A new deep learning framework developed by researchers from Contextvision, a medical technology company based in Sweden, can help doctors identify and segment lung cancer.
“The microscopic evaluation of cancer is the cornerstone of clinical diagnostics, and particularly in lung cancer where the evaluation is highly dependent on the experience of a pathologist. The evaluation can vary considerably between individual pathologists,” the researchers mentioned in their paper.
To help eliminate some of the variability, the team developed a deep learning framework capable of automatically detecting the disease.
Using NVIDIA TITAN Xp GPUs with the cuDNN-accelerated TensorFlow deep learning framework, the team trained a semantic segmentation network on images from 712 lung cancer patients operated at Uppsala Hospital.
“The trained segmentation models were evaluated qualitatively by pathologists. The predictions demonstrated striking accuracy at the pixel level,” the researchers said. The network achieved 80 percent accuracy, which is on par with other state-of-the-art methods.
The research was recently published on ArXiv. In future work, the team plans to test the trained models on biopsies and images from more clinical settings.
The Contextvision team worked together with researchers from Uppsala University in Sweden