Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Radiomics Analysis and Prediction Using Image‑based Deep Learning (RAPID) for Clinical Diagnosis and Outcomes Modeling: Retrospective Study
ABSTRACT
Background:
Machine learning (ML)-based medical image analysis is increasingly prominent in surgical research. However, implementation often requires specialized expertise in artificial intelligence and software development, a substantial barrier for surgeon-scientists wishing to implement them for clinical use. Simple solutions that allow clinicians to perform rapid image analysis within a unified framework are lacking.
Objective:
In this context, we developed the RAPID software platform to facilitate image processing analysis with reduced technical burden while maintaining methodological rigor.
Methods:
RAPID is divided into two components. First is RAPID Core which contains routines to convert DICOM images into useable NIFTI format, automate segmentation of target anatomy using the open-source MONAI library, preprocess images with region of interest cropping and pixel preservation from segmentations, extract radiomics from preprocessed images using PyRadiomics, and train and test machine learning models. Second is RAPID GUI, a graphical user interface that organizes and parameterizes RAPID Core components to form a pipeline of tasks. RAPID GUI generates ready-to-use Python scripts for the RAPID pipeline that can be run on any computer containing RAPID Core.
Results:
We tested the RAPID software platform through two tasks. The first identified radiomic features(n=112) predictive of aortic aneurysm in a balanced cohort(n=1,114). The second identified features(n=225) of psoas muscles that correlate with sarcopenia as a surrogate for frailty and predictor of mortality for patients(n=116) undergoing thoracoabdominal aortic aneurysm repair. Pipelines of RAPID Core processes were designed for each of these tasks with RAPID GUI. Radiomics-based Extreme Gradient Boosting (XGBoost) models produced AUROC=0.956 and 0.929 for these tasks, respectively.
Conclusions:
RAPID is an accessible platform that shortens time from design to implementation. It enables surgeon-scientists with minimal computer programming experience to perform advanced radiomics and ML analysis, facilitating broader adoption of medical image analytics that enhances surgical research. Clinical Trial: n/a
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.