Analyzing the Effect of $k$-Space Features in MRI Classification Models
This addresses the critical need for trust and reliability in clinical settings by making AI models more transparent, though it appears incremental as it builds on existing CNN and UMAP techniques.
The paper tackled the problem of model opacity in AI for medical diagnostics by developing an explainable AI methodology for MRI classification, resulting in enhanced early training efficiency and improved interpretability for diagnostic inferences.
The integration of Artificial Intelligence (AI) in medical diagnostics is often hindered by model opacity, where high-accuracy systems function as "black boxes" without transparent reasoning. This limitation is critical in clinical settings, where trust and reliability are paramount. To address this, we have developed an explainable AI methodology tailored for medical imaging. By employing a Convolutional Neural Network (CNN) that analyzes MRI scans across both image and frequency domains, we introduce a novel approach that incorporates Uniform Manifold Approximation and Projection UMAP] for the visualization of latent input embeddings. This approach not only enhances early training efficiency but also deepens our understanding of how additional features impact the model predictions, thereby increasing interpretability and supporting more accurate and intuitive diagnostic inferences