Metaverse-based deep learning framework for coronary artery stenosis classification using Monte Carlo Dropout-based...

Metaverse offers an immersive healthcare platform that combines virtual reality (VR) and artificial intelligence (AI), providing a new approach to medical diagnostics. However, difficulties such as inadequate spatial resolution, uncertainty management, and ignoring virtual events for patients still exist. To solve these problems, this work introduces an extraordinary way of recognizing coronary artery stenosis and the metaverse to create interactive 3D models for patient-centric virtual events. The process begins with collecting data through Invasive Coronary Angiography (ICA). Then, preprocessing involves the Quantum-Adapted Diffusion (QAD) method to remove noise and a cross-correlation method for motion artifacts to get a clearer ICA image. After preprocessing, the accountable semantic segmentation technique was used to isolate coronary arteries from surrounding tissues, and features were extracted by the Gray Level Co-occurrence Matrix (GLCM). It takes textural and shape features to identify abnormalities. Then, the extracted features are selected using the Stochastic Gradient Descent (SGD) optimization algorithm with the Adam algorithm rates to improve the feature selection model. Finally, the selected features are classified with Monte Carlo Dropout-based ResNet-152 (MCD-ResNet-152) to determine the presence of stenosis. These results are discussed in the metaverse with the patient to deliver a VR examination. The proposed method achieves an accuracy level of 99.20 % in the diagnosis of stenosis highlighting its improvement in diagnostic precision over existing approaches, potentially leading to better patient outcomes.