Anish Gudur

I spent week 1 of CIP in the Urology clinic. I saw a mix of new consults, follow-ups with patients, and in-office procedures. Clinic visits in Urology are comparable to other out-patient clinics, with a mixture of history taking and physical exams.

One process unique to Urology is the collection of urine samples. I observed that urine samples were collected from almost every patient, regardless of past conditions and/or diseases on the differential. Urine analysis and urine culture seem to be used as a screening tool. A major issue with urine samples/cultures is improper collection. The penis needs to be properly sanitized to avoid skin contamination and the first 90% of urine needs to be discarded. Improper urine collection leads to many false positives on urine culture, leading to unreliable results to evaluate potential UTIs. The current method for collecting urine doesn’t seem to be user friendly, and most patients seem to have a knowledge gap for how to properly collect a sample.

The two cystoscopies that I watched were quick and efficient, with just a urologist and a nurse in the room. They were each completed in under 10 min and lead to helpful insights for the Urologist to come to a diagnosis.

One of the main symptoms that patients came into the clinic was urinary incontinence. I noticed that dealing with incontinence has a huge effect on quality of life, and many patients seem to come up with their own ways of dealing with incontinence. However, there are many options of dealing with these symptoms that patients may not be aware of, such as condom catheters/external urinary catheter. It seems that educational materials on the current landscape of lifestyle management devices for urinary incontinence seems to be lacking.

During Week 2 of the Clinical Immersion Program, I primarily observed procedures in the operating room, including bladder Botox injections for incontinence, sacral nerve stimulator implantation, robotic prostatectomy using the Da Vinci SP system, and ureteroscopy with laser lithotripsy. One topic that stood out to me was the clinical uncertainty surrounding kidney stone management. In the outpatient clinic, I observed a urologist explain to patients that there is currently no reliable method to predict when a kidney stone will pass. As a result, patients are often left to choose between preemptively removing the stone when they may not need to yet or waiting for symptoms to develop.

A recent study by Xiao, Bai, and Zhang (2024) explored the use of deep learning models to predict spontaneous stone passage (SSP) using CT imaging. The researchers developed two convolutional neural networks—2D and 3D ResNet29 architectures—that analyzed CT images from over 1,200 patients. Their models significantly outperformed traditional stone-size–based heuristics, achieving up to 90.6% accuracy and 95.1% specificity with the 3D model. This innovation suggests that image-based AI could play a critical role in informing the clinical decision of whether to observe or intervene and when to intervene in asymptomatic stone cases.

The Cleveland Clinic has patented a slightly different, but complimentary approach to preemptively treat kidney stones (U.S. Patent US20150031992A1). They classified kidney stone composition using CT imaging features, such as attenuation values from the stone’s center and periphery. Machine learning algorithms are applied to these features to determine the likely stone type—for example, differentiating uric acid from calcium-based stones. Currently, stone type cannot be determined until they are collected which guides treatment to prevent recurring stones. This patent could allow for long-term prevention of stones and avoid procedures. While this tool focuses on stone type rather than passage prediction, both approaches both approaches aim to enhance early, non-invasive decision-making.

When a patient presents to the ED with flank pain, CT confirms a ureteral stone but tells us little else. Today, urologists must choose between early surgery (ureteroscopy + laser lithotripsy) or watchful waiting without reliable data on (a) whether the stone will pass on its own or (b) whether it is a medically dissolvable uric-acid stone. This clinical uncertainty drives unnecessary procedures, delayed care, and patient anxiety.

Storyboard 1 — Status Quo Pathway

1. Flank-pain ED visit ➜ CT scan identifies a 6 mm stone.

2. Consult: “Surgery or wait?” — no predictive tool.

3. Uncertainty → most patients schedule ureteroscopy weeks out.

4. Pre-op, OR lithotripsy, post-op recovery; stent placed.

5. Stone type learned only after lab analysis; prevention advice given late.

Pain points: extra OR time, stent discomfort, delayed prevention, high cost.

Storyboard 2 — AI-Guided Pathway

1. Same CT scan, but AI instantly predict:

a) Stone-Type Model → identifies type of stones (calcium oxalate, struvite, uric acid, cystine, etc.)

b) Spontaneous-Passage Model → predicts ≥/< 50 % pass chance.

2. Branching care:

• Uric-acid OR ≥50 % pass → medical therapy + follow-up US; many stones dissolve (uric acid) or pass, avoiding surgery
• Non-dissolvable & <50 % pass → early ureteroscopy + laser lithotripsy (shorter symptom window, fewer ED returns).

3. Preventive consult happens sooner for every path.

Benefits: data-driven decisions, fewer unnecessary surgeries, targeted prevention