Building a national repository of dural-based lesions: clinical, pathological, and demographic insights from the Indian population.

BACKGROUND

Meningiomas are the most common dural-based intracranial tumors, yet Indian literature is predominantly composed of limited single-center studies, restricting nationwide representation and data-driven decision making. With artificial intelligence (AI) becoming increasingly relevant in neuro-oncology for diagnosis, segmentation, and outcome prediction, the lack of a large, standardized national dataset poses a major barrier. The Medical Imaging Datasets for India (MIDAS) initiative, a collaborative national effort involving ICMR, IISc, and ARTPARK, aims to create high-quality, annotated medical imaging repositories that can support clinical research and AI model development. As a part of this initiative, we developed a multicenter national repository of dural-based lesions.

METHODS

This ambispective study included patients with radiologically suspected and histopathologically confirmed dural-based lesions from seven neurosurgical centers across India (January 2022-July 2025). Standardized de-identified demographic, clinical, imaging, and pathological data were collected. Imaging was archived in DICOM format and annotated using ITK-SNAP, while histopathology followed WHO-2021 CNS tumor guidelines. Statistical analysis was performed using descriptive and comparative measures.

RESULTS

Among 586 patients, women constituted two-thirds of the cohort, with a mean age of 47.2 years. Meningiomas accounted for 98.3 % of cases and were predominantly WHO Grade I, most commonly of transitional and meningothelial subtypes. Convexity, parasagittal, and falcine locations were most frequently involved. A small but important proportion of lesions were non-meningiomatous, including schwannomas, solitary fibrous tumors, granulomatous, and metastatic lesions. Simpson Grade II resection was the most common surgical outcome, and a subset of patients underwent postoperative adjuvant radiosurgery.

CONCLUSION

This MIDAS-linked national repository represents the largest structured dataset of dural-based lesions from India, integrating standardized clinical, imaging, and pathological information across multiple centers. In addition to defining national disease patterns, the availability of curated imaging and volumetric segmentations provides a strong translational platform for future artificial intelligence-based research, including automated segmentation, diagnostic classification, and outcome prediction.