Year: 2024
Type: article
Abstract: Early and accurate detection of deadly diseases from medical images is a major challenge in healthcare. In this paper, we develop a new methodology to solve this problem by using “Generative Adversarial Networks (GANs), Convolutional Neural Networks (CNNs) and Explainable Artificial Intelligence (XALI)” techniques. Using GANs to generate high quality synthetic medi- cal images extends the limited training data baseline, allowing the CNN model to improve performance. Next we designed our CNN model layered with multiple convolutional and pooling layers that can classifies the images into different disease categories such as brain tumor. “XAI techniques, such as Grad-CAM is incorporated to increase the understandable and transparency of the CNN model’s decision making process.” Grad-CAM produces localization maps distinguishing the prominent regions of the input images that are important for the classification. Combining GAN based data augmentation, CNN based image classification and XAI powered explainability have synergistically integrated to produce a robust and reliable detection system of deadly diseases from medical images. Overall, this research may serve as a major contribution to the field of medical imaging analysis and help towards designing more reliable, more impactful, AI driven diagnostic systems
Authors: Neeraj Jaiswal, Sonu Choubey, Amika ., Divya Sharma
Institution: Chandigarh University
Language: English
Subfield: Neurology
Field: Neuroscience
Domain: Life Sciences
Open Access status: bronze
OpenAlex ID: W4405085767