Abstract:The analysis of animal cross section images, such as cross sections of laboratory mice, is critical in assessing the effect of experimental drugs such as the biodistribution of candidate compounds in preclinical drug development stage. Tissue distribution of radiolabeled candidate therapeutic compounds can be quantified using techniques like Quantitative Whole-Body Autoradiography (QWBA).QWBA relies, among other aspects, on the accurate segmentation or identification of key organs of interest in the animal cross section image such as the brain, spine, heart, liver and others. We present a deep learning based organ segmentation solution to this problem, using which we can achieve automated organ segmentation with high precision (dice coefficient in the 0.83-0.95 range depending on organ) for the key organs of interest.

Automated Mouse Organ Segmentation: A Deep Learning Based Solution