Integrity-Aware Wound Classification Under Image Degradation
Undergraduate Research Assistant, NH-INBRE
Keene State College
Faculty Mentor: Dr. Wei Lu
May 2026 – Present
This research investigates how image degradation affects the
reliability of machine-learning models used for wound
classification in healthcare IoT environments.
-
Evaluate wound-classification models under blur, Gaussian noise,
JPEG compression, reduced resolution, brightness loss, and other
image-quality conditions.
-
Develop image-preprocessing and controlled-degradation pipelines.
-
Analyze model accuracy, precision, recall, F1-score, confidence,
and failure patterns.
-
Identify image-quality conditions that reduce prediction
reliability.
-
Contribute to experimental design, analysis, visualization, and
manuscript preparation.
Key Result:
The model achieved 86.05% accuracy on original test images.
Brightness loss caused the largest decline, reducing accuracy to
73.26%.
Manuscript:
“Integrity-Aware Wound Classification Under Image Degradation in
Healthcare IoT Systems,” submitted to WIDECOM 2026.
A Task-Based Quantitative Model for Predicting Occupational Exposure
to Artificial Intelligence
Undergraduate Researcher
Keene State College
Faculty Mentor: Dr. Wei Lu
March 2026 – Present
This research develops a quantitative model for estimating how
occupations may be affected by artificial intelligence and
automation.
-
Clean and integrate occupational, task-level, and
technology-related datasets.
-
Select predictive variables associated with job tasks, skills,
technology use, and automation exposure.
-
Develop task-based measures of artificial-intelligence exposure.
-
Compare exposure patterns across occupations and job categories.
-
Evaluate model outputs and identify patterns in automation
vulnerability.
-
Contribute to paper development and conference presentation.
Research Contribution:
The model evaluates individual job tasks rather than assigning one
automation score to an entire occupation.
Conference Presentation:
International Conference on Complex, Intelligent, and Software
Intensive Systems, CISIS 2026.
Smart Syringe Pump Design
Undergraduate Researcher
Keene State College
February 2025
This project involved designing and implementing an Arduino- and
C++-based embedded control system for automated fluid delivery.
-
Designed and programmed an Arduino-based fluid-delivery system.
-
Integrated hardware, software, electronic, and mechanical
components.
-
Evaluated dispensing accuracy, consistency, and system
reliability.
-
Assessed cybersecurity vulnerabilities in the control
architecture.
-
Implemented access-control protections against unauthorized
manipulation.
-
Conducted hardware-software integration testing under multiple
operating conditions.
Research Contribution:
The project examined both functional reliability and cybersecurity
requirements for embedded healthcare systems.
Presentation:
Nebraska Conference for Undergraduate Women in Mathematics.