Artificial Intelligence-Driven Blood Loss Prediction in Large-Volume Liposuction: Enhancing Precision and Patient
Mauricio E Perez Pachon1,2, Jose T Santaella3,4, Carlos Oñate5,6
1From the Department of Plastic Surgery, Total Definer.
Background:
Over 2.3 million liposuctions are performed annually with a complication rate of approximately 5%, including a death rate of 1 in 5000 because of blood loss. Artificial intelligence models offer potential for improving blood loss prediction and management in these procedures, analyzing extensive data to identify risk factors and accurately estimate blood loss.
Methods:
Data from 721 large-volume liposuction patients at 2 centers in Bogotá, Colombia, and Loja, Ecuador, between 2019 and 2023 were evaluated. Both centers followed identical perioperative protocols. The data set was split into training (621 patients) and testing (100 patients) sets. A supervised machine learning model was trained to predict blood loss. Model predictions were compared with clinical data using statistical validation metrics.
Results:
Most patients were women (79.2%), with median values of age of 37 years; weight, 65 kg; height, 165 cm; body mass index, 24.34 kg/m²; volemia, 3924.41 mL; infiltrated volume, 5800 mL; and aspirated volume, 3900 mL. Previous liposuction was noted in 32%. No significant differences were found between training and testing cohorts. The model achieved a mean absolute error of 22.09 mL, root mean square error of 34.13 mL, and an R ² value of 0.974, indicating high predictive accuracy and excellent model fit.
Conclusions:
The authors' study has developed and validated an accurate artificial intelligence-based model to predict blood loss in large-volume liposuction, showing 94.1% accuracy. The authors' model enhances preoperative planning and intraoperative management, potentially reducing complications and improving outcomes.
