Making Microscopy More Accessible with Smartphone-Guided Slide Scanning
Most people haven’t realized smartphones can be a part of clinical workflow. For Sixuan Wu, that assumption is exactly the problem.
At TSRB in Tech Square, Wu, a PhD student in computer science and research assistant in Dr. Alexander Adams’s Uncommon Senses Lab, is working on µMobileScan, a smartphone-based system that transforms a standard microscope setup into a more accessible tool for analyzing blood. The goal is not to replace existing methods, but to improve how they work in practice.
“Without getting high-quality signals, AI is just lost,” Wu said.
That perspective drives the project. µMobileScan focuses less on automation for its own sake and more on something foundational: capturing better data from tools that already exist.
A Common Test, Still Done by Hand
Red blood cell counts are one of the most routine and important measurements in healthcare. They help clinicians identify conditions like anemia or detect abnormalities in the body. Yet in many settings, the process still relies on manual counting.
Technicians load a blood sample onto a hemocytometer (a specialized thick glass microscope slide used for counting blood cells), place it under a microscope, and count cells by eye across a grid. It works, and it remains widely used because the tools are affordable and accessible. But it is slow, labor-intensive, and difficult to verify later. There is often no visual record beyond the final number.
High-end alternatives exist. Whole slide imaging systems can digitize samples and allow for storage, sharing, and reanalysis. But those systems are expensive and out of reach for many clinics.
µMobileScan sits between those two worlds.
“We wanted to use the existing approach and combine it with computation to make it more recordable, retrievable, accessible, accurate,” Wu said.
A Simple Setup with Real Impact
The system itself is straightforward. A smartphone is mounted in front of a microscope eyepiece. As a user moves the slide, the app provides real-time guidance on how to scan properly. It tells them when to move, when to stop, and how to maintain the right overlap between images.
Behind the scenes, the app captures a series of images and stitches them into a single, larger view of the sample. This creates a digital version of the slide that would otherwise only be visible in fragments through the microscope’s narrow field of view.
From there, the system identifies the relevant region and estimates red blood cell counts automatically.
That combination matters. Microscopes offer precision, but only in small windows. Smartphones offer computation, storage, and connectivity. µMobileScan bridges the two.
Designed for Real-World Conditions
The research does not stop at the concept. Wu and his collaborators tested how well the system performs across different users and conditions. In usability studies, participants improved quickly. After about five trials, scanning time stabilized and success rates increased. Accuracy also held up. Across different devices, users, and samples, µMobileScan achieved a mean absolute percentage error of about 1.40 percent in cell counting. That level of performance suggests the system is not just a controlled experiment. It can function under varied, real-world conditions, which is critical for broader adoption.
Rethinking AI’s Role in Healthcare
Wu’s path to the project helps explain its emphasis on practicality. With a background in computer vision and machine learning, he arrived at Georgia Tech already familiar with AI capabilities. He also saw its limits.
“I already have lots of experience with AI and machine learning, but AI is often limited in the healthcare domain,” Wu said. “This is mainly due to the lack of data and the distinction of data distribution, especially when new hardware is introduced.”
“Healthcare problems often depend on physical signals: biological samples, sensors, imaging systems. Without reliable ways to capture that data, even the best algorithms struggle.”
That is where µMobileScan fits. It does not try to replace clinical expertise or automate everything. It focuses on improving the quality and usability of the input itself.
Access, Not Reinvention
The broader promise of µMobileScan is access.
Automatic cell counting devices like flow cytometers are accurate but expensive. Manual methods are accessible but inefficient. µMobileScan offers a third path: keep the low-cost tools, then add a digital layer that improves how they are used.
The implications extend beyond a single test. A system that can capture, store, and share microscope data opens the door to remote analysis, better recordkeeping, and more consistent workflows across clinics.
Wu sees potential applications in digital pathology and education as well. Future versions could help pathologists scan and review samples or allow students to work with realistic, digitized slides.
µMobileScan does not introduce entirely new hardware or require specialized infrastructure. It builds on tools that are already in place and asks how they can work better together. That approach reflects a broader shift in healthcare technology. Progress is not always about replacing systems. Sometimes it comes from connecting them.