Disentangling Synthetic Edge Artifacts and Spatial-Context Sensitivity in Brain MRI Tumour Classification: A Multi-Control Spatial Intervention Study
Year: 2026
Type: software
Abstract: Research Release v1.0.0 — Brain MRI Dataset Preparation and Audit This release supports reproducible research on brain MRI tumour classification, with a focus on synthetic edge artifacts, spatial-context sensitivity, and controlled spatial interventions. Research Scope Prepare the Cheng Brain Tumor Dataset for systematic analysis. Extract MRI images, tumour masks, and related annotations while preserving source data. Generate metadata, dataset inventories, duplicate checks, and quality-control reports. Create reproducible patient-grouped cross-validation splits where supported by the available patient identifiers. Document dataset limitations and potential confounding factors for future controlled experiments. Reproducibility and Validation The Python-based workflow includes dataset preparation, cross-validation inspection, patient-split generation, and validation scripts. It produces structured outputs and audit reports to support transparent, repeatable experimentation. Intended Use This release is intended for academic research, educational purposes, dataset preparation, and experimental development. It does not provide a clinically validated diagnostic system, and it must not be used for medical diagnosis or treatment decisions. Research Disclaimer Results from future experiments must be interpreted carefully. Changes in model predictions under spatial interventions do not, by themselves, establish causal tumour reliance or clinical validity. Researchers should report dataset limitations, preprocessing choices, evaluation methodology, and potential sources of bias. Release: v1.0.0 Project: Brain Tumor Prediction Using Deep Learning Focus: Reproducible dataset preparation, quality assurance, and controlled research on spatial-context sensitivity.
Institution: Manipal University Jaipur
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7223246331