Enhancing Interpretability in Lung CT Image Clustering with HFMSTM-Enhanced Explainable CNN for Disease Detection

L. Dhanapriya, S. Preetha

Abstract: The importance of interpretability in medical image clustering has grown significantly as a foundation for trustworthy clinical decision-making and healthcare transparency in Artificial Intelligence systems. Although Convolutional Neural Networks (CNN) and other Deep Learning (DL) methods have demonstrated remarkable results in distinctive feature extraction from CT scans of the lungs, the majority of clustering algorithms are unreliable, leaving no information about the clustering process or the significance of the features. Furthermore, traditional clustering algorithms are unable to preserve the topologic relationships of complex image representations, thus making the clustering result difficult to interpret and unreliable. In order to address the above drawbacks, this study proposed a novel approach for lung Computed Tomography (CT) image interpretation and clustering called the HFMSTM Enhanced Self Organizing Map (SOM)?Explainable CNN (HSECNN) framework. To obtain informative deep feature representations, the proposed system employed an attention-based Explainable CNN, and to perform topology preserving clustering, an HFMSTM based SOM was used. The HFMSTM enhancement enabled neighborhood connectivity and cluster organization which enabled complex patterns in lung CT images to be well represented. In order to emphasize transparency even more, SHapley Additive exPlanations (SHAP) was employed for analyzing the global importance of features, while Local Interpretable Model-Agnostic Explanations (LIME) was used for providing local justifications for each cluster participation. Moreover, visualization modules were added to gain insight into clusters, feature distributions and topological relationships. The suggested HSECNN framework outperformed the other methods in terms of precision, recall, accuracy, and F1-Score, according to the experimental data. The performance of the clustering was improved through the use of topology-preserving clustering and Explainable AI (XAI) methods, which also helped to make the clustering transparent and trusted by users. As a result, medical image analysis and clustering of CT scans of the lungs are both improved by the suggested framework.

Keywords: Lung CT Images, Interpretable Clustering, Explainable CNN, SOM, HFMSTM, SHAP, LIME, XAI.

How to Cite?: L. Dhanapriya, S. Preetha, "Enhancing Interpretability in Lung CT Image Clustering with HFMSTM-Enhanced Explainable CNN for Disease Detection", Volume 15 Issue 7, July 2026, International Journal of Science and Research (IJSR), Pages: 1088-1098, https://www.ijsr.net/getabstract.php?paperid=SR26713155355, DOI: https://dx.doi.org/10.21275/SR26713155355

Download Citation: APA | MLA | BibTeX | EndNote | RefMan