The Chilkoti Lab (Duke BME) has built a D4 assay that can be used to run ELISA like tests rapidly at point of care. This technology involves the use to labelled antibodies to detect diagnostically relevant antigens. These labelled antibodies are then measured fluorometrically and the image produced by the detector can then be analyzed by this software to provide the necessary concentration intensity profile for that assay. The limited factor in making the D4 assay truly "point-of-care" in time-sensitive, limited-resource environments is an automated image processing script.

What it does

D4 assay results consist of an image of bright, fluorescent spots against a darker background. Our application: (1) successfully detects spot locations from these images, (2) calculates average fluorescence intensity of each spot, (3) plots the data on loglog axes, and (4) computes a sigmoid fit from which unknown sample values can later be determined and used to quantify biomarker concentration.

How we built it

We used the python OpenCV library to extract spot locations and radii from the images. We used the numpy and matplotlib libraries for data analysis, plotting, and curve-fitting.

Challenges we ran into

Using the OpenCV library was quite fun to use, and a little complex initially, but once we got the hang of it, we realized how powerful it was. Since it's an open-source library, there are so many way to apply it without anything being hidden from us programmatically!

Accomplishments that we're proud of

Writing code in python without too many hitches!

What we learned

We learned that image processing is unbelievably easy with Python with simple to moderately difficult tasks-- The future of image processing is bright because of these amazing image processing resources that are available to anyone.

What's next for Point of Care Microassay Image Processing

We plan to continue developing our application until it can be readily and reliably deployed alongside the D4 assay in clinical field trials (conducted in partnership with global partner organizations).

Built With

Share this project: