Keywords: FOCUS, Rapid Cardiac Assessment, AI ultrasound support, multiparametric ultrasound cardiac assessment, POCUS in frontline physicians
Background:
FOCUS could be a complement to the clinical exam for evaluating the structural and functional abnormalities of the heart. Just a few studies have assessed the value and accuracy of focused cardiac ultrasound (FOCUS) performed by family physicians.
Aim of the study:
This study aimed to evaluate the diagnostic accuracy of FOCUS performed by family doctors compared to cardiologists' evaluation as a "Gold-Standard-method" on patients with a high cardiovascular risk and to find a new AI-supported-algorithm for early detection of heart-pathology.
Methodology:
We conducted a prospective-observational-cross-sectional-study on1780-patients with high-risk(Score2/OP2). Patients were first examined by family physicians(FP) with expertise and subsequently by cardiologists to determine the accuracy of this application. We used the five standard-cardiac-scans:Subxiphoid-view,/Parasternal-long/short-axis,/Apical-four-chamber-view,/IVC-assessment with AI-support(self-detected-LVEF). We have developed a computerized diagnostic supported by AI-Software. The agreement was evaluated using Cohen’s-kappa coefficient. We did a Logistic regression analysis to assess the impact of clinical-ultrasound variables on cardiovascular risk and to test the performance of the AI algorithm in the cardiovascular risk prediction model based on independent variables (risk factors/clinical/ultrasound parameters).
Results:
We identified 585 patients with pathology subsequently confirmed by the cardiologist. We did the descriptive analysis. The accuracy was 94.33%, with a sensitivity-89.91%, specificity-96.49%and prevalence-32.87%. We did the Chi-square(χ²)test to compare the results of the AI-combined FOCUS algorithm proceeded by FP with traditional clinical assessments. The statistical significance level was very high, p<0.0001. Reports of the two groups for identifying cardiac pathology showed very good agreement(k=0.88; 95%CI=0.81–0.95),standard-error:0,037. The logistic-regression showed strong-significance-predictors like: hypertension/LVEF<50%(AI)/valvular-abnormalities with p-values(<0.0001).
Conclusions:
FOCUS performed by trained-FP AI-assisted was comparable to cardiologist's results. The logistic regression reveals that several factors significantly predict high cardiovascular risk including hypertension/diabetes/smoking/clinical symptoms/AI-assisted-LVEF/and atheromatous-plaque on the carotid artery. . The AI-assisted LVEF measurement, in particular, shows a strong association with cardiovascular risk, highlighting the importance of AI in assisting with the diagnosis in primary care settings.
#17