Abstract
Abstract Introduction The development of a predictive algorithm for determining total small intestine length using artificial intelligence represents a significant advancement in tailoring bariatric surgical techniques. A robust data registry is fundamental to validate the algorithm and ensure its effective application. This study aims to present a protocol (document, management, and centralization) for training the predictive algorithm. Methods Algorithm Development: The dataset (n=1090) was recorded in an Excel database using Filemaker for initial registration. Statistical analysis was conducted in Python (Google Colaboratory Notebook format). The methodology involved data transformation and scaling (using MinMaxScaler), clustering (unsupervised machine learning), data interpolation (machine learning with SMOTE oversampling), and algorithm modeling (XGBoost model via PyCaret). Data Recording Protocol An anonymous electronic document was created to centralize data collection. This document includes sociodemographic data, medical history, type of bariatric surgery, and measurements of intestinal loop lengths. New data and variables are progressively added to the database for validation and testing using the K-means clustering model. Results The registration document will present sociodemographic data, medical history, type of bariatric surgery, and the length of intestinal loops. Conclusion The development of a predictive algorithm for determining total small intestine length through machine learning requires a robust and comprehensive data collection process. A structured and simplified protocol ensures the efficient collection of multicentric data, supporting model training and validation. This framework facilitates the implementation of AI-driven individualization in bariatric surgery.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call