Moderate evidence
More
A 181-person study found AI-read urine sorted two bladder conditions at up to 92% accuracy
Study title: Multiclass Diagnostic Improvement of Interstitial Cystitis (IC) and Overactive Bladder (OAB) via Machine Learning Pipeline on Surface-Enhanced Raman Spectroscopy (SERS)
- Evidence type
- Controlled or validated comparison
- Sample
- n = 181
- Sources
- 1 sourcePeer-reviewed
- Published
- 1 min read
- Truvace record
- TRV-2026-1256cite or check it
Study title: Multiclass Diagnostic Improvement of Interstitial Cystitis (IC) and Overactive Bladder (OAB) via Machine Learning Pipeline on Surface-Enhanced Raman Spectroscopy (SERS)
On October 2, 2026, a peer-reviewed study in ACS Sensors reported a urine-based diagnostic pipeline for interstitial cystitis and overactive bladder. Samples from 117 healthy controls, 19 IC patients, and 45 OAB patients were analyzed with Au-ZnO nanorod SERS chips, and spectra were classified using PCA-PLS-DA, PCA-LDA, XGBoost, and LightGBM, reaching up to 92% accuracy for three-way discrimination.
Read the study abstract
Interstitial cystitis (IC) and overactive bladder (OAB) are chronic pelvic conditions with shared symptoms of urinary urgency, increased frequency, and nocturia with pain in the bladder. IC is characterized by bladder pain and inflammation without a clear etiology, whereas OAB involves detrusor overactivity in the absence of infection. Despite this symptomatic overlap, IC and OAB differ in underlying pathophysiology and require distinct treatment and medical care. A specific diagnostic strategy is currently lack…

The gain
In one small study of 181 people (117 healthy, 19 with interstitial cystitis, 45 with overactive bladder), machine learning on urine spectra separated the three groups with up to 92% accuracy.
In brief
The result matters because IC and OAB share urgency, frequency, and bladder pain but require distinct care, and current diagnosis often relies on exclusion. A noninvasive, interpretable spectral classifier could enable disease-specific decisions, though the reported accuracy is from a small, single-cohort study and remains a promising platform rather than a validated clinical test.
Main points
- Study collected urine from 117 healthy controls, 19 IC patients, and 45 OAB patients for SERS acquisition.
- Raman spectra were acquired using Au-ZnO nanorod surface-enhanced Raman spectroscopy chips.
- Linear models PCA-PLS-DA and PCA-LDA and nonlinear models XGBoost and LightGBM were tested.
- Best models reached up to 92% accuracy for three-way classification with biologically interpretable features.
The full story
Researchers collected urine from 117 healthy individuals, 19 IC patients, and 45 OAB patients and measured molecular fingerprints using Au-ZnO nanorod SERS chips. The work addresses the lack of a specific diagnostic strategy for two conditions that share urgency, frequency, and nocturia but differ in pathophysiology and treatment.
Four classifiers were evaluated: PCA-PLS-DA and PCA-LDA as linear models and XGBoost and LightGBM as tree-based nonlinear models. The pipeline aimed to obtain biologically interpretable Raman spectra per group while maximizing classification performance.
The models discriminated among healthy control, IC, and OAB with up to 92% accuracy and yielded interpretable and pathologic relative Raman features. Authors frame urine-based SERS combined with machine learning as a promising platform to support disease-specific clinical decisions rather than exclusion-based diagnosis.
Sources
1 on file- ACS Sensorsdoi:10.1021/acssensors.6c02725
Talk it over
The study had 117 healthy people but only 19 with interstitial cystitis and 45 with overactive bladder. How might that uneven mix change what the 92% accuracy means?
A 181-person study found AI-read urine sorted two bladder conditions at up to 92% accuracyhttps://truvace.com/item/multiclass-diagnostic-improvement-of-interstitial-cystitis-ic-and-overactive-bla
The study had 117 healthy people but only 19 with interstitial cystitis and 45 with overactive bladder. How might that uneven mix change what the 92% accuracy means?
Answer in the threadThe researchers say current diagnosis often relies on exclusion. What would a urine test that separates the three groups change for someone with urgency, frequency and bladder pain?
Answer in the threadThe accuracy comes from a single cohort at one site. What finding from a separate group of patients would make you trust the urine classifier as a clinical test?
Answer in the thread
If you have more time
Look up how a classifier such as XGBoost or LightGBM is checked for overfitting in small samples. What would you ask the authors about it?
Answer in the threadIf a clinic you know offered this urine test, who should decide how its result is used alongside exclusion-based diagnosis, and what would they need to know first?
Answer in the thread
ace
What readers say