Turning a Smartphone into a Kidney Test
In the middle of Tech Square at TSRB, a smartphone sits snug inside what looks like a slightly oversized case. A small cuvette of urine slides into place. The flashlight kicks on. A few seconds later, the screen displays something that a dipstick (or even a trained eye) simply cannot produce: a precise, quantitative number.
That number is the whole point.
A Different Kind of Robotics Lab
Uncommon Sense Lab: Kefan Song (left) and Professor Alex Adams (right) (Image: Uncommon Sense Lab/GT College of Computing)
Kefan Song did not set out to build a urine test. He studied biomedical engineering at Bucknell, earned his master's at Johns Hopkins, and arrived at Georgia Tech drawn to medical robotics, the surgical kind, the kind that operates inside people. But research has a way of redirecting you."I always wanted to do something with healthcare," Song says. “The fundamental reason behind my research direction has always been helping people with their medical issues. Having come from a painful surgical experience during high school, my wish is to relieve the pain and suffering of patients to make their experience better.” At Georgia Tech, he found his way to Dr. Alexander Adams and the Uncommon Senses Lab, a group less interested in robots that cut and more interested in sensors that listen. The lab's focus is on hardware-forward healthcare: devices that gather biological signals in the real world, not just in controlled clinical settings. For Song, that translated into a question both obvious and underexplored: what if the most sophisticated diagnostic tool most people owned was already in their pocket?The result is SpectraPhone.
The Problem with a Color Change
Chronic kidney disease (CKD) is, in many ways, a disease of delayed discovery. It affects roughly 10 to 13 percent of adults worldwide, yet fewer than one in five cases is caught before the disease has already progressed to kidney failure, which is a condition requiring lifelong dialysis or transplantation. Early CKD has no symptoms. It doesn't announce itself. And for many people, especially in rural or low-resource settings, the infrastructure to detect it quietly doesn't exist either.Outside of a clinical lab, the standard fallback is the urine dipstick: a paper strip dipped in a sample, read by color change, and discarded. They're fast and cheap; however, they have a limited shelf life and provide a very coarse measurement (typically -, ±,+, ++, or +++). The issue isn't just cost or shelf life. It's that dipsticks are fundamentally qualitative, relying on users to interpret the depth or darkness of a color as a proxy for concentration. They can tell you something is present. They cannot tell you how much. In testing, the SpectraPhone team found that dipsticks often appear to overestimate concentrations, or flag a "large" amount of blood in urine even at concentrations so low they fall below the clinical threshold for concern, rendering every higher reading meaningless by comparison. You can't track disease progression with a color that doesn't change.
Internal components of SpectraPhone: top view and cross section view of the lightpipe configuration.
Internal components of SpectraPhone: top view and cross section view of the lightpipe configuration.
Light Through Urine
SpectraPhone approaches the problem through spectroscopy — the same physics that lets scientists identify the composition of distant stars by analyzing their light.Urine, it turns out, is an information-rich fluid. "Urine is abundant and has a lot of information," Song says. "And from an at-home perspective, it should be fair to treat urinalysis as non-invasive." For kidney disease specifically, two important biomarkers are hematuria (the presence of red blood cells) and albuminuria (the presence of protein). Both signal kidney damage. Both are difficult to quantify precisely outside of a lab.SpectraPhone's hardware is built around that challenge. The phone's flashlight serves as the light source. A custom-designed case channels that light through a cuvette (small transparent lab vessel) containing the urine sample, then routes it through a diffraction grating — a component that splits light into its component wavelengths, the way a prism creates a rainbow. The phone's camera captures the resulting spectrum. An algorithm does the rest.The entire system runs off the phone. No external power source. No bench equipment. No trained technician.What makes it work is what the algorithm is looking for. Different substances absorb light at different wavelengths. Hemoglobin, the protein in red blood cells, has known absorption peaks in the visible spectrum. Albumin is nearly colorless, but even its subtle optical signature can be detected through careful signal processing. Low albumin levels often signal liver or kidney disease. By applying partial least squares regression to spectral data, SpectraPhone can predict biomarker concentrations with striking precision.
Who It's Actually For
The technical performance matters. But Song is quick to anchor SpectraPhone in a more practical question: who needs this, and where are they?
He describes two primary scenarios. The first is the patient managing chronic kidney disease from a distance; someone who lives far from a nephrology clinic and currently has no way to monitor their condition between appointments. The second is the clinician working in a resource-limited environment: a rural outpost, a mobile unit, a medical mission abroad.
In both contexts, what SpectraPhone offers isn't just accuracy, it's frequency. A patient who can test at home gains a longitudinal view of their own kidney health that a twice-yearly clinic visit simply cannot provide. A clinician in the field gains a screening tool that can guide real-time decisions without a centrifuge in sight.
Further down the road, Song envisions the data feeding directly into hospital systems. A future where a reading taken on a phone flows automatically into the clinical record and is flagged for review if a threshold is crossed. That future carries regulatory weight: FDA approval, cross-device validation, and clinical trials on real patient populations, not laboratory-prepared samples. The team is already thinking in that direction, with plans to partner with Emory Midtown to test the system against actual clinical assays.
What Comes Next
Front side view of the SpectraPhone prototype showing the spectrum of an arbitrary sample.
SpectraPhone is a prototype, and the current validation was conducted with controlled, laboratory-prepared samples. Real urine from real patients introduces a more complicated picture: medications, metabolites, underlying conditions, and contaminants that no lab bench can fully anticipate. Whether SpectraPhone's algorithm holds up against that variability is still an open question.
There are also practical hurdles. The system was tested on iPhone models specifically. Adapting it across the full landscape of Android hardware with varying camera sensors, flash outputs, and image processing pipelines will require standardized calibration. The 3D-printed case that houses the optics was designed for iteration, not mass production.
None of this is a dismissal. It's a roadmap.
The broader vision of smartphone-based biomarker sensing is one that Song and the Uncommon Senses Lab are working toward methodically. The next step is more biomarkers. After that, anonymized clinical samples. After that, the regulatory process.
For now, the device does something worth sitting with. It takes a biological signal that has historically required a laboratory to interpret, and it brings that interpretation to the user.