Abstract
1628
Introduction: Artificial intelligence tools are emerging as supporting tools for physicians because their ability to improve the workflow and to reduce systemic errors specially in busy hospitals. Myocardial perfusion scintigraphy (MPS) is the most frequent exam in many nuclear medicine departments. Our aim was to evaluate the MPS diagnostic agreement between the report by medical specialists and by an artificial intelligence software (AIsR), checking its specificity and applicability in the population of the Nuclear Medicine department of an University Hospital. Method: 51 consecutive patients previously analyzed with QPS/QGS software were consecutively included. The reports by two specialist physicians were retrospectively compared to the reports automatically generated by ECTb V4.0, using the AIsR tool. We analyzed whether the overall examination was normal or abnormal, as well as all 17 myocardial segments represented on the polar map, which could also be classified as normal or abnormal, and the abnormal ones as affected by ischemia or fibrosis. The AIsR has an High-Specificity (SP), High-Sensitivity (SN), or TradeOff setting for the automatic analysis that was evaluated independently. It is an ongoing study and the researchers who are responsible for the study also intend to increase its population to 100 patients. Results: About 39.21% of the patients in the population were men, 88.23% had systemic arterial hypertension, 35.29% had diabetes mellitus, 39.21% had had acute myocardial infarction and 28.43% were smokers or former smokers. A prevalence of 47,1 % for CAD was estimated. When the AIsR tool was set to TradeOff, there was an agreement of 70,6 %, specificity of 92,6 %, and sensitivity of 45,8 %. It was considered the best setting for this population. For SN and SP, respectively, the following was estimated: agreement (66,7 % and 66,7 %), specificity (81,5 % and 92,6 %), and sensitivity (50,0 % and 37,5 %). Conclusions: The report created by this artificial intelligence tool has satisfactory sensitivity and agreement when compared to the report created by medical specialists, especially for TradeOff setting. Its role for workflow improvement was demonstrated.
