Noninterpretive Uses of Artificial Intelligence in Radiology.

We deem a computer to exhibit artificial intelligence (AI) when it performs a task that would normally require intelligent action by a human. Much of the recent excitement about AI in the medical literature has revolved around the ability of AI models to recognize anatomy and detect pathology on medical images, sometimes at the level of expert physicians. However, AI can also be used to solve a wide range of noninterpretive problems that are relevant to radiologists and their patients. This review summarizes some of the newer noninterpretive uses of AI in radiology.

CT Pulmonary Angiography: Using Decision Rules in the Emergency Department.

PURPOSE: The aim of this study was to assess the appropriateness of utilization and diagnostic yields of CT pulmonary angiography (CTPA), comparing two commonly applied decision rules, the pulmonary embolism (PE) rule-out criteria (PERC) and the modified Wells criteria (mWells), in the emergency department (ED) setting. METHODS: Institutional review board approval was obtained for this HIPAA-compliant, prospective-cohort, academic single-center study. Six hundred two consecutive adult ED patients undergoing CTPA for suspected PE formed the study population. The outcome was positive or negative for PE by CTPA and at 6-month follow-up. PERC and mWells scores were calculated. A positive PERC score was defined as meeting one or more criteria and a positive mWells score as >4. The percentage of CT pulmonary angiographic examinations that could have been avoided and the diagnostic yield of CTPA using PERC, mWells, and PERC applied to a negative mWells score were calculated. RESULTS: The diagnostic yield of CTPA was 10% (61 of 602). By applying PERC, mWells, and PERC to negative mWells score, 17.6% (106 of 602), 45% (273 of 602), and 17.1% (103 of 602) of CT pulmonary angiographic examinations, respectively, could have been avoided. The diagnostic yield in PERC-positive patients was higher than in mWells-positive patients (10% [59 of 602] vs 8% [49 of 602], P < .0001). Among PERC-negative and mWells-negative patients, the diagnostic yields for PE were 1.9% (2 of 106) and 4% (12 of 273), respectively (P = .004). The diagnostic yield of a negative PERC score applied to a negative mWells score was 1.9% (2 of 103). CONCLUSIONS: The use of PERC in the ED has the potential to significantly reduce the utilization of CTPA and misses fewer cases of PE compared with mWells, and it is therefore a more efficient decision tool.