Computer-Assisted Detection Of Proximal Arterial Stenosis On Doppler Ultrasound

The primary aim of this study was to determine if use of a novel computer-generated quantitative measure, effective acceleration time (effAT), can improve diagnostic accuracy for detecting a proximal arterial stenosis on spectral Doppler ultrasound. We hypothesized that use of an effAT cutoff value would have superior accuracy for detecting stenosis compared to qualitative Doppler waveform assessment. This was a retrospective, case-control study, whereby aortic stenosis (AS) was used as a model to detect distal tardus parvus (TP) physiology. Patients with echocardiography-confirmed AS (n=132; 60 mild, 44 moderate, 28 severe) and matched controls (n=48) who underwent carotid ultrasound within 90 days were identified through a diagnostic imaging database at a single medical center. A custom-built computerized spectral analysis program generated effAT values for spectral Doppler waveforms in the extracranial carotid arteries and a receiver operating characteristic (ROC) analysis was performed to determine the optimal median effAT cutoff value to detect AS. Two radiologists, blinded to subject disease status, reviewed all carotid sonograms for presence of TP waveforms. Inter-rater variability was measured, and the accuracy of the radiologists to detect AS with and without use of the effAT cutoff was calculated. There were no significant differences between cases and controls with regards to age, sex, body mass index, or ejection fraction. Accuracy of radiologist detection of AS via waveform interpretation ranged from 43-61%. Inter-rater agreement in the detection of TP waveforms was 76% (136/180 cases, K=0.44, p\u3c0.001). ROC analysis revealed an optimal effAT cutoff of greater than or equal to 48 ms to detect AS with a corresponding area under the curve of 0.77 (95% CI: 0.75-0.84). Use of the effAT cutoff independent of radiologist waveform interpretation demonstrated an accuracy of 72%. Radiologist detection of a proximal arterial stenosis though visual interpretation of spectral Doppler waveform morphology is limited by low accuracy and moderate inter-observer variability. Use of a computer-generated median effAT cutoff value markedly improves diagnostic accuracy and eliminates observer variability