Intensity-based Segmentation of Tissue Images Using a U-Net with a Pretrained ResNet-34 Encoder: Application to Mueller Microscopy
A new AI model is automating a critical, time-consuming step in medical imaging analysis.
Deep Dive
Researchers have developed an AI model that automates the segmentation of tissue images in Mueller microscopy, a task traditionally done manually. Using a U-Net with a pretrained ResNet-34 encoder, the model requires only the total intensity data from images. Trained on just 70 cervical tissue sections, it achieved 89.71% pixel accuracy and an 80.96% mean Dice coefficient on test data, enabling accurate analysis with minimal training data.
Why It Matters
This drastically reduces analysis time for medical researchers, making advanced microscopy techniques more scalable and accessible.