which synergizes Neural Networks with Fuzzy Logic to offer a highly precise and efficient identification of bleeding regions.

Abstract

Timely and precise classification and segmentation of gastric bleeding in endoscopic imagery are pivotal for the rapid diagnosis and intervention of gastric complications, which is critical in life-saving medical procedures. Traditional methods grapple with the challenge posed by the indistinguishable intensity values of bleeding tissues adjacent to other gastric structures. Our study seeks to revolutionize this domain by introducing a novel deep learning model, the Dual Spatial Kernelized Constrained Fuzzy C-Means (Deep DuS-KFCM) clustering algorithm. This Hybrid Neuro-Fuzzy system synergizes Neural Networks with Fuzzy Logic to offer a highly precise and efficient identification of bleeding regions. Implementing a two-fold coarse-to-fine strategy for segmentation, this model initially employs the Spatial Kernelized Fuzzy C-Means (SKFCM) algorithm enhanced with spatial intensity profiles and subsequently harnesses the state-of-the-art DeepLabv3+ with ResNet50 architecture to refine the segmentation output. Through extensive experiments across mainstream gastric bleeding and red spots datasets, our Deep DuS-KFCM model demonstrated unprecedented accuracy rates of 87.95%, coupled with a specificity of 96.33%, outperforming contemporary segmentation methods. The findings underscore the model's robustness against noise and its outstanding segmentation capabilities, particularly for identifying subtle bleeding symptoms, thereby presenting a significant leap forward in medical image processing.

Figures (3)

• Figure 1: Model Overview of Deep DuS-KFCM: (a) Overview of the Feature Extraction System using DuS-KFCM. This figure illustrates the initial phase of segmentation with the DuS-KFCM method, highlighting the algorithm's process for distinguishing potential gastric lesions from surrounding tissues by analyzing GLCM and fuzzy clustering for GB feature classification. The workflow encapsulates the progress from lesion identification to cancer boundary specification, aiming for efficient classification. (b) Network Architecture Diagram: DeepLabv3+ with ResNet50 for Enhanced Gastric Lesion Segmentation. Showcasing the systemic architecture employed for refining the boundaries of segmented gastric lesions, this figure details how the DeepLabv3+ model, backed by ResNet50, navigates through the complexities of image segmentation-mitigating misclassification through precise lesion boundary refinement. (c) Pre-Segmentation and Refinement Process for Gastric Endoscopy Imaging with DLv3+ and ResNet50, Based on DuS-KFCM. Providing a comprehensive view of the neural network's block diagram, this figure illustrates the dual-stage segmentation process. Initially leveraging the DuS-KFCM technique for basic lesion identification, it further demonstrates the deep learning model's region-enhancing capabilities, highlighting the encoder and decoder modules' roles in achieving a refined and highly accurate lesion segmentation outcome.
• Figure 2: Comparative Performance Analysis of Clustering Models on GB/GRS Database. This figure portrays the reliability of our novel Deep DuS-KFCM model against traditional methods across both GB and GRS datasets. Key performance metrics—accuracy, precision, and IoU—under differential evaluation scenarios are plotted respectively.
• Figure 3: Segmentation Efficacy on Synthetic Pseudo-color Images. Displayed are the results for the GB and GRS datasets within red and blue delineations, respectively. Here, the juxtaposition of our deep DuS-KFCM method with traditional threshold techniques illuminates the finesse of its clustering accuracy. Lesion boundaries are discriminatively marked: system-detected margins are drawn in magenta, whereas gastroenterologist-annotated ground truths are green, and reference information-fused areas appear white.