Multi-site Benchmarking of Deep Learning Models for Intraparenchymal Hemorrhage Segmentation on NCCT

Kaue Tartarotti Nepomuceno Duarte, Abhijot Singh Sidhu, Murilo Costa de Barros, Taha Aslan, Donghao Zhang, Jianhai Zhang, Devansh Bhatt, Brij Karmur, Mohamed AlShamrani, Wu Qiu, Aravind Ganesh, Bijoy Menon

Keywords: stroke, intraparenchymal, hemorrhage, artificial intelligence, medicine, computed tomography

Abstract: Intraparenchymal hemorrhage (IPH) is a critical and often fatal subtype of hemorrhagic stroke, requiring rapid and accurate diagnosis on non-contrast computed tomography (NCCT) scans for effective treatment. While deep learning (DL) models, more specifically using convolutional neural networks (CNNs), offer potential for automating IPH segmentation, their real-world clinical utility is often limited by poor generalization across diverse hospital sites with varying imaging protocols. This study conducted a multi-site benchmarking of three prominent CNN architectures: baseline U-Net, Attention U-Net, and Feature Pyramid Network (FPN), for IPH segmentation on a heterogeneous dataset from 17 clinical sites. Models were rigorously evaluated using F-measure, Intersection over Union (IoU), and 95\% Hausdorff Distance ($d_{H95}$). Both advanced CNN variants (Attention U-Net and FPN) significantly outperformed the baseline U-Net across all metrics (\textit{e.g.}, FPN F-measure: 0.868 vs. U-Net: $0.819$, $p<0.001$), with no significant difference between them. The Attention U-Net and FPN also demonstrated a substantial 53\% reduction in boundary error (measure by $d_{H95}$). These models exhibited robust generalization across different sites and hemorrhage volumes, with minimal site-specific effects on performance. This study demonstrates that advanced CNN variants like Attention U-Net and FPN can be easily adopted for IPH segmentation in real-world, multi-site clinical settings, providing a validated basis for their independent implementation across diverse hospitals to standardize, and potentially accelerate, stroke diagnosis.

Primary Subject Area: Segmentation

Secondary Subject Area: Detection and Diagnosis

Registration Requirement: Yes

Visa & Travel: Yes

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Submission Number: 40