Patient-Level Intracranial Aneurysm Detection in Volumetric Neuroimaging Using a 2.5D Deep Learning Framework
Journal of Smart Algorithms and Applications (JSAA) · 26.06.2026 · 4 (2) · 61-72
DOI: 10.66279/na055t34 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Smart Algorithms and Applications (JSAA), online erschienen am 26.06.2026. Verfasst von Santhosh Pandi C und 2 weiteren Autor:innen. Die Arbeit nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Intracranial aneurysms (IAs) have a case-fatality rate greater than 50% following rupture, and the majority are clinically silent until hemorrhage occurs. An important research direction in neurovascular imaging informatics is the development of automated screening tools that can identify potentially high-risk cases for expedited expert review. In this work, we propose a binary deep learning framework for patient-level aneurysm detection from volumetric DICOM neuroimaging. The proposed architecture adapts EfficientNetV2-S to a 2.5D input representation of channel-stacked 32 axial slices resized to 384 × 384 pixels. Percentile-based intensity normalization is employed to reduce cross-scanner variability. We measure model generalization using stratified five-fold cross-validation and improve prediction stability using weighted ensemble averaging, test-time augmentation, and temperature-scaled probability calibration. On the RSNA 2025 Intracranial Aneurysm Detection Challenge dataset of 1,147 clinical cases, the framework achieved a weighted AUC-ROC of 0.694 with a cross-validation mean of 0.681±0.005. These results indicate that the proposed method can be considered as a first research baseline for automated aneurysm pre-screening, but the performance obtained is still not enough to be used in the clinical routine without assistance. Before its integration into clinical radiology workflows, it needs external validation on independent multi-site cohorts, threshold-specific sensitivity and specificity analysis, and lesion-level evaluation when annotations are available.
Zitieren
APA
C, S. P., Chakrabarti, P., & Gupta, S. K. (2026). Patient-Level Intracranial Aneurysm Detection in Volumetric Neuroimaging Using a 2.5D Deep Learning Framework. Journal of Smart Algorithms and Applications (JSAA), 4(2), 61-72. https://doi.org/10.66279/na055t34
Vancouver
C SP, Chakrabarti P, Gupta SK. Patient-Level Intracranial Aneurysm Detection in Volumetric Neuroimaging Using a 2.5D Deep Learning Framework. Journal of Smart Algorithms and Applications (JSAA). 2026;4(2):61-72. doi:10.66279/na055t34
BibTeX
@article{C2026,
author = {C, Santhosh Pandi and Chakrabarti, Prasun and Gupta, Shashi Kant},
title = {Patient-Level Intracranial Aneurysm Detection in Volumetric Neuroimaging Using a 2.5D Deep Learning Framework},
journal = {Journal of Smart Algorithms and Applications (JSAA)},
year = {2026},
volume = {4},
number = {2},
pages = {61--72},
doi = {10.66279/na055t34}
}
EUVIMED