Introduction:
Urinalysis (UA) and urine microscopy are essential diagnostic tools commonly performed in both inpatient and outpatient settings. A comprehensive evaluation of UA findings, coupled with appropriate clinical correlation, can provide valuable insights into underlying nephrology or urological conditions. Similar to the role of AI in ECG interpretation, an initial AI assessment of UA and urine microscopy findings could be a powerful aid in guiding clinicians towards accurate diagnoses and appropriate management strategies. Moreover, this technology may highlight critical UA or urine microscopy findings that clinicians might overlook during busy clinical sessions, given the multifaceted nature of these test results. We evaluated ChatGPT for its potential to interpret UA and urine microscopy results.
Methods:
Objective: Evaluate the accuracy of GPT-4o in interpreting urinalysis and urine microscopy results Methods: GPT-4o response was evaluated to 15 different UA and urine microscopy (results provided, not image of urine microscopy slides) scenarios. A variety of scenarios were provided including Pyuria with and without hematuria, bacteruria & WBC cast, hematuria with and without dysmorphic RBC, RBC cast, proteinuria and atypical cells, isolated proteinuria 1+, 2+, 3+ with and without associated muddy brown cast or free fatty oval bodies, Urine microscopy with multiple squamous epithelial cells, and Bland urinalysis with normal microscopy. The following options were provided for all cases A. Normal; B. UTI; C. Glomerular disease, nephritic; D. Glomerular disease without hematuria; E. ATN; F. AIN; G. Malignancy; H. Contaminated urine sample; I. Malignancy or glomerular disease, nephritic possible; J. AIN or UTI possible; K. ATN or glomerular disease possible; L. Liver Disease; I. UTI or glomerular disease possible. Analysis: Data was analyzed for the 1. Number and % of correct responses, 2. Rationale was assessed on incorrect responses.
Results:
ChatGPT 4o answered 86.6% questions on interpretation of UA and urine microscopy finding correctly when tested on 2 separate occasions 24 hour apart. The 2 questions that it answered incorrectly involved ruling out the possibility of Glomerulonephritis in the presence of hematuria but absence of dysmorphic RBC and RBC cast, it is important to note this as dysmorphic RBC or RBC cast may not be seen in all cases of Glomerulonephritis.
Conclusions:
Evaluation of urinalysis and urine microscopy findings by AI platforms could be a valuable tool for clinicians in their fast-paced practice. ChatGPT-4o correctly answered 86% of questions when presented with options on UA and urine microscopy interpretation. However, it occasionally ruled out important diagnoses based on the presence or absence of findings that, while having high specificity, do not necessarily have high sensitivity for specific conditions. As AI platforms continue to evolve, ChatGPT may not yet be fully reliable for urinalysis and urine microscopy interpretation, but it holds potential as a useful tool in the future.
I have no potential conflict of interest to disclose.
I used generative AI and AI-assisted technologies in the writing process.
During the preparation of this work the author(s) used Chat GPT 4o in order to improve readability. AI was not used in drawing inference and conclusion from the study results. After using this tool/service, the author(s) reviewed and edited the content as needed and take full responsibility for the content of the publication.