Tiny-objective segmentation for spot signs on multi-phase CT angiography via contrastive learning with dynamic-updated positive-negative memory banks

Jianhai Zhang; Mackenzie Horn; Koji Tanaka; Fouzi Bala; Nishita Singh; Faysal Benali; Aravind Ganesh; Andrew M Demchuk; Bijoy K Menon; Wu Qiu

doi:10.1016/j.compbiomed.2025.110613

Tiny-objective segmentation for spot signs on multi-phase CT angiography via contrastive learning with dynamic-updated positive-negative memory banks

Comput Biol Med. 2025 Jul 11;196(Pt A):110613. doi: 10.1016/j.compbiomed.2025.110613. Online ahead of print.

Authors

Jianhai Zhang¹, Mackenzie Horn², Koji Tanaka³, Fouzi Bala⁴, Nishita Singh⁵, Faysal Benali⁶, Aravind Ganesh⁷, Andrew M Demchuk⁸, Bijoy K Menon⁹, Wu Qiu¹⁰

Affiliations

¹ School of Life Science and Technology Huazhong University of Science and Technology, China; Department of Clinical Neurosciences, Cumming School of Medicine, University of Calgary, Canada. Electronic address: jianhaizhang@ucalgary.ca.
² Department of Clinical Neurosciences, Cumming School of Medicine, University of Calgary, Canada. Electronic address: mackenzie.horn1@ucalgary.ca.
³ Department of Clinical Neurosciences, Cumming School of Medicine, University of Calgary, Canada. Electronic address: koji.tanaka@ucalgary.ca.
⁴ Diagnostic and Interventional Neuroradiology Department, University Hospital of Tours, France. Electronic address: balafouzi@gmail.com.
⁵ Department of Internal Medicine, Neurology division, Rady Faculty of Health Sciences, University of Manitoba, Canada. Electronic address: nishitaneurology@gmail.com.
⁶ Department of Radiology, Nuclear Medicine, Maastricht University Medical Center (MUMC+), Netherlands. Electronic address: faysal.benali1@ucalgary.ca.
⁷ Department of Clinical Neurosciences, Cumming School of Medicine, University of Calgary, Canada. Electronic address: aganesh@ucalgary.ca.
⁸ Department of Clinical Neurosciences, Cumming School of Medicine, University of Calgary, Canada. Electronic address: ademchuk@ucalgary.ca.
⁹ Department of Clinical Neurosciences, Cumming School of Medicine, University of Calgary, Canada. Electronic address: docbijoymenon@gmail.com.
¹⁰ School of Life Science and Technology Huazhong University of Science and Technology, China; Advanced Biomedical Imaging Facility, Hubei, China; Faculty of Applied Sciences, Macao Polytechnic University, Macao Special Administrative Region of China. Electronic address: qiu.wu.ch@gmail.com.

PMID: 40651197
DOI: 10.1016/j.compbiomed.2025.110613

Abstract

Background and objective: Presence of spot sign on CT Angiography (CTA) is associated with hematoma growth in patients with intracerebral hemorrhage. Measuring spot sign volume over time may aid to predict hematoma expansion. Due to the difficulties that imaging characteristics of spot sign are similar with vein and calcification and spot signs are tiny appeared in CTA images to detect, our aim is to develop an automated method to pick up spot signs accurately.

Methods: We proposed a novel collaborative architecture of network based on a student-teacher model by efficiently exploiting additional negative samples with contrastive learning. In particular, a set of dynamic-updated memory banks is proposed to learn more distinctive features from the extremely imbalanced positive and negative samples. Alongside, a two-steam network with an additional contextual-decoder is designed for learning more contextual information at different scales in a collaborative way. Besides, to better inhibit the false positive detection rate, a region restriction loss function is further designed to confine the spot sign segmentation within the hemorrhage.

Results: Quantitative evaluations using dice, volume correlation, sensitivity, specificity, area under the curve show that the proposed method is able to segment and detect spot signs accurately. Our proposed contractive learning framework obtained the best segmentation performance regarding a mean Dice of 0.638 ± 0211, a mean VC of 0.871 and a mean VDP of 0.348 ± 0.237 and detection performance regarding sensitivity of 0.956 with CI(0.895,1.000), specificity of 0.833 with CI(0.766,0.900), and AUC of 0.892 with CI(0.888,0.896), outperforming nnuNet, cascade-nnuNet, nnuNet++, SegRegNet, UNETR and SwinUNETR.

Conclusion: This paper proposed a novel segmentation approach that leverages contrastive learning to explore additional negative samples concurrently for the automatic segmentation of spot signs on mCTA images. The experimental results demonstrate the effectiveness of our method and highlight its potential applicability in clinical settings for measuring spot sign volumes.

Keywords: Contrastive learning; Hematoma growth; Spot sign segmentation; U-shape neural network.