Demo Cases
Diabetes Risk Assessment
Age: 55, BMI: 28, FBG: 6.2, Family History: Yes. Requesting risk prediction.
Health Report Analysis
Transforming unstructured physical examination reports into structured, actionable preventive insights.
Processing Workflow:
- Stage 1: Indicator Extraction: Extracting structured indicators from document.
- Stage 2: Reference Data Loading: Loading normal reference ranges and guidelines.
- Stage 3: Indicator Analysis: Comparing extracted values against reference data.
- Stage 4: Abnormality Interpretation: Multi-agent system identifies and interprets abnormal values.
- Stage 5: Examination Suggestions: Recommending further clinical tests based on findings.
- Stage 6: Department Recommendation: Triaging to the appropriate medical departments.
- Stage 7: Generating Summary: Synthesizing the final structured report and advice.
Please analyze my physical exam report. Blood glucose is 6.5 mmol/L, LDL is 3.8 mmol/L, BP 135/85.
The preliminary analysis is complete. Your fasting glucose (6.5) and LDL (3.8) are elevated. To finalize the summary, do you have a family history of diabetes or cardiovascular disease?
Final Summary & Advice:
Combined with your father's type 2 diabetes history, your metabolic risk is significantly elevated. Based on the workflow analysis, here is your tailored plan:
- Department: Schedule a follow-up with Endocrinology.
- Further Exams: HbA1c test and OGTT recommended.
- Lifestyle: Start a Mediterranean diet and exercise 150 mins/week.
Diabetes Risk Assessment
Providing calibrated risk estimation and feature-level interpretability via integrated quantitative models.
Processing Workflow:
- Data Parsing: System parses patient demographics, lab results, and family history from dialogue.
- Model Invocation: Automatically triggers the specific Diabetes Risk Model from the 20+ available models.
- Risk Calculation: Computes a calibrated probability score.
- Interpretation: Generates a SHAP plot to explain feature importance to the patient.
Based on your symptoms (polydipsia) and family history, there is a potential risk for diabetes. To provide a precise quantitative analysis, please fill out this assessment form.
I've submitted the required information.
The model calculates your Diabetes Risk Score as 72% (High Risk). Here is the feature attribution (SHAP) explaining why:
Suggestions:
- Perform an Oral Glucose Tolerance Test (OGTT) immediately.
- Aim for a 5-10% weight loss.
- Consult a dietitian for a low-GI meal plan.
Medical Imaging Triage
Multi-modal agent capability integrating visual anomaly detection with conversational context.
Processing Workflow:
- Multi-modal Input: User uploads an imaging file (e.g., Chest CT) alongside textual symptoms.
- Vision Analysis: Vision Model analyzes the scan to identify anomalies.
- Context Integration: Synthesizes imaging findings with user's dialogue history.
- Triage & Advice: Outputs follow-up recommendations based on retrieved medical protocols.
I have a mild cough, what does this scan show?
I have analyzed the imaging and found a 5mm subsolid nodule in the right upper lobe, with a probability of 82.81%. Do you experience any discomfort, and if so, how long has it lasted? According to the clinical manifestations, it is most likely a breast cyst or cystic hyperplasia/simple small cyst, all of which are benign features.
Analysis:
Based on the characteristics you described and the appearance of the 5mm subsolid nodule, the current risk is low to moderate, so you do not need to worry excessively. There are no obvious signs of acute infection or large malignant lesions on the imaging. This indicates that, according to the current test results, your breast condition is most likely benign (non-cancerous) with a low risk. However, please note that no test result can completely rule out abnormalities. It is recommended that you consult a doctor for further examination and confirmation.
Suggestions:
- According to the retrieved Fleischner Society Guidelines (RAG), a 5mm subsolid nodule requires follow‑up but no immediate surgical intervention.
- Diet: Reduce intake of high‑fat and high‑sugar foods. Increase consumption of vegetables, fruits and whole grains, and maintain a healthy weight.
- Exercise: Engage in at least 150 minutes of moderate‑intensity exercise per week (e.g., brisk walking, swimming) and avoid prolonged sitting.
- Lifestyle: Quit smoking and limit alcohol consumption. Maintain a regular schedule without staying up late. Relax your body and mind through meditation, yoga, etc., to relieve stress.
- Breast care: Perform regular monthly breast self‑examinations (recommended 7–10 days after menstruation) and watch for new lumps or abnormal changes.
Open-Domain QA with RAG
Conversational agent leveraging Retrieval-Augmented Generation to answer open-ended health queries using validated medical literature.
Processing Workflow:
- Query Analysis: AI comprehends user's open-ended health question and personal context.
- Literature Retrieval: Searches embedded vector database of clinical guidelines and medical papers (RAG).
- Evidence Synthesis: Consolidates multi-source information to reduce hallucination.
- Answer Generation: Delivers personalized advice with accurate literature citations.
Analysis:
We can combine Western symptomatic relief with TCM syndrome differentiation, while ensuring drug safety...
Suggestions:
- Western Medicine: Use antipyretics (Paracetamol...) and antihistamines; antibiotics are ineffective against viruses.
- TCM Approach: Choose formulas based on patterns (Wind-Cold, Wind-Heat, etc.); avoid drug-drug interactions.
- Care & Prevention: Rest, hydration, and flu vaccination; consult a doctor if symptoms worsen or complications arise.
- *Citation: AHA 2024 Guidelines on Primary Prevention; Lewis et al., 2024 (RAG)
