Analysis & Recommendations

Overview of Analysis & Recommendations

The HCG AI application provides an automated medical risk assessment engine designed for the first trimester of pregnancy. By combining quantitative data from $\beta$-hCG blood tests and qualitative measurements from ultrasound scans, the system generates standardized diagnosis summaries, calculates Z-scores, and assigns a clinical risk level.

Clinical Risk Levels

The system categorizes all analysis results into three distinct risk tiers:

• LOW: Measurements are within expected physiological ranges for the reported gestational age.
• MEDIUM: Minor deviations detected or inconsistent data points that require closer monitoring.
• HIGH: Significant deviations from standard growth curves (e.g., abnormal HCG doubling times or discordant ultrasound measurements) that suggest a potential clinical complication.

Automated Analysis Types

1. HCG Trend Analysis

This module tracks the progression of $\beta$-hCG levels over time. Users can provide "Advanced Details" to increase the accuracy of the analysis.

• Standard Inputs: HCG value (mIU/mL) and test date.
• Advanced Inputs: Last Menstrual Period (LMP), average cycle length, and known ovulation dates.
• Output: The system calculates doubling times and compares results against standardized growth curves to generate a Z-Score.

2. Ultrasound Scan Analysis

This module interprets measurements from early pregnancy scans to determine if fetal development is synchronous with gestational age.

• Key Measurements:
  • GS Size: Gestational Sac diameter (mm).
  • YS Size: Yolk Sac diameter (mm).
  • CRL: Crown-Rump Length (mm).
  • Heart Rate: Fetal heart tones (BPM).
• Multiple Sac Support: The system allows for individual measurement entries for multiple gestational sacs (twins/triplets), storing this in a structured JSON format.

AI Chatbot Support

The application integrates with the OpenAI API to provide an interactive "AI Chat" tab. This feature allows users to:

• Ask clarifying questions about their analysis results.
• Understand medical terminology used in the recommendations.
• Receive general guidance on first-trimester symptoms.

Note: The AI Chatbot is an informational tool and does not replace professional medical consultation. It utilizes the OPENAI_API_KEY configured in your .env file.

API Reference: Analysis Results

If you are extending the application or querying the backend directly, the analysis data is managed via the following schema and endpoints.

Analysis Data Object

The analysis_results table stores the processed interpretation of user entries:

| Field | Type | Description | | :--- | :--- | :--- | | entryType | text | Either 'hcg' or 'ultrasound'. | | diagnosis | text | A plain-language summary of the findings. | | riskLevel | text | LOW, MEDIUM, or HIGH. | | recommendations | jsonb | A list of suggested next steps (e.g., "Repeat HCG in 48 hours"). | | zScores | jsonb | Statistical deviation markers for HCG values. |

Fetching Analysis

To retrieve the analysis for a specific user, the frontend interacts with the authenticated data routes:

// Example: Fetching the analysis for the current user
async function getAnalysis() {
  const response = await fetch('/api/hcg-entries', {
    method: 'GET',
    headers: { 'Content-Type': 'application/json' }
  });
  
  const data = await response.json();
  // Results include associated analysisResults via the Drizzle ORM relations
  console.log(data);
}

Usage Guidelines

1. Toggle Advanced Details: For the most accurate risk assessment, users should enable the "Advanced Details" toggle in the UI and provide their LMP (Last Menstrual Period).
2. Sequential Entries: Analysis is most effective when multiple HCG entries are recorded, allowing the system to calculate the rate of change (doubling time) rather than a single static value.
3. Viewing Recommendations: Recommendations are displayed in the Analysis Tab immediately after data is synchronized with the PostgreSQL database.