Data for AI Uncertainty Quantification in Radiotherapy Applications Scoping Review

<b>Datasheets for Uncertainty Quantification in Machine Learning Radiotherapy Applications: A Scoping Review.</b><br>This dataset accompanies the scoping review titled <b>"Uncertainty Quantification in Machine Learning Radiotherapy Applications: A Scoping Review"</b> (preprint link: https://www.medrxiv.org/content/10.1101/2024.05.13.24307226v1; now accepted at Radiotherapy and Oncology, in press). <br>Contact: Kareem Wahid, PhD; <b>[e-mail]</b>. The scoping review systematically explores the current landscape of uncertainty quantification (UQ) in machine learning (ML) applications within radiotherapy (RT), identifying existing trends, areas for improvement, and future directions in this emerging field.<br>Included files:<b>final_manuscript_data.csv</b>: This CSV file contains extracted data from the 56 studies included in the scoping review. The data encompasses four main categories. The General Study Characteristics include information on publication type, year, geographic origin, and code or data availability. The Radiotherapy Characteristics provide details on radiotherapy application domains—such as auto-contouring and dose prediction—the disease sites studied, and the types of data used, including CT scans, MRI images, and RT dose information. The AI Characteristics section offers information about the machine learning approaches employed, the training, validation, and testing sample sizes, and the validation methods used. Lastly, the Uncertainty Quantification Characteristics detail the UQ methods used, the uncertainty metrics reported, the types of uncertainty addressed (aleatoric vs. epistemic), and the UQ applications investigated, such as failure detection and calibration.<b>overlap_data.xlsx</b>: This Excel file provides sheets for our review and previous systematic/scoping reviews related to UQ in medical applications where Study IDs, Titles, and digital object identifiers are used for comparative overlap analysis. These files are to be used in conjunction with the corresponding code repository at: https://github.com/kwahid/RT_UQ_scoping_review/tree/main.<br>