Authors:
K. R. Chaithra, Sreeja Rajesh, T. Shreekumar
Addresses:
Department of Computer Science and Engineering, School of Engineering and Technology, Sapthagiri NPS University, Bengaluru, Karnataka, India. Department of Information Science and Engineering, Mangalore Institute of Technology and Engineering, Mijar, Moodbidre, Karnataka, India. Department of Computer Science and Engineering, Mangalore Institute of Technology and Engineering, Mijar, Moodbidre, Karnataka, India.
Abstract:
Measurement of aortic root dimensions with high accuracy is a critical factor in the diagnosis and follow-up of aortic aneurysms, which, if not detected in time, can result in life-threatening complications. Manual interpretation of CT and MRI scans is a common practice in clinical settings, but it is subject to variability, time-consuming, and dependent on the operator. Within this study, we introduce a Convolutional Neural Network (CNN)-based system for automatic detection and measurement of aortic root size from cardiac MRI images. The research employed the Multi-Centre, Multi-Vendor, Multi-Disease (M&Ms) Challenge Dataset, which comprises 375 patient cases: 150 for training, 125 for validation, and 100 for hidden test cases, with the training set containing approximately 3,288 2D images. Pre-processing operations included Z-score normalisation and data augmentation via rotation, translation, and flipping to enhance generalizability. The CNN model utilised convolutional and pooling layers for feature extraction, activation functions for non-linearity, and fully connected layers for regression-based prediction of aortic diameters. The model was trained with the Adam optimiser and tested against human-annotated ground truth with metrics such as accuracy, precision, recall, F1-score, Matthews Correlation Coefficient (MCC), and absolute error. The CNN consistently outperformed baseline machine learning models such as SVM, KNN, and NN, achieving 91.25% accuracy, 92.31% recall, an MCC of 0.824, and absolute errors of 0.0–0.1 mm.
Keywords: Quantum Segmentation; Convolutional Neural Network; Multimodal Fusion; Aortic Measurement; Active Learning; Real-Time Visualisation; CNN Model; Matthews Correlation Coefficient (MCC).
Received: 06/07/2024, Revised: 18/08/2024, Accepted: 30/09/2024, Published: 03/06/2025
DOI: 10.64091/ATICL.2025.000148
AVE Trends in Intelligent Computer Letters, 2025 Vol. 1 No. 2 , Pages: 73-85