Content Tags

There are no tags.

Multiclass Weighted Loss for Instance Segmentation of Cluttered Cells.

RSS Source
Authors
Fidel A. Guerrero-Pena, Pedro D. Marrero Fernandez, Tsang Ing Ren, Mary Yui, Ellen Rothenberg, Alexandre Cunha

We propose a new multiclass weighted loss function for instance segmentationof cluttered cells. We are primarily motivated by the need of developmentalbiologists to quantify and model the behavior of blood T-cells which might helpus in understanding their regulation mechanisms and ultimately help researchersin their quest for developing an effective immuno-therapy cancer treatment.Segmenting individual touching cells in cluttered regions is challenging as thefeature distribution on shared borders and cell foreground are similar thusdifficulting discriminating pixels into proper classes. We present two novelweight maps applied to the weighted cross entropy loss function which take intoaccount both class imbalance and cell geometry. Binary ground truth trainingdata is augmented so the learning model can handle not only foreground andbackground but also a third touching class. This framework allows trainingusing U-Net. Experiments with our formulations have shown superior results whencompared to other similar schemes, outperforming binary class models withsignificant improvement of boundary adequacy and instance detection. Wevalidate our results on manually annotated microscope images of T-cells.

Stay in the loop.

Subscribe to our newsletter for a weekly update on the latest podcast, news, events, and jobs postings.