New deep learning algorithm achieves human-level accuracy in identifying cell boundaries
In short: Researchers have developed Mesmer, a deep-learning algorithm capable of segmenting individual cells in tissue images with performance comparable to human experts. By training on TissueNet—a massive new dataset containing over one million manually labeled cells—the system can accurately identify precise cell boundaries across a wide variety of tissue types and imaging technologies.
The study introduced TissueNet, a dataset an order of magnitude larger than previous collections, to train the Mesmer algorithm. This approach allows for the automated extraction of complex cellular features, such as protein localization. While the model demonstrates high accuracy and generalization, its performance relies on the large-scale manual annotations provided in the training set.
The researchers also demonstrated the tool's utility by quantifying morphological changes in cells during human gestation.
Why it matters: Large-scale data annotation can enable AI to match human expertise in complex biological image analysis.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Summary by Intellect · source: Research: Greenwald et al. (2022). Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning. Nature biotechnology.
To cite this card, link to https://intellectplus.co.uk/app/health-research-new-deep-learning-algorithm-achieves-human-level-accuracy-in.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
Related
More in Health · Latest · A–Z · All topics
