International Research Center for Medicinal Administration, Peking University
Dept. of Real-World Evidence and Pharmacoeconomics

Author Of 1 Presentation

Machine Learning/Network Science Poster Presentation

P0010 - Integration of the Extreme Gradient Boosting model with clinical data to enable the early diagnosis of multiple sclerosis (ID 715)

Speakers
Presentation Number
P0010
Presentation Topic
Machine Learning/Network Science

Abstract

Background

Delayed multiple sclerosis (MS) diagnoses are not uncommon, especially in Asia, where MS is relatively rare. Therefore, an early MS diagnostic tool is urgently warranted.

Objectives

We aimed to develop an effective tool through machine learning techniques for the early diagnosis of MS in Chinese people.

Methods

Two case sets were established. The first MS cohort had 239 cases and 1142 controls (the training set), and the second MS cohort had 23 cases and 92 controls (the test set). The Extreme Gradient Boosting (XGBoost), Random Forest (RF), Naive Bayes, K-nearest-neighbor (KNN), Support Vector Machine (SVM) algorithms were fitted using Bayesian optimization, and the best parameter sets were assessed using the F1 scores of 5-fold cross-validations. The utility of machine learning algorithms in MS early diagnosis was evaluated by precision, recall, specificity, accuracy and F1 score through 5-fold cross validation.

Results

The XGBoost algorithm performed better than the other algorithms in 5-fold cross-validation. Thirty-four variables were set for the XGBoost algorithm, and their relative importance with MS were ranked. The training set recall was 0.632, with a specificity of 0.903, and the test set recall was 0.609, with a specificity of 0.902. Our study found that 61%, 51%, and 49% of the patients could be diagnosed with MS, 1, 2, and 3 years earlier than their real diagnostic time point, respectively.

Conclusions

A diagnostic tool for early MS recognition based on the XGBoost model and medical record data were developed to help reduce diagnostic delays in MS patients.

Collapse