SPC365

Improving pacemaker implantation prediction after TAVR: creation and validation of a machine learning based model

Session: SESSÃO DE COMUNICAÇÕES ORAIS 15 - INTELIGÊNCIA ARTIFICIAL EM CARDIOLOGIA: APROVEITAR O POTENCIAL!
Speaker: Francisco Barbas De Albuquerque

Congress: CPC 2025
Topic: N. E-Cardiology / Digital Health, Public Health, Health Economics, Research Methodology
Theme: 33. e-Cardiology / Digital Health
Subtheme: 33.3 Computer Modeling and Simulation
Session Type: Comunicações Orais
FP Number: ---
Authors: Francisco Barbas De Albuquerque; Miguel Marques Antunes; Tomás Barbas de Albuquerque; Barbara Teixeira; André Grazina; Fernando Ferreira; Inês Rodrigues; António Fiarresga; Rúben Ramos; Rui Ferreira; Duarte Cacela; Mário Oliveira

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><strong><span style="font-size:10.0pt">Background</span></strong></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">Pacemaker (PM) implantation (I) is a common complication after TAVR. Artificial intelligence (AI)- and machine-learning (ML) technologies may contribute to developing better prediction models in this clinical context.</span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><strong><span style="font-size:10.0pt">Aim</span></strong></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">To develop a ML-based Binary Classification Model for predicting PMI after TAVR, compare it with a regression-based model and validate it in a prospective cohort.</span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><strong><span style="font-size:10.0pt">Methods</span></strong></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">Single-center retrospective study on patients (P) that underwent TAVR between 2018 and 2024.&nbsp; A full review of demographic, clinical, electrocardiographic, echocardiographic, cardiac CT scan and intra-procedural data was performed. Both pre- and intra-procedural variables were included in the dataset to train the model.&nbsp;</span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">A Python script was developed to build a Binary Classification model. Due to dataset imbalance, a SMOTE-based upsampling technique was performed on the minority class. The XGBoost (eXtremeGradient Boosting) open-source software library and algorithm was used to train the ML-based prediction model. To achieve better performance, we implemented an Ensemble Model approach consisting of 21 Binary Classifiers. For each P, the final prediction was determined by aggregating the predictions from all classifiers and selecting the most frequently predicted value. Both testing and validation model performance metrics were computed using the confusion matrix of predictions and are as follows: weighted precision (WP), weighted recall (WR) and weighted f1-score (WF1). In addition, a logistic regression was executed for performance comparison between models. ROC curves AUC were developed for both models. </span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><strong><span style="font-size:10.0pt">Results</span></strong></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">From a total of 770 TAVR procedures during the study period, 611 P entered the analysis. Mean age was 82 years and 44% were male. PM implantation occurred in 170 (27.8%) P. </span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">Using our XGBoost Ensemble ML algorithm a scoring model was generated.The highest weighted variables were the presence of right bundle branch block, QRS duration, peripheral artery disease, male gender and left bundle branch block (figure 1A).</span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">The ML-based model performance metrics were: WP of 58,47%, WR of 59,07% and WF1 of 58,69%. The logistic regression model had the following metrics: WP of 48,45%, WR of 54,80% and WF1 of 51,43%. The XGBoost AUC was 0,73 and the LogRegression AUC was 0.63 (Figure 1B)</span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">Seventy-one P enter the prospective validation cohort. PMI occurred in 23 (32%) P. The metrics from our ML-based model in the validation cohort were: WP of 66,17%, WR of 64,48% and WF1 of 65,42%.</span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">The metrics from logistic-regression based model were: WP of 58,22%, WR of 52,28% and WF1 of 55,09%.</span></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><strong><span style="font-size:10.0pt">Conclusion</span></strong></span></span></p>

<p style="text-align:justify"><span style="font-size:12pt"><span style="font-family:Aptos,sans-serif"><span style="font-size:10.0pt">We created and validated a ML-based prediction model for PMI after TAVI. This model outperformed the traditional used regression-based model. This underscores the move towards a more personalized medicine, where AI and ML-based models may enhance clinical decision-making for better patient outcomes.</span></span></span></p>