Graduate Students Seminar

Session Chair:   Gargi Chaudhuri
Discussant:        Dr. Mohammadisiahroudi

Speaker 1:  Maliha Noushin

Title Linking Within-Host Viral Dynamics to Population-Level HIV Spread: A Multi-scale Agent-Based Modeling Framework

Abstract
This work develops a multi-scale modeling framework that connects an ODE-based within-host model to an agent-based model (ABM) to illustrate disease dynamics at the population level. The study aims to understand how a person’s viral load dynamics and treatment responses affect disease propagation in a population-level model. Viral load, CD8+ cells, and both infected and uninfected CD4+ T-cells constitute the within-host model, with each agent having its own within-host dynamics. Our agent-based model characterizes individual autonomous “agents” and then programs them with fundamental behavioral principles to observe how their interactions result in complex, population-level patterns. Within-host dynamics differ among agents in the ABM. Based on their average weekly sexual activity, the agents are divided into two classes: low and high activity. Agents move around, select partners, and interact with one another through sexual contacts. The probability of transmission is determined by the viral load of the infected individuals, and the initially infected agents transmit the virus to susceptible agents through sexual contacts. Simulation results demonstrate how population-level dynamics for the susceptible and infected people change during the course of infection in the presence and absence of treatment

Speaker 2: Vishal Subedi

Title Sequential Change detection in Mammogram 

Abstract
The problem addressed here is whether there exists any way by which the onset of growth of any lesions could be detected from mammogram images as early as possible while avoiding unnecessary false alarms. Here, the problem is posed in terms of change point detection and the classical cumulative sum control chart approach is extended to sequences of images. There are three strategies used in this study to summarize each image including mean intensity level, spatial variability of intensity levels, and a summary statistic based on topological data analysis. For setting up an alarm threshold value, there are no theoretical assumptions involved in it. Rather, it is extracted from the actual data. By simulating mammograms having growing tumors of different velocities, the objective here is to find out which summary statistics raise an alarm before visually detecting the change.