The AutoScope: An Automated, Point-of-Care Urinalysis System

Project: MIT EECS Master’s Thesis Research
Advisor: Prof. Charlie Sodini, LeBel Professor of Electrical Engineering, MIT
Sponsorship: Medical Electronic Device Realization Center (MEDRC) - Analog Devices, Inc.

I spent 2 years working on my Master's in Computer Science at MIT. I developed my own low-cost microscope (the Autoscope) and used neural networks to automatically classify particles in urine. My work enables doctors to do low-cost urinalysis at the point-of-care instead of sending it off to a laboratory and waiting a few days for the results.

My low-cost microscope does not have any magnification and so it shouldn't be possible to detect red blood cells. But the cool part is that... it does.

This work highlights the power of neural networks to take advantage of information that we, as humans, cannot.

Blog Post that Describes Research (with GIFs): Available on Medium
Video of Final Project Presentation (26min): Available on YouTube
Slides of Final Project Presentation: Available on Slideshare
Master's Thesis: Available here


Over 200 million urine tests are ordered each year in the US alone. Due to the cost and complexity of microscopic urinalysis tests, the majority are conducted at a central medical lab instead of the point-of-care. The AutoScope is an automated, low-cost microscopic urinalysis system that can accurately quantify red blood cells (RBCs) and white blood cells (WBCs) at the point-of-care. Even without any magnification, we achieved sensitivity, specificity, and R-squared values that are comparable (and mostly better) than the same metrics for the iQ-200, a $100,000-$150,000 state-of-the-art semi-automated urinalysis system. Specifically, the AutoScope’s particle counts and the reference particle counts (cross-validated through medical laboratory results) were linearly correlated to each other (r2= 0.980) for RBCs and WBCs. Furthermore, the AutoScope has an estimated sensitivity of 88% (RBCs) and 91% (WBCs) and an estimated specificity of 89% (RBCs) and 97% (WBCs).