VITA-D: A Radiomic Web Tool for Predicting Vitamin D Deficiency Levels

Background: Vitamin D deficiency is a significant risk factor for several chronic conditions. This study aims to predict vitamin D deficiency levels in a private database, collected from the southern part of Loja-Ecuador using a graphical web interface tool based on artificial intelligence algorithms. Methods: Two databases were processed using ML training models: SVM, Random Forest (RF), Linear Regression (LR). (i) Private data collection was undertaken on 465 patients from a local university, where vitamin D levels were measured through a blood sample collection to calculate the concentration of 25-hydroxy vitamin D in plasma and determine it by enzyme-linked immunosorbent assay, and (ii) public data collection was obtained from the FigShare database. Then, a survey was conducted from April 2022 to June 2023, identifying 157 variables, 18 of which were used for ML training models. Results: Vitamin D deficiency levels in private patients reached 18.10 ng/mL and 20.42 ng/mL in the public. The RF algorithm achieved (87.73%) accuracy, the SVM (80.0%), and LR (70.70%). RF was selected as the best performance model for web application design in binary levels classification: deficiency (Class 0) indicates vitamin D levels below 15 ng/mL, and sufficiency (Class 1) indicates vitamin D levels above 15 ng/mL. Conclusions: The “VITA-D” web application was used to monitor and predict vitamin D levels and deficiency factor risk based on clinical and sociodemographic data, providing an efficient and cost-effective alternative to traditional vitamin D testing methods.