Low Level Feature Extraction for Cilia Segmentation

Cilia are organelles found on the surface of some cells in human body that sweep rhythmically to transport substances. Dysfunction ciliary motion is often indicative diseases known as ciliopathies, which disrupt functionality macroscopic structures within lungs, kidneys and other organs li2018composite. Phenotyping an essential step towards understanding ciliopathies; however, this generally expert-intensive process quinn2015automated. A means automatically parsing recordings cilia determine useful information would greatly reduce amount expert intervention required. This not only improve overall throughput, but also mitigate error, accessibility cilia-based insights. Such automation difficult achieve due noisy, partially occluded potentially out-of-phase imagery used represent cilia, well fact occupy a minority any given image. Segmentation mitigates these issues, thus critical enabling powerful pipeline. However, notoriously properly segment most imagery, imposing bottleneck Experimentation evaluation alternative methods for feature extraction hence provide building blocks more potent segmentation model. Current experiments show up 10\% improvement over base models using novel combination extractors.