A bibliometric and thematic review of deep learning approaches for brain MRI analysis
Deep learning (DL) has become a prominent approach to brain magnetic resonance imaging (MRI) analysis. This bibliometric and thematic review examines a revision-stage inventory of 585 records with publication-year labels spanning 2016–2026 and describes architectures, tasks, imaging modalities, datasets and research themes. Record-level bibliographic and abstract information was normalized using explicit category mappings and analyzed with overlapping labels. Each record contributes at most once to each category; all main percentages use 585 as their denominator. Convolutional neural network (CNN) labels occur in 403 records (68.9%), transformer labels in 53 (9.1%) and diffusion labels in 16 (2.7%). Classification, detection and segmentation labels occur in 419 (71.6%), 344 (58.8%) and 218 (37.3%) records, respectively. A dataset or cohort is named in the available extraction for 137 records (23.4%). These frequencies describe the documented inventory rather than comparative effectiveness. Original database-specific search histories were not retained, and categories were not uniformly adjudicated from full texts. The supplementary matrix and analysis materials make the descriptive results inspectable. Incomplete search provenance, heterogeneous publication types and limited study-level validation extraction restrict field-wide and clinical conclusions.
Authors
- Rashmi Dahiya (ORCID: https://orcid.org/0000-0002-7985-0862)
- Deepika Sharma (ORCID: https://orcid.org/0000-0003-1376-538X)
- Binod Bhattarai (ORCID: https://orcid.org/0009-0006-6691-5264)
Institutions
- Noida International University (IN)
Publication Details
- Journal
- Discover Artificial Intelligence
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1007/s44163-026-02728-5
- Primary Topic
- Brain Tumor Detection and Classification
- Type
- article
- Field-Weighted Citation Impact
- 0.00