- MATLAB 69.1%
- Python 30.9%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
Updated references from 'RatCortex-IOS' to 'MouseCortex-IOS' and revised abstract details to reflect dataset specifics. Adjusted figure descriptions and contact email for dataset inquiries. |
||
| IOS dataset code | ||
| README.md | ||
| tracking sample.mp4 | ||
MouseCortex-IOS
Home page of RatCortex-IOS dataset sharing.
A Mouse Cortex Video Segmentation Dataset for Intrinsic Optical Signal Tracking and Neural Activity Analysis
From MixLab, SIAT, ShenZhen, China.
Abstract
Intrinsic optical signal imaging (IOSI) enables non-invasive monitoring of neural activity in the mouse cortex, yet quantitative analysis remains hindered by low signal intensity, complex spatiotemporal patterns, and the lack of standardized benchmarks. To tackle this challenge, we present MouseCortex-IOS, a novel open accessible video segmentation dataset specifically designed for standardizing IOS analysis in awake rodent models, comprising 5732 images from 14 experimental subjects. Meanwhile, we implement an efficient processing pipeline leveraging foundation models to ensure annotation consistency while minimizing manual intervention. The dataset supports quantitative characterization of neural activation parameters including signal propagation velocity and trajectory, serving as a critical benchmark for developing automated analysis tools. Furthermore, this resource facilitates technique development in neuroimaging studies and accelerates the integration of computational approaches in IOS-based neuroscience research.
Fig. 1 Data processing framework. Overall, raw data undergoes preprocessing and subsequent processing via the SAM2 model to generate our dataset. Three stages of preprocessing convert massive indistinguishable raw grayscale images into interpretable color-coded signal maps while reducing data volume by two orders of magnitude. The preprocessed image sequences are imported into an open-sourced annotation tool, where a preloaded SAM2 model enables manual or automated annotation of prompts for the initial frames. A single-click inference operation then generates annotations for all frames. The annotated data is used for downstream tasks such as signal analysis and tracking.
Fig. 5. Display of some extracted frames from a vagus nerve stimulation signal video segment. The F num in blue box means frame order in the 25 clip, and in up left region of every image are instant velocity, average velocity, signal coverage area that from original point to the end and signal lasting time. The red line is the trajectory of signal center and the green line is direction of signal from start to the end. The numbers in parentheses after the points are pixel coordinates.
Due to original data are in TB level, we will upload part of the proccessed dataset in figshare, https://doi.org/10.6084/m9.figshare.28601813.v4. If you are interested in the dataset, for more details please email: w_zhangstudy@163.com .