Otolaryngologist perceptions of AI-based sinus CT interpretation

CT imaging plays a critical role in the work up of chronic rhinosinusitis (CRS). Sinus CT has been used for decades in this regard, and the presence of inflammation as evidenced by mucosal thickening or sinus cavity opacification remains a key diagnostic criterion for CRS [1,2]. However, visual-based assessment of CT imaging as presented in typical radiologist reporting is a subjective exercise and is plagued by variable interpretation. Results are often presented in non-standard or vague language that may or may not be useful for the treating clinician [3]. There exists a need for objective, standardized, and quantitative sinus CT reporting for both clinical and research arenas.

Recent developments in artificial intelligence (AI)-driven image processing have allowed for fully automated, quantitative sinus CT analysis. This novel technology, using convolutional neural networks, permits rapid and precise volumetric quantitation of sinus disease burden. Initial studies from our group have demonstrated a strong correlation of this deep learning-enabled metric to the Lund-Mackay score, one of the most commonly used methods of visual-based disease characterization [4]. The technology has been further extended to investigate other important radiological features such as neo-osteogenesis [5].

Research interest in AI-driven healthcare delivery has seen a dramatic uptick in the past several years, particularly in our own field [6]. It is predicted that this technology will have increasingly commonplace applications in a number of areas, with initial FDA-approved uses in diagnostic fields such as radiology and pathology [7]. However, general acceptance and familiarity with AI technologies at a global level among healthcare professionals is tenuous at best [8].

We are thus presented with a situation in which satisfaction with the current status quo radiology reporting of sinus CT appears low among end-users, but readiness for acceptance of an AI-based solution is unclear. We set out to investigate perceptions of conventional radiology interpretations and how AI technology may add value to this important clinical tool for clinicians in our field.

留言 (0)

沒有登入
gif