Session Title: Virtual Poster Hall
Session Time: None. Available on demand.
Disclosures: Dae Youp Shin, n/a: No financial relationships or conflicts of interest
Objective: To find the most relevant predictor for DSPN among patients with type 2 DM using machine learning (ML) algorithms, and whether ML-based method is better than statistics for the prediction of DSPN.
Design: Retrospective case control study Setting : 434 DM patients were analyzed, and patients who had polyneuropathies other than DSPN, mononeuropathies, or radiculopathies were excluded. Subjects were divided into two groups according to the electrophysiological results based on the guidelines of the American Diabetes Association. Participants : 434 patients were analyzed. DSPN group (n=118) and control group without DSPN (n=316).
Interventions: ML was performed using XGBoost (XGB), Support Vector Machine (SVM) and Random Forest (RF). The initial and mean values, and pattern of change in each laboratory data were included. In order to compare the predictability of ML results, general statistical methods were also carried out as follows; independent t-test, chi-square test, factor analysis, and regression analysis.
Main Outcome Measures: A total of 120 parameters including clinical data (n= 30) and laboratory data (n= 90) were analyzed.
Results: When we combined XGB and RF, area under the curve (AUC) and accuracy were higher (0.8340 and 81.163% respectively) than any other single algorithms or combination of each other. When we selected 30 parameters in order of highest important score, 0.8526 AUC and 81.86% accuracy were achieved. Disease duration, Hemoglobin A1c, urine microalbumin, Body Mass Index were the most relevant predictors. The statistical methods showed that no meaningful parameter of DSPN was not derived. Conclusions: We confirmed that the ML algorithm, especially the combination of XGB and RF, was effective in predicting DSPN, which was not achieved by general statistical methods. ML-based diagnostic prediction would be useful for screening polyneuropathies in diabetic patients.
Level of Evidence: Level I
To cite this abstract in AMA style:Shin DY, BORA L, YOO WS, PARK JW, Kim T, Lee SJ, Kim SY, Hyun JK. Machine Learning-Based Prediction of Diabetic Sensorimotor Polyneuropathy (DSPN) [abstract]. PM R. 2020; 12(S1)(suppl 1). https://pmrjabstracts.org/abstract/machine-learning-based-prediction-of-diabetic-sensorimotor-polyneuropathy-dspn/. Accessed October 23, 2021.
« Back to AAPM&R Annual Assembly 2020
PM&R Meeting Abstracts - https://pmrjabstracts.org/abstract/machine-learning-based-prediction-of-diabetic-sensorimotor-polyneuropathy-dspn/