The 2018 Data Science Bowl: spot nuclei, speed cures
The 2018 Data Science Bowl was organized Booz Allen Hamilton and Kaggle. The task of the competition was to create a computer model that can identify a range of nuclei across varied conditions. For this competition, I modified Matterport's implementation of Mask R-CNN deep neural network for object instance segmentation. My solution was awarded prize winnning 5th place among 816 teams worldwide. More details can be found in my github repository.
Automated Segmentation of Structures from MR images
Deep learning methods have been widely used to produce state-of-the-art results for object recognition. One of the main advantages of deep learning is they eliminate the need for the manual feature engineering step, which is required for traditional statistical learning approaches. Although the adoption of deep learning in biological applications has been comparatively slow, they have recently been used in biomedical image segmentation, cancer phenotype to genotype matching to name a few. Motivated by the recent findings in the application of deep learning in object segmentation, in this work, we wish to apply up-to-date deep learning techniques for automated segmentation of structures from MR images. More details can be found in my github repository.
egtplot: A Python Package for 3-Strategy Evolutionary Games
Evolutionary game theory is a very broad modeling framework that effectively describes many aspects
of biological cooperation and competition. Visualization of three-strategy evolutionary games has historically
been difficult within the Python ecosystem. We have created a package to ease visualization efforts that is
capable of displaying both static and animated dynamics with the game space.
For detailed software usage instructions we refer readers to our interactive
jupyter notebook.
More details are available in publication in Journal of Open Source Software where you can
also find installation instructions.
