Classification-Segmentation Pipeline for MRI via Transfer Learning and Residual Networks
Nghia Duong‐Trung, Dung Ngoc Le Ha, Hiep Xuan Huynh · Annals of Computer Science and Information Systems · 2022
Artificial intelligence association into brain magnetic resonance imaging (MRI) and clinical practices embrace substantial cancer diagnosis improvement.The advancement of deep learning has improved the processing and analysis of MRI, boosting models' performance, decreasing the destructive effects of data sources overload, and increasing accurate detection and time efficacy.However, that specific dataset leads to diverse research fields such as image processing and analysis, detection, registration, segmentation, and classification.This paper proposes a decision-making pipeline for MRI data by combining image classification and segmentation.First, the pipeline should correctly produce a correct decision given an MRI image.If the figure is classified as defective, the pipeline can extract defect regions and highlight them accordingly.We have implemented several advanced convolutional neural networks with transfer learning and residual techniques to address two broad clinical concerns in one decision-making workflow.