Plastic and Reconstructive Surgery: Artificial Intelligence Predicts Blood Loss and Could Make Liposuction Safer

Scientists have developed an artificial intelligence (AI) model capable of predicting blood loss volume in patients undergoing large-volume liposuction with high accuracy, which could significantly improve the safety of this widely performed cosmetic procedure. Liposuction is one of the most common plastic surgery operations worldwide (more than 2.3 million procedures performed annually), and although it is generally considered safe, excessive blood loss remains a serious risk—especially when large volumes of fat and fluid are removed.

The study was published in Plastic and Reconstructive Surgery, the official journal of the American Society of Plastic Surgeons (ASPS), and was conducted by an international team of researchers.

The AI model was trained on data from 721 patients who underwent liposuction involving the removal of more than four liters of fat and fluid in total at two clinics that followed identical surgical protocols. The researchers used a wide range of demographic, clinical, and surgical variables to enable the algorithm to predict expected blood loss. The data were randomly split: 621 patients were used to train the model, and the remaining 100 were used to test its accuracy.

The results were impressive: the AI model predicted blood loss volume with approximately 94% accuracy. The average discrepancy between predicted and actual volumes was small (a root mean square error of about 26 ml), and the maximum difference was approximately 188 ml. Such accuracy allows surgeons to assess risks in advance and plan procedures based on the individual characteristics of each patient.

The authors note that AI-based predictions can assist in making critical perioperative management decisions, such as determining the need for blood transfusions, managing fluid balance, and implementing other patient care measures. This, in turn, may reduce complications, improve postoperative recovery, and enhance the quality of informed consent.

The researchers plan to further develop the model by incorporating data from surgeons around the world, with the goal of making the tool even more universal and accurate across different clinical settings.