PreprintArticleVersion 1Preserved in Portico This version is not peer-reviewed
Comparative Performance of Autoencoders and Traditional Machine Learning Algorithms in Clinical Data Analysis for Predicting Post-Staged GKRS Tumor Dynamics
Version 1
: Received: 30 August 2024 / Approved: 30 August 2024 / Online: 2 September 2024 (11:20:28 CEST)
How to cite:
Volovăț, S. R.; Popa, T. O.; Rusu, D.; Ochiuz, L.; Vasincu, D.; Agop, M.; Buzea, C. G.; Volovăț, C. C. Comparative Performance of Autoencoders and Traditional Machine Learning Algorithms in Clinical Data Analysis for Predicting Post-Staged GKRS Tumor Dynamics. Preprints2024, 2024082245. https://doi.org/10.20944/preprints202408.2245.v1
Volovăț, S. R.; Popa, T. O.; Rusu, D.; Ochiuz, L.; Vasincu, D.; Agop, M.; Buzea, C. G.; Volovăț, C. C. Comparative Performance of Autoencoders and Traditional Machine Learning Algorithms in Clinical Data Analysis for Predicting Post-Staged GKRS Tumor Dynamics. Preprints 2024, 2024082245. https://doi.org/10.20944/preprints202408.2245.v1
Volovăț, S. R.; Popa, T. O.; Rusu, D.; Ochiuz, L.; Vasincu, D.; Agop, M.; Buzea, C. G.; Volovăț, C. C. Comparative Performance of Autoencoders and Traditional Machine Learning Algorithms in Clinical Data Analysis for Predicting Post-Staged GKRS Tumor Dynamics. Preprints2024, 2024082245. https://doi.org/10.20944/preprints202408.2245.v1
APA Style
Volovăț, S. R., Popa, T. O., Rusu, D., Ochiuz, L., Vasincu, D., Agop, M., Buzea, C. G., & Volovăț, C. C. (2024). Comparative Performance of Autoencoders and Traditional Machine Learning Algorithms in Clinical Data Analysis for Predicting Post-Staged GKRS Tumor Dynamics. Preprints. https://doi.org/10.20944/preprints202408.2245.v1
Chicago/Turabian Style
Volovăț, S. R., Călin Gheorghe Buzea and Cristian Constantin Volovăț. 2024 "Comparative Performance of Autoencoders and Traditional Machine Learning Algorithms in Clinical Data Analysis for Predicting Post-Staged GKRS Tumor Dynamics" Preprints. https://doi.org/10.20944/preprints202408.2245.v1
Abstract
This study investigates the comparative performance of autoencoders and traditional machine learning (ML) algorithms in predicting tumor dynamics following Gamma Knife radiosurgery (GKRS) for brain metastases (BMs). The research retrospectively analyzed clinical data from 77 patients (median age: 64 years; 45 males, 32 females) treated at the “Prof. Dr. NicolaeOblu” Emergency Clinic Hospital-Iasi. The dataset comprised 12 variables, including socio-demographic, clinical, treatment, and radiosurgery-related factors. The primary outcome was tumor progression or regression within three months post-GKRS, with 71 cases of regression and 6 cases of progression observed. Traditional ML models—Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Extra Trees, Random Forest, and XGBoost—were evaluated, achieving accuracy rates ranging from 0.91 (KNN) to 1.00 (Extra Trees). When incorporating features extracted by autoencoders, the performance of these models generally improved, especially with compression in the bottleneck layer. For instance, Logistic Regression's accuracy increased from 0.91 to 0.94, while SVM's accuracy improved significantly from 0.85 to 0.96. XGBoost consistently performed well across both traditional and hybrid setups, maintaining an accuracy of 0.94 with an AUC of 0.98. The findings suggest that hybrid approaches combining deep learning with traditional ML techniques can significantly enhance predictive accuracy and robustness, particularly in the context of GKRS for BMs. This improved performance has potential implications for more informed clinical decision-making in personalized medicine.
Medicine and Pharmacology, Neuroscience and Neurology
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.