In a recent report, published in Vascular Medicine, researchers evaluated whether a deep neural network-based machine analysis using resting Doppler waveforms was viable for peripheral artery disease detection. The study’s authors, led by Robert D. McBane II, were motivated to improve accurate peripheral artery disease (PAD) identification to help reduce major adverse cardiac and limb events. The researchers ultimately reported that their predictive model was able to identify PAD at a clinically relevant level.
The study enrolled a total of 3,432 patients undergoing rest and post-exercise ankle-brachial index (ABI) testing between 2015 and 2020, of which 1,941 had PAD and 1,491 did not. Patients were randomized to training, validation, and testing cohorts for development of predictive models.
The validation cohort consisted of patients who underwent testing between January 1, 2021 and March 31, 2021, after the model had been created. Authors summarized that the “deep neural networks were trained on resting posterior tibial arterial Doppler waveforms to predict normal (>0.9) or PAD (≤0.9) using rest and post-exercise ABI.”
Reportedly, the authors’ most effective predictive model demonstrated an area under the curve (AOC) of 0.94 (95% confidence interval [CI], 0.92–0.96), as well as a sensitivity of 0.83, specificity of 0.88, accuracy of 0.85, and a positive predictive value (PPV) of 0.90. After additional analyses, authors reported that “results were similar for the validation dataset: AUC 0.94 (95% CI, 0.91–0.98), sensitivity 0.91, specificity 0.85, accuracy 0.89, and PPV 0.89 (post-exercise ABI comparison).”
Ultimately, the authors “an artificial intelligence-enabled analysis of a resting Doppler arterial waveform permits identification of PAD at a clinically relevant performance level.”
View More Related Articles and Videos at the Atherosclerotic Disease Knowledge Hub