Dual-Color Tracer Segmentation with Mean-Teacher Learning and Bleed-Through Correction

Peerawat Pannattee, Akiya Watakabe, Tomomi Shimogori, Henrik Skibbe

Keywords: Dual-color tracer imaging, Segmentation, Bleed-through correction, Signal unmixing

Registration Requirement: Yes

Abstract: Dual-color tracer imaging enables simultaneous visualization of multiple brain pathways, reducing experimental time. However, segmentation for quantitative signal estimation remains challenging due to the lack of ground-truth annotations, domain shift, and signal bleed-through. We propose a preliminary pipeline that combines a U-Net segmentation model trained with a mean-teacher framework and domain-specific augmentations, followed by a clustering-based method for bleed-through correction. Experimental results show improved segmentation performance and reasonable separation of tracer signals under these challenging conditions. These findings demonstrate the promise of the proposed approach for analyzing dual-color tracer data.

Visa & Travel: No

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Submission Number: 78