@elab @rami 📚 Understanding the status quo: Tissue clearing and fluorescent imaging techniques have revolutionized protein expression analysis in whole specimens. By immunostaining for immediate early genes like c-Fos, we get a comprehensive view of neuronal activity.
3/n

@elab @rami 🎯 But there are challenges: Current methods often lack sensitivity and specificity, and while deep learning offers solutions, it requires a lot of training data. Moreover, many of the existing deep learning methods aren't user-friendly for biologists.
4/n

@elab @rami 💡 Introducing DELiVR - our answer to these challenges: a deep learning solution for brain-wide cell analysis, it uses a more accurate and faster VR-based annotation and comes with a user-friendly FIJI plugin.
5/n

@elab @rami 🕹️ The VR advantage: The bottleneck for training a powerful deep learning solution is often the amount of annotated data. Traditional 2D annotation methods are time-consuming, so we turned to VR for a faster and more reliable approach.
6/n
http://discotechnologies.org/DELiVR/Supplementary_Video_2_Segmentation_Syglass_v01.mp4

@elab @rami 🏁 The results? Our VR approach improved annotation accuracy and sped up the process by 7X on average! We fully annotated our training data in VR, revolutionizing the ground truth data generation process.
7/n

@elab @rami 🎉 Comparing DELiVR to existing solutions: Our method outperforms filter and threshold-based segmentation techniques. With 79.18% instance Dice (+38.66% increase) and 84.70% instance sensitivity (+57.79% increase), it clearly stands out from the crowd.
8/n

@elab https://mstdn.science/@rami 🚀 Next level with Allen Brain Atlas: DELiVR produces whole-brain segmentation output in the original image and atlas space. Cells are assigned values matching the Area ID of the Allen Brain Atlas, enabling brain region-based color-coding. Dive into syGlass video for a closer look!
9/n
http://discotechnologies.org/DELiVR/Supplementary_Video_4_DELiVR_Results_overlay_v01.mp4