Optimization CT images using GANs: Stroke context in Senegal

Generative adversarial networks (GAN) are an emerging technology in the field of artificial intelligence (AI). It is an unsupervised learning framework that involves automatically learning a model from input data to produce a synthetic output that closely resembles the original input. GANs create datasets by making copies of false outputs based on specific inputs. GANs allow machines to generate unique content that is different from human-generated data. GANs are widely used in various image-related fields, such as image-to-text conversion, next-generation faces, high-resolution image conversion, etc. The general growth of artificial intelligence and its medical applications is increasing the efficiency of the healthcare system for patients who need rapid care. Chronic diseases, especially stroke, are now the leading cause of death worldwide. Diagnosis of stroke is often based on imaging techniques such as computed tomography (CT) and magnetic resonance imaging (MRI). However, CT scans are more common and readily available in many African hospitals, especially in Senegal. However, the need for rapid image interpretation of stroke is sometimes problematic due to variable image quality, highlighting the need for improvements to achieve better image quality. This article focuses on improving the resolution of CT images, a commonly used method in medical imaging. However, obtaining a high-quality CT scan often requires lengthy wait times and the use of limited laboratory equipment. In addition, radioactive radiation poses health risks and results in poorer image quality. Super-resolution GAN has the potential to solve this problem. Super-resolution is a method of generating high-resolution images from low-resolution images. We use the input CT dataset to generate higher quality images while keeping parameters unchanged. We found that using ResNet with GANs produced higher quality images than the original input images.