Early Detection of Mesothelioma Using Machine Learning Algorithms

Abstract

Early detection of mesothelioma, a severe form of cancer commonly associated with asbestos exposure, is a significant challenge that greatly affects prognosis. This study addresses this issue using Machine Learning (ML) algorithms, including Gradient-Boosted Trees (GBT), Support Vector Machines (SVM), and Logistic Regression (LR). This study uses the mesothelioma dataset from the UCI Machine Learning Repository to evaluate the proposed models, achieving 100% accuracy and F1 score in detecting the disease, accurately classifying 98 patients with 30 true positives and 68 true negatives. Further analysis using the AUC-ROC score showed that the ‘duration of symptoms’ feature was most informative for the GBT model, with a score of 0.595. In contrast, ‘C-reactive protein’ was the most significant feature for the SVM and LR models, each achieving an AUC-ROC of 0.603. Despite these promising results, validating these findings with additional datasets is critical to confirm their generalisability. However, this study provides strong support for using ML algorithms in the early detection of mesothelioma, potentially leading to improved patient diagnoses.

Keywords

• mesothelioma
• machine learning
• diagnosis
• gradient-boosted trees
• support vector machines
• logistic regression