Development of Deep Learning Model for Detection of Abdominal Diseases using Radiological Images

As the abdominal and chest injuries are high morbidity and mortality around the globe, these pose an acute challenge in emergency care. Current data from recently released sources indicate that more than half of the patients visiting ER and also being admitted into hospital account for trauma; and the burden of both abdominal and chest injuries is quite high. Imaging plays an important role in diagnosis and treatment planning in the management of abdominal and chest injuries. The most widely practiced modality is radiograph imaging simply because it produces quality and easy results. Improving Automated pathology object detection of things in radiographs with deep learning algorithms promises much for the precision of diagnosis and optimization in health care. This paper discusses and provides a very detailed review of optimized CNNs, primarily ResNet18, ResNet34, along with a custom-built neural network for the identification of disease from X-ray pictures. Such models are trained and validated to reach excellent sensitivity and specificity on a wide variety of diseases by using the annotated RSNA 2023 dataset by professionals in radiology. It extended the scope of earlier research in this field by looking at other imaging conditions and diseases as well. This puts its contribution into the larger, ongoing discussion of AI in radiology. The results of this study help to contribute to the thesis that the promise of deep learning in radiological diagnosis is substantial, and it could be used to better enhance patient outcomes in the practice of emergency radiology.