Liedl, David

Introduction: Abdominal aortic aneurysms (AAA) are common among individuals over age 65 with a prevalence of 9%. Most AAAs are small and require surveillance imaging to monitor size and growth rates. Tools capable of predicting growth rates would be clinically useful. Hypothesis: We investigated how computational hemodynamics derived from CT angiography (CTA) differentiated slow and rapid-growing AAAs. Methods: Using readily available computing power in the clinical workflow, 3D models were generated for 25 subjects with AAA from available 3D CTA data. AAA growth rates were categorized as slow ( less than 5 mm/year) or rapid (greater than 5 mm/year) based on serial imaging. A geometrical analysis was performed using 7 indices (Volume, Surface Area, Length, Max/Min radius, Expansion Ratio [Mean radius/vessel radius], Proximal vessel area, Voronoi diagram-based metrics) for each AAA. Computational hemodynamics were performed using commercial software (ANSYS-Fluent). Computational hemodynamics were completed for all cases (see Fig. 1). In-house software assessed wall shear stress-(WSS) and vortex-based hemodynamics. Support vector machine (SVM) was run with cross-validation (50 iterations) to assess predictive strength using these 16 parameters. Results: Among 25 AAAs studied, 7 were rapid and 18 were slow-growing. The combination of proximal vessel area, spatiotemporally averaged WSS, and spatially averaged OSI provided the best accuracy for growth differentiation. The area under receiving operating curve (AUROC) and total accuracy are 0.770 and 0.765, respectively, with 48% and 86% of rapid and slow-growing AAAs correctly identified, respectively. Conclusions: Computational hemodynamics appears feasible with reasonable growth rate prediction. With optimization, this tool may enable resource utilization refinement for imaging surveillance strategies by distinguishing rapid from slow-growing AAAs.

Background Patients with peripheral artery disease are at increased risk for major adverse cardiac events, major adverse limb events, and all‐cause death. Developing tools capable of identifying those patients with peripheral artery disease at greatest risk for major adverse events is the first step for outcome prevention. This study aimed to determine whether computer‐assisted analysis of a resting Doppler waveform using deep neural networks can accurately identify patients with peripheral artery disease at greatest risk for adverse outcome events. Methods and Results Consecutive patients (April 1, 2015, to December 31, 2020) undergoing ankle–brachial index testing were included. Patients were randomly allocated to training, validation, and testing subsets (60%/20%/20%). Deep neural networks were trained on resting posterior tibial arterial Doppler waveforms to predict major adverse cardiac events, major adverse limb events, and all‐cause death at 5 years. Patients were then analyzed in groups based on the quartiles of each prediction score in the training set. Among 11 384 total patients, 10 437 patients met study inclusion criteria (mean age, 65.8±14.8 years; 40.6% women). The test subset included 2084 patients. During 5 years of follow‐up, there were 447 deaths, 585 major adverse cardiac events, and 161 MALE events. After adjusting for age, sex, and Charlson comorbidity index, deep neural network analysis of the posterior tibial artery waveform provided independent prediction of death (hazard ratio [HR], 2.44 [95% CI, 1.78–3.34]), major adverse cardiac events (HR, 1.97 [95% CI, 1.49–2.61]), and major adverse limb events (HR, 11.03 [95% CI, 5.43–22.39]) at 5 years. Conclusions An artificial intelligence–enabled analysis of Doppler arterial waveforms enables identification of major adverse outcomes among patients with peripheral artery disease, which may promote early adoption and adherence of risk factor modification.

Introduction: Predictive algorithms using multiple clinical variables can identify patients at higher risk for abdominal aortic aneurysms (AAA), however they are cumbersome and cannot used directly by patients. A simple, cheap, noninvasive, and readily available (virtual) screening tool for identification of AAA could aid in screening algorithms. Aims: We sought to determine if a machine learning algorithm could predict the presence of an AAA by using a image of the face Methods: Diagnostic imaging studies were extracted from the electronic health record (EHR) from 5/7/2018 through 1/1/2023 and analyzed by a natural language processing algorithm previously validated to identify abdominal aortic aneurysms. Patient EHR profile pictures were extracted and matched to imaging studies. Various deep neural network (DNN) architectures were explored, all trained as classifiers on the face images along with clinical variables to predict clinical outcomes. All models used were CNNs with an EfficientNet architecture. Model performance was evaluated by standard metrics such as the area under the curve (AUC), specificities (Sp), and sensitivities (Sn). All performance metrics reported results from the test subset of the primary dataset and did not contain any observations considered during training. Results: A total of 5522 patients with facial images and AAA diagnostic studies were analyzed, among whom 1314 (23.8%) had a AAA. The optimal model had an AUC of 0.72 (Sn 0.44, Sp 0.83) for predicting AAA and remained similar in women (AUC 0.7), men (AUC 0.65), age >65 (AUC 0.63), and age <=65 (AUC 0.72) (Table). Conclusions: Using only an image of a patients face, the presence of an AAA could be identified with moderate performance in men, women, and in those less than 65. The model’s performance in demographics not typically screened for AAA may allow for targeted screening programs in these groups.

Introduction: Warfarin is a commonly prescribed anticoagulant for treating atrial fibrillation, mechanical valves, and venous thromboembolism. Warfarin dose management remains challenging due to dosing variability between patients and warfarin’s narrow therapeutic window. Time in therapeutic range (TTR) is critical to warfarin’s safety and efficacy, but TTR typically remains low (40-50%) in community practices. Specialized anticoagulation clinics and protocolized approaches can increase TTR but have great administrative burdens and health care costs. Aim: To develop standardized optimal warfarin dose decision support using a machine learning model based on time series anticoagulation data and patient demographic characteristics Methods: The dataset included 12,497 warfarin patients monitored in the anticoagulation tracker of electronic medical records across the Mayo Clinic Enterprise. The dataset contained time series anticoagulation data (warfarin dose, INR, INR target range) and patient demographic characteristics. We implemented an offline deep reinforcement learning model (DRL) to predict the cumulative warfarin dose for the days until the next INR test based on a patient’s historical anticoagulation data. Prior approaches utilized traditional supervised learning methods such as regression or Long Short Term Memory (LSTM) neural networks to approximate a function mimicking the behavior of the training data. On the other hand, DRL learned an optimal dosing policy through continuous interaction and feedback from the training data. The key advantage of DRL is the model can learn to behave differently (and potentially better) for suboptimal clinical states in the data such as overdosing or underdosing. To evaluate the DRL model we compared the predicted warfarin doses with the physician-prescribed doses Results: DRL model’s prediction accuracy was 96.96%, outperforming our implementation of a baseline LSTM model with a prediction accuracy of 70.58%. Further evaluation of the DRL model indicated that the model correctly adjusted the warfarin dose at time steps when patients had out-of-range INRs. Conclusions: Offline deep reinforcement learning demonstrates potential in supporting warfarin dose management to maximize TTR.

Summary: Background: Peripheral artery disease (PAD) is a risk factor for adverse limb events (LE) and cardiovascular events (CVE) that coexists with type 1 (T1) and 2 (T2) diabetes mellitus (DM). Little is known about comparative risk of LE and CVE in T1/T2 DM patients with PAD. Patients and methods: We queried our database of 40,144 patients ≥18 years old who underwent ankle brachial index (ABI) measurement from 01/1996-02/2020. We isolated T1/T2 DM patients with PAD diagnosed by ankle brachial index (ABI; low [<1.0] or elevated [>1.4]) and retrieved demographics including glycated hemoglobin (HbA1c). Primary outcomes were LE (critical limb ischemia/vascular amputation) and CVE (myocardial infarction/ischemic stroke). All-cause mortality was a secondary outcome. Multivariable Cox proportional regression yielded hazard ratios (HR) with 95% confidence intervals (CI) after adjusting for pertinent risk factors including age, hypertension, hyperlipidemia, smoking, and HbA1c. Results: Our study found 10,156 patients with PAD and DM (34% T1DM, 66% T2DM) with median follow-up time 34 mo (IQR 85 mo). T1DM patients were younger than T2DM (mean age 67 vs. 70 years), with higher median HbA1c (7.7 [IQR 1.9] vs. 6.7% [IQR 1.6]), and more prevalent hypertension, hyperlipidemia, CAD, and CKD. Antiplatelet and statin use was equivocal. Elevated ABI was more common in T1DM (47 vs. 28%). LE occurred in 23% and CVE in 12% patients. LE risk was higher in T1 than T2 DM patients (HR 1.58 [95% CI 1.44, 1.73], p<0.0001), but CVE and all-cause mortality were equivocal. These observations were preserved across ABI and HbA1c subgroup analyses. Conclusions: PAD patients with T1DM had a higher LE risk than those with T2DM, even after adjustment for glycemic control and pertinent risk factors, but CVE risk and all-cause mortality were equivocal. These data suggest a potential role for more intensive LE risk modification in PAD patients with T1DM, but further investigation is needed.

Introduction: In seniors, the prevalence of abdominal aortic aneurysms (AAA) is high (9-10%). We aim to characterize AAA growth rates (slow vs. rapid) to help inform management strategies, including individualized imaging surveillance frequency. Hypothesis: Modeling thrombosis may improve the characterization of AAAs' growth status. Methods: 3D geometrical AAA models with and without thrombosis (vessel lumen only) were generated for 70 human subjects using available contrast-enhanced CTA data. AAA growth rates were categorized as slow (< 5 mm/year) or rapid (≥ 5 mm/year) based on serial imaging. Python scripts were used to calculate geometrical parameters (83) with and without thrombosis. Patient-specific relevant health information was retrieved through a review of medical records. Support vector machine (SVM), a well-established machine learning method, was run with 10-fold cross-validation (100 iterations) to assess the predictive accuracy. Results: Among 70 AAAs studied, the ratio between rapidly-growing and slowly-growing was nearly 1:2. The combination of antihypertensive medication, coronary artery disease, juxta-aneurysm aortic size, and five geometrical parameters quantifying the extent of thrombosis provided the best accuracy for AAA's growth status. The area under receiving operating curve (AUROC) was 0.85, and total accuracy was 0.81, with 57% and 93% of rapidly- and slowly-growing AAAs correctly identified, respectively. Excluding patients with aneurysm thrombus, the AUROC and accuracy in predicting rapidly-growing AAAs decreased to 0.82 and 48%, respectively. Conclusions: machine-learning-based predictive modeling appears feasible for characterizing AAA's growth rate status. Manual aortic image segmentation, required for data analytics, is labor-intensive (on average, 1 hour). Leveraging artificial intelligence-based segmentation methodology may provide an automated solution to improve workflow and feasibility.

Introduction: Peripheral artery disease (PAD) is a risk factor for adverse limb events (LE) and cardiovascular events (CVE) that often coexists with type 1 (T1) and type 2 (T2) diabetes mellitus (DM), which differ mechanistically. There is a paucity of data on comparative risk of LE and CVE in T1DM and T2DM patients with PAD. Methods: We queried our database of 40,144 patients ≥18y old who underwent ankle brachial index (ABI) measurement from 01/1996-02/2020. We isolated patients with T1DM or T2DM and with PAD diagnosed by ABI (low [<1.0] or elevated [>1.4]). We retrieved demographics including hemoglobin A1c (A1c). Primary outcomes were LE (critical limb ischemia or vascular amputation) and CVE (myocardial infarction or ischemic stroke). All-cause mortality was a secondary outcome. Multivariable Cox proportional regression yielded hazard ratios (HR) with 95% confidence intervals (CI) after adjusting for age, sex, A1c, hypertension, hyperlipidemia, coronary artery disease (CAD), chronic kidney disease (CKD), and tobacco use. Results: Our study found 9,271 patients with PAD and DM (36% T1DM, 64% T2DM) with median follow-up time 34 mo (interquartile range [IQR] 85 mo). T1DM patients were younger than T2DM (mean age 67 vs 70y), with higher median A1c (7.7 [IQR 1.9] vs 6.7% [IQR 1.6]), and higher prevalence of hypertension, hyperlipidemia, CAD, and CKD. Antiplatelet and statin use did not differ. Elevated ABI was more common in T1DM (47 vs 27%). LE occurred in 23% and CVE in 12% of patients. LE risk was higher in T1DM patients than T2DM (HR 1.59[95% CI 1.45,1.74], p<0.0001), but CVE and all-cause mortality did not differ significantly. Conclusions: PAD patients with T1DM had a higher LE risk than those with T2DM, even after adjustment for A1c and pertinent risk factors, but CVE risk and all-cause mortality were equivocal. These data suggest a potential role for more intensive LE risk modification in PAD patients with T1DM, but further investigation is needed to minimize limb morbidity.

Summary: Background: Peripheral artery disease (PAD) impacts 3–12% of patients worldwide and is characterized by endothelial dysfunction and inflammatory pathways which are also common to venous thromboembolism (VTE), but there is a paucity of evidence regarding VTE risk in PAD patients. We investigated whether PAD is an independent risk factor for VTE. Patients and methods: We reviewed medical records of patients undergoing ABI studies at Mayo Clinic from 01/1996-02/2020. We classified patients by ABI (low [<1.0], normal [1.0–1.4], or elevated [>1.4]), as well as by specific low ABI subgroup: severely reduced (ABI: 0.00–0.39), moderately reduced (0.40–0.69), mildly reduced (0.70–0.90), and borderline reduced (0.91–0.99). The primary outcome was incident VTE event (acute lower extremity deep vein thrombosis or pulmonary embolism) after ABI measurement. Multivariable Cox proportional regression was used to calculate hazard ratios (HR) with 95% confidence intervals (CI) after adjusting for age, sex, active smoking, cancer, previous VTE, thrombophilia, anticoagulation, and revascularization. Results: 39,834 unique patients (mean age 66.3±14.3 years, median follow-up 34 months) were identified. 2,305 VTE events occurred in patients without PAD (13.0%), 2,218 in low ABI patients (13.0%), and 751 in elevated ABI patients (14.8%). After risk factor adjustment, VTE risk was modestly increased for PAD overall (HR: 1.12, 95% CI [1.06, 1.18]), including low ABI (HR: 1.11, 95% CI [1.04, 1.18]) and elevated ABI groups (HR: 1.15, 95% CI [1.04, 1.26]), compared to patients without PAD. The greatest VTE risk was in severely low ABI patients (HR: 1.46, 95% CI [1.31, 1.64]). Conclusions: In a large longitudinal cohort, we present strong clinical evidence of PAD, with low and elevated ABI, as an independent VTE risk factor, with the highest risk seen in patients with severely low ABI. Continued research is required to further investigate this relationship and its intersection with functional performance status to optimize VTE risk reduction or anticoagulation strategies in the PAD population.

Background: Patients with peripheral artery disease (PAD) are at increased risk for major adverse limb and cardiac events including mortality. Developing screening tools capable of accurate PAD identification is a necessary first step for strategies of adverse outcome prevention. This study aimed to determine whether machine analysis of a resting Doppler waveform using deep neural networks can accurately identify patients with PAD. Methods: Consecutive patients (4/8/2015 – 12/31/2020) undergoing rest and postexercise ankle–brachial index (ABI) testing were included. Patients were randomly allocated to training, validation, and testing subsets (70%/15%/15%). Deep neural networks were trained on resting posterior tibial arterial Doppler waveforms to predict normal (> 0.9) or PAD (⩽ 0.9) using rest and postexercise ABI. A separate dataset of 151 patients who underwent testing during a period after the model had been created and validated (1/1/2021 – 3/31/2021) was used for secondary validation. Area under the receiver operating characteristic curves (AUC) were constructed to evaluate test performance. Results: Among 11,748 total patients, 3432 patients met study criteria: 1941 with PAD (mean age 69 ± 12 years) and 1491 without PAD (64 ± 14 years). The predictive model with highest performance identified PAD with an AUC 0.94 (CI = 0.92–0.96), sensitivity 0.83, specificity 0.88, accuracy 0.85, and positive predictive value (PPV) 0.90. Results were similar for the validation dataset: AUC 0.94 (CI = 0.91–0.98), sensitivity 0.91, specificity 0.85, accuracy 0.89, and PPV 0.89 (postexercise ABI comparison). Conclusion: An artificial intelligence-enabled analysis of a resting Doppler arterial waveform permits identification of PAD at a clinically relevant performance level.