Skip to content
AI & VisionResearch

Digital Pathology & Allergy Wheal CV Research

Role
Undergraduate AI/ML Researcher
Dates
2026
Stack
Python · OpenCV · TensorFlow · Scikit-Image

01.Executive Summary

Computer vision research developed at The Ohio State University to automate allergy skin prick test (SPT) wheal and flare quantification, replacing manual millimeter calipers with reproducible pixel-calibrated metrics.

02.Problem & Constraints

Manual millimeter ruler measurements of skin prick test wheals exhibit high inter-observer variability in clinical settings, resulting in inconsistent diagnostic grading for environmental and food allergens.

03.System Architecture

Engineered an image-processing pipeline combining spatial color thresholding, convex hull edge contouring, and elliptical fitting to calculate major/minor axis lengths and total surface area against reference calibration markers.

04.Engineering Decisions & Tradeoffs

Prioritized deterministic geometric contouring over deep convolutional segmentation networks to ensure transparent, reproducible measurements that clinical researchers could audit directly without black-box opacity.

Decision 1: Deterministic Geometric Contouring vs Deep Segmentation

Option chosen

Convex hull & Otsu color thresholding with reference marker scaling.

Alternative rejected

U-Net deep segmentation network.

Cost accepted

Requires clean lighting and clearly visible skin calibration marker.

Measured impact

100% explainable metric extraction with zero training data hallucination.

05.Evaluation & Verified Metrics

Key performance indicators and latency profiles benchmarked under explicit workload and hardware constraints.

Measurement Metric

Calibrated mm²

Area calculation against physical marker

Processing Speed

< 80 ms

Deterministic OpenCV pipeline per photograph

06.Personal Contribution

Authored OpenCV image segmentation scripts, calibrated measurement calculations, and conducted clinical test comparisons.

07.Limitations & Next Steps

Known limitations:

Requires consistent flash lighting to prevent specular skin highlights from distorting contour boundaries.

Next engineering steps:

Validating across diverse patient Fitzpatrick skin phototypes and varying erythema intensities.