PROJECT TITLE :

Automatic Video Analysis Framework for Exposure Region Recognition in X-Ray Imaging Automation

ABSTRACT:

The deep learning-based automatic recognition of the scanning or exposing region in Medical Imaging automation is an exciting new technique that has the potential to improve image quality, reduce the burdensome workload of radiographers, and optimize imaging workflow. X-ray imaging, on the other hand, has only a limited amount of relevant research and practice. In this paper, our primary focus is on two important issues pertaining to the automation of X-ray imaging: the automatic recognition of the exposure moment and the exposure region. As a result, we propose a framework for automatic video analysis that is based on the hybrid model and has performance that is nearly on par with real time. Body Structure Detection, Motion State Tracing, and Body Modeling are the three interdependent components that make up the framework. In order to obtain the patient's corresponding body keypoints and body Bboxes, Body Structure Detection must first disassemble the patient. When these two distinct types of body structure representations are combined and analyzed, rich spatial location information about the patient's body structure can be obtained. Motion State Tracing is an image analysis technique that focuses on determining the motion state of an exposure region in order to identify the correct exposure moment. When the exposure moment occurs, the Body Modeling process determines the exposure region to be in effect. In order to verify that the proposed method produces accurate results, a comprehensive dataset of X-ray examination scenes is being compiled. Extensive testing demonstrates that the proposed method is superior when it comes to automatically recognizing the exposure moment and exposure region. This is the first method that, when applied to X-ray imaging, makes it possible to automatically and accurately recognize the exposure region without the assistance of the radiographer. This paradigm provides this capability.