|
Activity Number:
|
355
- Advanced Bayesian Topics (Part 4)
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #319036
|
|
|
Title:
|
A Bayesian Ising model to infer the hidden lattice structure of spatial point pattern data
|
|
Author(s):
|
Esteban Fernández* and Qiwei Li
|
|
Companies:
|
The University of Texas at Dallas and The University of Texas at Dallas
|
|
Keywords:
|
Ising model;
layer detection;
spatial point pattern;
Markov random field;
Metropolis-Hastings;
oral cancer
|
|
Abstract:
|
Recent developments in deep-learning methods have enabled us to identify and classify individual tumor cells from digital pathology images. This provides new opportunities to study the grouping of cells and its relation to cancer progression. One immediate question is how to detect the layers formed by these groupings. Most of the current methods are manual and prone to errors. To provide an alternative, we introduce a model-based approach in a Bayesian framework. The key idea is to infer the lattice structure of the cells via an Ising model. The results allow us to easily detect and count the cell layers. We use a Metropolis-Hastings algorithm to update the lattice and improve the posterior density. We demonstrate how this model-based analysis can lead to sharper inferences than manual techniques, by means of application to two benchmark datasets and a case study on the pathology images of oral cancer patients. Simulation results show that our modeling approach led to recreating the shapes in the benchmark datasets, while the case study shows that the number of layers formed by the lattice predicts patient prognosis.
|
Authors who are presenting talks have a * after their name.
