RYGB Sucess Estimator (RYSUES)

Introduction

This tool was developed as a joint effort of the Metabolic Research Laboratory, Clínica Universidad de Navarra and Tecnun School of Engineering, Universidad de Navarra, to help healthcare professionals in the estimation of the likelihood of Roux-en-Y bariatric surgery success based on patient-specific pre and perioperative data (see below for the needed input variables). By leveraging three Machine Learning (ML) models to choose from, this tool provides success/failure predictions along with probability success score. These models have been trained using surgery outcome at 18 months measured by the following two metrics:

Prediction based on Body Mass Index (BMI) metric, computed as weight in kilograms divided by the square of height in meters. A BMI equal to or less than 30 kg/m² is considered success (ML models based on BMI are referred to as Model 1 & Model 2).

Prediction based on Excess Weight Loss (EWL) metric, calculated according to the formula EWL = (body weight (BW) before – BW after)/excess BW before, where excess BW = total BW – ideal BW. An EWL% equal or greater than 65% for males and 75% for females is considered success (ML model based on EWL is referred to as Model 3).
Research group
Research Team

members of the research group from left to right

  • Dra. Idoia Ochoa Álvarez
  • Rocío Marugán Pinos
  • Dr. Javier Gómez Ambrosi
  • Dra. Gema Frühbeck

How it works

Option 1: multiple patient prediction

1.  Prepare your CSV file containing patient data. ML models Model 1 & Model 2, based on BMI, need as input 16 variables (see provided test_data_1_2.csv file in GitHub for an example). ML model Model 3, based on EWL, needs as input 58 variables (see provided test_data_3.csv file in GitHub for an example). The tool expects a .csv file following the structure of the provided examples. In the first case, predictions for Model 1 and Model 2 will be computed; whereas in the second case only predictions for Model 3 will be computed. In all cases, the models will analyze the input and predict the probability of treatment success.

2.  Upload your CSV file to the dropzone on the right.

3.  Results of the predictions will be added as new columns in a newly generated .csv (one per model run), including:

  • Computed probability of success (0 - 1).
  • Final predicted label: 1 for “success” and 0 for “failure”.
  • Final written label: “success” or “failure”.

Option 2: single patient prediction

1. Manually introduce each variable’s values for one patient in the windows included below.This option will only provide a prediction using Model 1 & Model 2, with their 16 input variables. Decimal used must be a dot “.”.

2. Press “send” and the prediction will appear on screen

Important note: 

Variables:

“Roux Limb” and “Treitz” are perioperative anatomical landmarks, that is, defined during the surgery. Please enter in your .csv file whichever value you would like to evaluate for each individual patient, taking into account that a higher value means a more aggressive surgery. In particular:

  • Roux Limb (alimentary limb): part of the small intestine that bypasses the stomach and duodenum. It connects the gastric pouch with the anastomosis with the biliopancreatic limb to allow contact between food and digestive enzymes. Value usually ranges from 40 to 400 cm. A higher value indicates a more aggressive surgery.
  • Treitz: Part of the jejunum bypassed until the point of biliopancreatic limb anastomosis. Value usually ranges from 30 to 170 cm.

📂 Download your file example to run the model !

📂 Upload your file to get started!

...

Single Patient

This options is only for running the Models 1 & Model 2 with only 1 patient

1 Fat mass: Total weight of fat tissue in the body.

2 Waist circumference: Measurement around the abdomen, used to assess abdominal fat.

3 Hip circumference: Measurement around the widest part of the hips.

4 REE (r): Measured resting energy expenditure — the energy your body uses at rest.

5 REE (t): Theoretical resting energy expenditure — estimated energy use at rest based on the Harris-Benedict equation.

6 Basal glycemia: Blood sugar level after fasting, usually measured in the morning.

7 120' glycemia: Blood sugar level 120 minutes after consuming glucose (OGTT).

8 HOMA index: Estimate of insulin resistance, calculated from fasting glucose and insulin.

HOMA-IR = (Fasting insulin [µU/mL] × Fasting glucose [mg/dL]) / 405.

9 Triglycerides: A type of fat in the blood, high levels may increase heart disease risk.

10 Leptin: Hormone that helps regulate appetite and energy balance.

11 TSH: Thyroid-stimulating hormone, controls thyroid function and metabolism.