2021 Alumni Reunion: Virtual Poster Participants and Presentations

Lamia Alam, PhD Student in Cognitive and Learning Sciences

Email: lalam@mtu.edu

Area of focus: Explainable AI

Topic: Examining the Effect of Explainable AI in Medical Diagnosis Using LIME Algorithm

Project Summary: There is an increasing interest in Artificial Intelligence (AI) for supporting and assisting medical diagnosis. One potential application of AI is as the first point of contact for initial. Even though many AI-driven clinical prediction tools have achieved high accuracy performance, the lack of explainability of these tools continues to spark criticism (Amann et al., 2020). If these systems do not
provide explanation about why the diagnoses are made, the high-performance ones may be ignored or rejected. In this paper, we examine this problem using a simulation experiment. We tested an evolving clinical diagnosis scenario and used LIME (Ribeiro et al., 2016) algorithm to generate feature-based graphical explanations for probable clinical conditions. We found that explanation helps improve satisfaction measures during the critical re-diagnosis period but had little effect later when an alternate diagnosis resolved the case successfully. Furthermore, we found that providing positive and negative examples (contrasts) along with the current case features were no more helpful than providing the current case features only. We also evaluated the understanding of the explanations using knowledge tests and found that explanation did not help improved understanding of the AI-generated diagnosis decisions.

Seth A. Kriz, PhD Student in Chemical Engineering

Email: skriz@mtu.edu

Area of focus: Bioseparations, Vaccine Manufacturing

Topic: Purifying viral vaccines by two-phase aqueous extraction

Project Summary: The World Health Organization (WHO) estimates as many as 650,000 people die annually from the seasonal influenza virus. Viral vaccines are an effective tool to combat this enormous problem, but current manufacturing purification methods suffer from high costs and yields of less than 30%. A switch from batch to fully continuous processing, which is acknowledged by the FDA to increase capacity and product consistency, is necessary to meet demand. Aqueous two-phase systems (ATPS) constructed of inexpensive, environmentally-friendly polymers and salts are an ideal method to replace traditional chromatography steps that rely on costly resins and operate discontinuously. Previously, we achieved over 80% recovery of two model viral products in the polymer phase of ATPS with high host cell protein and DNA removal. However, the purified viral product is too viscous for further polishing by traditional filtration methods. Thus, a polymer removal step is required. Here we developed a second stage of ATPS to back-extract the virus from the polymer-rich primary product into a gentle salt solution ready for polishing. Preliminary results demonstrate that back-extraction completes a fully continuous viral particle extraction process using ATPS.

Sushree S. Dash, PhD Student in Physics

Email:  ssdash@mtu.edu.

Area of focus: Magneto-optic Materials

Topic: : Boosting Optical Nonreciprocity – A Surface Reconstruction Phenomenon in Iron Garnets

Project Summary: : We are trying to understand the enhanced Faraday Rotation effect in Bi-LuIG materials by studying the surface effects.

Tauseef Ibne Mamun, PhD Student in Cognitive and Learning Sciences

Email: tmamun@mtu.edu

Area of focus: Explainable Artificial Intelligence

Topic: Collaborative Explainable AI: A non-algorithmic approach to generating explanations of AI

Project Summary: An important subdomain in research on Human-Artificial Intelligence interaction is Explainable AI (XAI). XAI attempts to improve human understanding and trust in machine intelligence and automation by providing users with visualizations and other information that explain decisions, actions, and plans. XAI approaches have primarily used algorithmic approaches designed to generate explanations automatically, but an alternate route that may augment these systems is to take advantage of the fact that user understanding of AI systems often develops through self-explanation. Users engage in this to piece together different sources of information and develop a clearer understanding, but these self-explanations are often lost if not shared with others. We demonstrate how this ‘Self-Explanation’ can be shared collaboratively via a system we call collaborative XAI (CXAI), akin to a Social Q&A platform such as StackExchange. We will describe the system and evaluate how it supports various kinds of explanations.

Karrar Takleef Alofari, PhD Student in Mechanical Engineering and Engineering Mechanics

Email: ktalofar@mtu.edu

Area of focus: Multi-phase Flow in Porous Media

Topic: The Impact of Relative Humidity on The Porosity and The Structure of PEM Fuel Cell Catalyst Layer

Project Summary: Understanding and modeling of mass transport limitations in the catalyst layers in PEM fuel cells remain a challenge despite decades of commercial development. That challenge has led to the development of a novel ex-situ test to characterize mass transport resistances in these extremely thin porous layers. This test characterizes radial percolation of gas and liquid at varying fluid injection rates and relative humidities. Liquid percolation exhibits a dominant capillarity influence at low injection rates with lower final wetted areas and saturation as compared to high injection rates. Changes in relative humidity have a significant effect on percolation behavior for both gas and liquid. There is a significant jump in resistance when the relative humidity exceeds 65%.

Jacob James Blazejewski, PhD Student in Mathematical Sciences

Email: jblazeje@mtu.edu

Area of focus: Numerical Analysis, Radial Basis Functions

Topic: Using Radial Basis Functions to Solve PDEs on Evolving Curves and Surfaces

Project Summary: What does modeling the growth of a cancer tumor have in common with telling a computer what part of an x-ray contains a brain? They are just two examples of many scientific questions that can use a partial differential equation (PDE) to model an evolving curve or surface. It is therefore imperative to always be developing accurate and computationally efficient algorithms to solve the PDE. Our research group is developing an approach using the Radial Basis Function (RBF) Method to accomplish this goal.

Manas Warke, PhD in Biological Sciences

Email: mwarke@mtu.edu

Topic: Using Phytoremediation to Lower Arsenic Accumulation in Rice Grains

Project Summary: Arsenic (As) is a naturally occurring element found in the earth’s crust. As is a Class I carcinogen. The soil and groundwater in South Asia, especially West Bengal, India, and Bangladesh are naturally contaminated with As. In many areas, contaminated groundwater is the only source for irrigation, drinking, and other household uses. Extensive use of As contaminated groundwater for irrigation has led to an increase in As in soil and crop. Rice is part of the diet for more than three billion people around the world. Due to the anaerobic conditions in rice fields, more As gets translocated compared to other crops which is a major human health concern We are developing a crop rotation system of alternatingly planting rice with an As hyperaccumulator Pteris vittata (PV) under simulated field conditions to reduce the concentration of As in the soil over a period of two years, and also reduce As accumulation in rice grains.