State of the Practice for Medical Imaging Software

• DOI: 10.48550/arxiv.2405.12171
• Publication Date: 2024-05-20
• Authors: Spencer Smith; Ao Dong; Jacques Carette; Michael D. Nosewo...
• Keywords: I.4.0; Software Engineering (cs.SE); FOS: Computer and information sciences; D.2.7; I.4.0; Computer Science - Software Engineering; Computer Vision and Pattern Recognition (cs.CV); Computer Science - Computer Vision and Pattern Recognition; D.2.7

We selected 29 medical imaging projects from 48 candidates, assessed 10 software qualities by answering 108 questions for each project, and interviewed 8 of the development teams. Based on quantitative data, we ranked MI with Analytic Hierarchy Process (AHP). The four top-ranked products are 3D Slicer, ImageJ, Fiji, OHIF Viewer. Generally, is in a healthy state as shown following: observed 88% documentation artifacts recommended research guidelines, 100% use version control tools, developers appear to common quasi-agile process. However, current practice deviates existing guidelines because rarity some artifacts, low usage continuous integration (17% projects), unit testing (about 50% room improvement (six nine felt their was not clear enough). From interviewing developers, identified five pain points two potential concern: lack time, funding, technology hurdles, ensuring correctness, usability, maintainability, reproducibility. interviewees proposed strategies improve practice, address points, quality. Combining ideas ours, have following list recommendations: increase documentation, enriching datasets, usage, move web applications, employ linters, peer reviews, design change, add assurance cases, incorporate "Generate All Things" approach.