3696 - MAMBA Toolbox: A Statistical Voxel-Based Analysis Solution for Radiation Therapy
Presenter(s)
G. Palma1, L. Cella2, and S. Monti2; 1National Research Council, Institute of Nanotechnology, Lecce, Italy, 2National Research Council, Institute of Biostructures and Bioimaging, Napoli, Italy
Purpose/Objective(s): To design a toolbox to enabling flexible and comprehensive statistical voxel-based (VB) analysis in radiation oncology. We hypothesized that the proposed MAMBA (Multi-pAradigM voxel-Based Analysis) will facilitate the investigation of spatially resolved dose-response relationships by providing robust statistical modeling tools for VB statistical analysis of radiation therapy (RT) data, assisting in the identification of spatial correlations between radiation dose and treatment outcomes.
Materials/Methods: MAMBA is implemented in a programming environment and offers an open-source suite of functions (https://github.com/pippipalma/MAMBA) for performing VB statistical modeling based on various regression schemes. It generates statistical VB maps reflecting the observed significance level by utilizing non-parametric permutation inference. The toolbox supports the inclusion of VB and global clinical outcomes, alongside an arbitrary number of explanatory variables. The Computing Toolbox in a programming environment is leveraged to optimize performance, exploiting the parallelizable nature of most computational tasks.
Results: The MAMBA toolbox provides an efficient and accessible platform for conducting advanced VB statistical analyses in RT research combining statistical and computational features within the same toolbox that are not found in other similar open-source toolboxes. It was successfully tested across various RT datasets enabling the detection of spatial patterns in dose-response relationships in different clinical scenarios. Through its implementation of parallel computing, MAMBA significantly reduces processing times, making large-scale analyses feasible. The availability of the source code ensures adaptability, allowing experienced users to integrate and expand the toolbox functionalities according to their requirements.
Conclusion: MAMBA offers a powerful, open-source solution for VB statistical analysis in RT research, enabling scientists to perform sophisticated spatial data analyses without requiring extensive programming expertise. Its flexibility, computational efficiency, and ease of use make it a valuable tool for advancing radiation oncology research by providing deeper insights into radiobiology patterns and at the same time allows standardization of VB methods opening the way to large validation studies.