BENCHMARKING COEFFICIENTS FOR FORECASTING WEIGHT LOSS AFTER SLEEVE GASTRECTOMY BIOMEDICAL ENGINEERING


Çelik S. , Sohail A., Arif F., Özdemir A.

BIOMEDICAL ENGINEERING-APPLICATIONS BASIS COMMUNICATIONS, cilt.32, 2020 (ESCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 32 Konu: 1
  • Basım Tarihi: 2020
  • Doi Numarası: 10.4015/s1016237220500040
  • Dergi Adı: BIOMEDICAL ENGINEERING-APPLICATIONS BASIS COMMUNICATIONS

Özet

Background/Aim: In treatment practice of obesity, losing excess weight and then maintaining an ideal body weight are very important. By the sleeve gastrectomy initial weight loss is easier, but the progress of patients have diverse variability in terms of maintaining weight loss. Predicting models for weight changes may provide doctors and patients a good tool to modify their approach to obesity treatment.The main objective of this research is to verify the dependence of weight loss on sleeve coefficients and to forecast the weight loss. The weight loss and its dependence on remnant gastric volume compartmants (antral and body parts), after laparoscopic sleeve gastrectomy (LSG) is discussed in this paper. Data was obtained from a previous study which included 63 patients. Deep analysis of weight loss after LSG and its relation with remnant gastric volume is still a challenge due to weight loss dependence on multiple factors. During this research, with the aid of machine learning regression classifier, the relationship(s) between the sleeve coefficients' formulae and weight loss formulae (%EWL and %TWL), are developed in a novel way. Other factors such as age and gender are also taken into account. A robust approach of artificial intelligence, i.e. the "Neural Network Bayesian Regularization" is adopted to utilize the third month, sixth month and first year weight loss data, to forecast the second year weight loss. Models are proposed to demonstrate the dependance of total weight loss on crucial parameters of components of remanat gastric volumes. A comparative study is conducted for the appropriate selection of artificial intelligence training algorithm.