Authors:

Kotekar Annapoorna Prabhu

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Zebrafish (Danio rerio) are small, striped fish that are important in biomedical research, primarily due to their genetic similarity to humans. One of the pigment cell types in zebrafish, melanophores, are closely related to human melanocytes, making them valuable in studying melanoma, a form of skin cancer. Zebrafish stripes take about 3 months to fully develop, making the process of studying their patterns time-consuming. To this end, mathematical models have been developed to simulate stripe formation in zebrafish. However, the simulated patterns produced by these models lack the realism of real fish. The goal of our research is to develop image-processing software that can create realistic zebrafish patterns using mathematical model data. We construct the simulated patterns by sampling different pigment cells from real zebrafish. To make the pattern appear realistic, we select each pigment cell in the simulated stripe by taking the age of the fish and the pigment cell's proximity to the center of the pattern into consideration. Augmenting this approach, we also use a conditional generative adversarial network (cGAN) to generate realistic zebrafish stripes using plots of pigment cell coordinates. In particular, we use 80 pairs of real zebrafish images and their corresponding plots of pigment cell coordinates. Our post-processing software can render a realistic timelapse of zebrafish stripe formation in minutes instead of months, and can be used to review mathematical models of zebrafish both quantitatively and qualitatively. Additionally, we are adapting our methods to produce realistic images for other biological systems, such as fern development.

Source:

Purdue University / 2023

Topics:

No topics listed

Co-authors:

Kotekar Annapoorna Prabhu