Calculate posterior probability using Bayes' Theorem. Args: prior: Prior probability sensitivity: Test sensitivity s

True when the field was included in the POST (disabled inputs are omitted).

Calculate the posterior probability of disease given a test result, using Bayes' theorem, with input validation.

Persist a single PatientHistory row. Returns the created entry, or None if the user is anonymous or if the database write fails. Anonymous us

Decide if the prediction is reliable. Parameters ---------- confidence_score : float Top prediction's confide

Log error message.

Escape user-controlled strings before putting them in HTML.

Predict disease probability based on selected symptoms.

Log info message.

Calculate posterior probability using Bayes' Theorem.

Create a User row and return (user_id, email, password). ⚠️ This fixture's body needs to match your actual User model. The repo's `User` doe

Generate single synthetic patient. Args: disease: Disease name (random if None) symptom_intensity: 0

Log warning message.

Validates that necessary environment variables and model files are present. In production mode, raises a ValueError/FileNotFoundError if any

Initialize application error. Args: message: Error message status_code: HTTP status code payload

Initialize request logger for Flask app. Args: app: Flask application instance

Mark the flask_login session as authenticated for `user_id`. Bypasses the /login route so this fixture doesn't depend on whatever URL / paylo

Normalize user-facing labels into model-friendly snake_case keys.

Convert error to dictionary.

Add request context to log entry. Args: extra: Additional context data Returns: Dictionary with con

Normalize and fuzzy match disease name to internal key.

Analyze a single set of vitals, flag clinical anomalies, and compute a vitals health index (0 to 1).

Get all unique symptoms across all diseases

Load and process data from CSV

Encode a uint8 RGB numpy array as a base64 PNG data URI.

Load the login page and parse the CSRF token from the rendered HTML.

Strip non-printable characters and enforce length limit.

Helper to determine if the client expects a JSON response.

Get or create global logger instance. Args: name: Logger name Returns: StructuredLogger instance

Compute the raw posterior value without rounding.

Get JSON formatter for structured logging.

Log error details. Args: error: Error instance status_code: HTTP status code include_traceback:

Load, resize, and preprocess an image using ResNet50's preprocess_input (ImageNet mean subtraction) — identical to what predict_disease_type_

Calculate dynamic survival/recovery probability. Args: disease_posterior: The posterior probability of having the diseas

Calculate posterior probability based on test result.

Compute SHAP-style symptom contribution scores for explainability. Uses a linear approximation: - Baseline = expected valu

Log debug message.

Display results in console

Alpha-blend the heatmap over the original image. overlay = alpha * heatmap + (1 - alpha) * original

Map a probability to a coarse risk band (low / medium / high).

Resize the raw heatmap to target (H, W) and apply the COLORMAP_JET colour palette, returning a uint8 RGB array of shape (H, W, 3).

Neutralize CSV/Formula Injection (CWE-1236). User-controlled free-text fields (e.g. disease, notes) are written into exported CSVs as-is

Run a single forward pass and return the output tensor as a float32 array. img_array must be shape (1, H, W, 3), dtype float32. allocate

Best-effort JSON serialisation. Returns None on failure.

Identify high-importance symptoms missing from present symptoms.

Calculate ML probability score for synthetic patient. Args: disease: Disease name symptoms: List of

Temperature-scaled sigmoid for probability calibration.

Configure Gemini API with the API key from environment variables.

"2026-05-20T11:30:00Z" → "May 20, 2026, 11:30 AM" (user locale).

Generate edge cases for stress testing.

Generate a Grad-CAM explanation for a single prediction. Parameters ---------- model : Loaded Keras model (eye diseas

Generate population of synthetic patients.

Generate rare symptom combinations across diseases.

Generate a Score-CAM explanation for a TFLite model prediction. Parameters ---------- tflite_path : Filesystem path to the

Return active thresholds (exposed via /api/ml/config).

Determine risk level based on probability percentage. Args: probability: Probability percentage (0-100) Returns: Dictio

Log API request with details. Args: endpoint: API endpoint method: HTTP method status_code: Resp