Date of Publication :21st May 2024

Abstract:Interstitial Lung Disease (ILD) is a prevalent and progressive respiratory condition that imposes a significant global health burden. Accurate diagnosis of ILD is vital for effective management and intervention strategies. In this paper, we present a novel classification approach for ILD using a Vision Transformer (ViT) model. Vision Transformer (ViT) presents an innovative deep learning framework that has consistently showcased exceptional performance across a wide range of computer vision applications. We aim to explore its applicability in the domain of medical imaging for ILD diagnosis. The proposed method leverages a dataset comprising chest X-rays and CT scans from patients with ILD, as well as healthy controls. Our methodology incorporates self-attention mechanisms to capture long-range dependencies within the images. This enables the model to effectively discern relevant patterns and features. By fine-tuning the pre-trained ViT model on this dataset, we employ transfer learning to adapt the model for the specific task of ILD classification. Our objective is to attain a notable accuracy without a substantial increase in parameter usage during the fine-tuning process. Accurate early-stage diagnosis of ILD through non-invasive imaging techniques holds the potential for timely interventions and improved patient outcomes. Our findings underscore the capability of Vision Transformers as a potent tool in medical image analysis. This research paves the way for enhanced diagnostic capabilities in the field of respiratory medicine.

Reference :