Publications and Presentations
Publications
Monte Carlo Efficacy of IRT Software Estimation
Title: "A Monte Carlo Comparison of the Efficacy of Mplus, flexMIRT, PROC IRT, ltm, and mirt in IRT Models Estimation."
Status: Upcoming
Role: Co-Author under supervision of Yi Zheng and Mark Reiser
Description: Engaging in various research projects with Yi Zheng and Mark Reiser focusing on managing preprocessing, conducting simulations, developing software, and documenting performance evaluations of R's mirt package under the 2PL IRT model. Created comprehensive visualizations of RMSE and bias metrics for all simulated subjects, enhancing the interpretability and impact of our findings.
Presentations
NASA Investigator's Workshop (IWS) - Poster
Title: "Using Natural Language Processing AI Tools to Analyze Mars Tasks"
Date: February, 2025
Authors: Kurth, A. M., Rehm, H., M. Matar.
Affiliations: NASA Glenn Research Center, CHP-PRA, PRisM.
Abstract: In order to plan and prepare for manned missions to Mars, it is essential to understand the tasks that astronauts will need to perform and incorporate this knowledge into our understanding of their performance capabilities. This research project utilizes Natural Language Processing (NLP) and Language Modeling (LM) techniques to assess the Mars Tasks (MT) and categorize them into various Human System Task Categories (HSTC). The goal is to accurately classify each MT with the correct HSTC, but given the severe class imbalance in the dataset, we will employ various techniques to improve the classification performance. The results of this research will be used to inform future mission planning and crew training.
NASA GRC LTX Branch Meeting - Presentation
Title: "Using Natural Language Processing AI Tools to Analyze Mars Tasks"
Date: August, 2025
Authors: Kurth, A. M., Rehm, H., M. Matar.
Affiliations: NASA Glenn Research Center, CHP-PRA, PRisM.
Abstract: Brief, non-technical introduction to concepts involved with HPM-NLP project for the CHP-PRA team.
In order to plan and prepare for manned missions to Mars, it is essential to understand the tasks that astronauts will need to perform and incorporate this knowledge into our understanding of their performance capabilities. This research project utilizes Natural Language Processing (NLP) and Language Modeling (LM) techniques to assess the Mars Tasks (MT) and categorize them into various Human System Task Categories (HSTC). The goal is to accurately classify each MT with the correct HSTC, but given the severe class imbalance in the dataset, we will employ various techniques to improve the classification performance. The results of this research will be used to inform future mission planning and crew training.
Experimental Peak Intensity Analysis for CXLS - Poster
Title: "Peak Intensity Analysis for Serial Femtosecond Crystallography Experiments at the Compact X-ray Light Source."
Authors: Kurth, A. M., Botha, S.
Affiliations: CXFEL Labs Biodesign Institute, Arizona State University; Department of Physics, Arizona State University
Abstract: This research introduces innovative Bragg peak integration methods to advance peak intensity analysis in X-ray crystallography, crucial for accurate structure determination. The Compact X-ray light source (CXLS) will produce femtosecond duration X-ray pulses, allowing the collection of “diffraction before destruction” data. However, the flux will be limited compared to traditional full-scale X-ray free electron laser facilities, warranting the development of unique data analysis tools to fully exploit the unique capabilities of these sources for macromolecular crystallography. By improving the differentiation between signal and noise, coupled with a novel data collection scheme, our approach will enable experiments at the compact X-ray light source at Arizona State University.
Repository: waterbackground_subtraction
Enhancing X-ray Peakfinding Through Deep Learning - Poster
Title: "Enhancing X-ray Peakfinding Through Deep Learning at the Compact X-ray Light Source (CXLS), Arizona State University"
Authors: Kurth, A. M., Everett, E., Botha, S.
Affiliations: The Biodesign Beus CXFEL Laboratory, Arizona State University; Department of Physics, Arizona State University
Abstract: This paper explores the application of deep learning techniques at Arizona State University's (ASU) Compact X-ray Light Source (CXLS) to analyze experimental data from various modalities, primarily focusing on X-ray crystallography using the Dectris Eiger 4M detector. Traditional methods of predicting photon energy and sample-detector distance are challenged by dynamic scattering, intrinsic noise, and the CXLS low flux X-ray beam, prompting the need for more advanced solutions. Utilizing the CrystFEL software, we simulate diffraction images for protein 1IC6.pdb across a matrix of nine variable combinations involving photon energies and camera length. Our approach employs convolutional neural networks (CNNs), testing various architectures for binary classification of peak detection and prediction of experimental parameters. The scope of this research wishes to further expand this with modifications in the architecture to accommodate for spectroscopy data, although this is beyond the extent of this manuscript. By integrating different experimental conditions, we anticipate broader applications and improved experimental outcomes.
Repository: cxls_hitfinder