The Role of Artificial Intelligence in Diagnostic Radiology: Knowledge, Attitude, and Practice of Radiologists and Radiology Residents in Kenya

Abstract

Background. Phenomenal developments in Artificial Intelligence/ Machine Learning (AI/ML) have led to the creation of powerful computerized algorithms with proven capabilities in the performance of some tasks in the radiology workflow. Predictions of the impact that AI/ML will have in the field of Diagnostic Radiology (DR) range from rendering radiologists obsolete to drastic changes in its practice. This has resulted in varied attitudes and perceptions of AI among radiologists and radiology residents. It is, therefore, key that radiologists be well versed with terminologies, concepts, and applications of AI/ML in DR to enable them to accurately project their potential effects and prepare them for the same. Objective This study assessed the knowledge, attitudes, and practice of radiologists and radiology residents towards AI/ML in the field of DR in Kenya. Methodology. A cross-sectional descriptive study method was used. The study was primarily conducted among members of the Kenya Association of Radiologists (KAR). Eligible persons included radiologists and radiology residents based in Kenya. Data was collected by sharing a web-based questionnaire on the association’s WhatsApp platform, which had a membership of 199. Total sampling technique was used. Study variables were be calculated by the use of percentages and frequencies. Pearson’s Chi-square and Mann-Whitney U test were utilized to compare categorical data and study groups, respectively. This study is of help in identifying the level of knowledge of AI in DR, its utilization in daily practice, and the prevailing attitudes and perceptions surrounding it. The data was analysed using Statistical Package for Social Sciences (SPSS) version 26. Results A considerable majority of the participants had basic knowledge on Artificial intelligence, for they had read/watched/attended an AI presentation (n = 73, 65.8%). Less than half of the participants were knowledgeable on machine learning, artificial neural networks and deep learning concept. The use of AI in detection in radiology emerged as the most mentioned application (37.4%), with the remaining applications such as segmentation, speech recognition, registration, workflow management, protocol optimization and others only accounting for less than 20% individually.......................................................................