Keywords: artificial intelligence, radiomics, precision diagnosis, clinical explainable AI (clinical x-AI)

"AI applications in routine clinical imaging: detection, segmentation, diagnosis, and prognosis" 

In this dissertation, applicability of Artificial intelligence (AI) in the medical imaging field has been investigated on respiratory, oncological and otolaryngology use-cases. AI-based model centric approaches were explored and validated to prove generalizability when compared to expert radiologists and clinicians. Methods to prove explainability of the AI models in the language of clinicians were also investigated. Overall, the thesis proves the overall hypothesis that (semi) automated AI based methodologies can produce generalizable performance equivalent to that of an expert human charged with the same tasks and is exemplified in detection, diagnosis, and treatment response prediction use cases. The (semi) AI based methodologies were explored specifically on the following clinical problem statements: 1) Covid-19 diagnosis and differential diagnosis using CT imaging, 2) Pulmonary embolism detection and diagnosis using CTPA imaging, 3) Treatment response prediction in patients with non-metastatic non-small cell lung cancer using CT imaging and 4) Menière’s disease diagnosis using MRI imaging.

Language: English