An International Multicenter Clinical Study on Application of UroMed AI Doctor Based on Large Language Models

Condition(s) studied

Investigational drug(s) / intervention(s)

UroMed AI Doctor-assisted Urological Diagnosis, Treatment and Preoperative Health Education: This urological intervention uses the independently developed UroMed AI Doctor, a urology-specialized large language model system distinct from generic medical AI tools. Trained \*\*exclusively\*\* on the latest international urology guidelines and high-quality literature, it has a built-in data cleaning system blocking non-standard knowledge sources, eliminating factual deviations and guideline misalignment common in general LLMs. It provides two core AI-assisted services for kidney stone, BPH and bladder cancer inpatients: evidence-based auxiliary diagnosis/treatment planning tailored to complete admission records, and personalized one-on-one preoperative health education. Uniquely equipped with ASEAN multilingual interaction and lightweight edge deployment for cross-border use, all AI outputs strictly adhere to urological clinical norms, ensuring professional accuracy and safety unavailable in non-specialized medical AI interventions.

Physician-led Urological Diagnosis, Treatment & Preoperative Education: This intervention consists of standard, physician-led urological care without any artificial intelligence support. Qualified urologists independently diagnose and create personalized treatment plans for inpatients with kidney stones, BPH, or bladder cancer, following official clinical guidelines and consensus. One-on-one preoperative education, including disease information, treatment procedures, and postoperative care, is provided directly by attending physicians. This arm represents routine clinical practice, serving as a clear, active comparator to the AI-assisted intervention, ensuring a direct, valid comparison in effectiveness and safety between traditional care and AI-supported care.

Study summary

This study evaluates our team's urology-specific AI (UroMed AI Doctor) for its safety, professionalism, knowledge and Q\&A ability, and tests its effectiveness against traditional manual urology care, to confirm if it can be a safe auxiliary tool and improve patients' preoperative experience.

Before the study, we will test the AI with urology questions, compare it to international AI models (DeepSeek, ChatGPT, Google Gemini), and have two senior chief physicians evaluate it.

In the clinical trial, patients at The First Affiliated Hospital of Guangxi Medical University will be randomly split into two groups: AI-assisted care or traditional care by a specialist.

Two senior specialists will evaluate both groups blindly; each group will get preoperative education (AI or physician), with anxiety and satisfaction surveyed.

Subsequently, a multi-center validation will be conducted with 11 domestic and international hospitals.

Eligibility

Primary outcome measure(s)

• Comprehension of Medical Cases — Baseline Day 1
  Evaluates the ability to extract and summarize patient condition information for kidney stones, benign prostatic hyperplasia, or bladder cancer, scored on a 1-5 Likert scale (1=loses almost all correct condition basis and cannot diagnose; 5=provides complete basis for correct condition summary).
• Adherence to Medical Guidelines and Consensus — Baseline Day 1
  Assesses the consistency of diagnostic and treatment suggestions with clinical guidelines, professional consensus and clinical practice, scored on a 1-5 Likert scale (1=completely deviates from guidelines; 5=fully complies with guidelines, consensus and clinical practice).
• Clinical Reasoning — Baseline Day 1
  Measures the logicality, evidence-based nature and comprehensiveness of the clinical reasoning process for urological diagnoses, scored on a 1-5 Likert scale (1=reasoning violates clinical logic with irrelevant conclusions; 5=comprehensive, systematic reasoning adhering to evidence-based medicine principles).
• Relevance of Differential Diagnoses — Baseline Day 1
  Evaluates the value of differential diagnosis in narrowing potential disease causes for definitive diagnosis of urological diseases, scored on a 1-5 Likert scale (1=no diagnostic value; 5=excellent value for accurate and reasonable medical decisions).
• Diagnostic Acceptability — Baseline Day 1
  Assesses the clinical rationality, completeness and accuracy of the definitive diagnosis (including staging/severity) for urological patients, scored on a 1-5 Likert scale (1=absurd diagnosis with serious errors; 5=comprehensive, accurate diagnosis meeting medical standards).
• Presence of Unrealistic Content — Baseline Day 1
  Measures the accuracy and authenticity of diagnosis and treatment plan content, evaluating the absence of fabrication or factual errors, scored on a 1-5 Likert scale (1=completely incorrect/fabricated content; 5=100% accurate content consistent with medical facts).
• Bias and Unfairness — Baseline Day 1
  Evaluates the absence of bias in diagnosis and treatment plans, and the full consideration of individual patient differences and diversity, scored on a 1-5 Likert scale (1=severe bias ignoring individual differences; 5=completely bias-free with full consideration of individual diversity).
• Potential Harm — Baseline Day 1
  Assesses the risk of misleading clinical practice or causing medical incidents from diagnosis and treatment suggestions, scored on a 1-5 Likert scale (1=completely incorrect content with high risk of serious medical incidents; 5=fully reliable content with no misleading or harm risk).

Trial sites (5)

More First Affiliated Hospital of Guangxi Medical University trials in China

Other trials for the same condition