endoreg-db 0.8.9.32__py3-none-any.whl

endoreg_db/models/metadata/sensitive_meta_logic.py

@@ -0,0 +1,1161 @@
+import logging
+import os
+import random
+import re  # Neu hinzugefügt für Regex-Pattern
+from datetime import date, datetime, timedelta
+from hashlib import sha256
+from typing import TYPE_CHECKING, Any, Dict, Optional, Type
+
+from django.db import transaction
+from django.utils import timezone
+
+from endoreg_db.utils import guess_name_gender
+
+# Assuming these utils are correctly located
+from endoreg_db.utils.hashs import get_patient_examination_hash, get_patient_hash
+
+# Import models needed for logic, use local imports inside functions if needed to break cycles
+from ..administration import Center, Examiner, FirstName, LastName, Patient
+from ..medical import PatientExamination
+from ..other import Gender
+
+if TYPE_CHECKING:
+    from .sensitive_meta import SensitiveMeta  # Import model for type hinting
+
+logger = logging.getLogger(__name__)
+SECRET_SALT = os.getenv("DJANGO_SALT", "default_salt")
+DEFAULT_UNKNOWN = "unknown"
+
+
+# Regex-Pattern für verschiedene Datumsformate
+ISO_RX = re.compile(r"^\d{4}-\d{2}-\d{2}$")
+DE_RX = re.compile(r"^\d{2}\.\d{2}\.\d{4}$")
+
+
+def parse_any_date(s: str) -> Optional[date]:
+    """
+    Parst Datumsstring mit Priorität auf deutsches Format (DD.MM.YYYY).
+
+    Unterstützte Formate:
+    1. DD.MM.YYYY (Priorität) - deutsches Format
+    2. YYYY-MM-DD (Fallback) - ISO-Format
+    3. Erweiterte Fallbacks über dateparser
+
+    Args:
+        s: Datumsstring zum Parsen
+
+    Returns:
+        date-Objekt oder None bei ungültigem/fehlendem Input
+    """
+    if not s:
+        return None
+
+    s = s.strip()
+
+    # 1. German dd.mm.yyyy (PRIORITÄT)
+    if DE_RX.match(s):
+        try:
+            dd, mm, yyyy = s.split(".")
+            return date(int(yyyy), int(mm), int(dd))
+        except ValueError as e:
+            logger.warning(f"Invalid German date format '{s}': {e}")
+            return None
+
+    # 2. ISO yyyy-mm-dd (Fallback für Rückwärtskompatibilität)
+    if ISO_RX.match(s):
+        try:
+            return date.fromisoformat(s)
+        except ValueError as e:
+            logger.warning(f"Invalid ISO date format '{s}': {e}")
+            return None
+
+    # 3. Extended fallbacks
+    try:
+        # Try standard datetime parsing
+        return datetime.fromisoformat(s).date()
+    except Exception:
+        pass
+
+    try:
+        # Try dateparser with German locale preference
+        import dateparser
+
+        dt = dateparser.parse(
+            s, settings={"DATE_ORDER": "DMY", "PREFER_DAY_OF_MONTH": "first"}
+        )
+        return dt.date() if dt else None
+    except Exception as e:
+        logger.debug(f"Dateparser fallback failed for '{s}': {e}")
+        return None
+
+
+def format_date_german(d: Optional[date]) -> str:
+    """
+    Formatiert date-Objekt als deutsches Datumsformat (DD.MM.YYYY).
+
+    Args:
+        d: date-Objekt oder None
+
+    Returns:
+        Formatiertes Datum als String oder leerer String bei None
+    """
+    if not d:
+        return ""
+    return d.strftime("%d.%m.%Y")
+
+
+def format_date_iso(d: Optional[date]) -> str:
+    """
+    Formatiert date-Objekt als ISO-Format (YYYY-MM-DD).
+
+    Args:
+        d: date-Objekt oder None
+
+    Returns:
+        Formatiertes Datum als String oder leerer String bei None
+    """
+    if not d:
+        return ""
+    return d.isoformat()
+
+
+def generate_random_dob() -> datetime:
+    """Generates a random timezone-aware datetime between 1920-01-01 and 2000-12-31."""
+    start_date = date(1920, 1, 1)
+    end_date = date(2000, 12, 31)
+    time_between_dates = end_date - start_date
+    days_between_dates = time_between_dates.days
+    random_number_of_days = random.randrange(days_between_dates)
+    random_date = start_date + timedelta(days=random_number_of_days)
+    random_datetime = datetime.combine(random_date, datetime.min.time())
+    return timezone.make_aware(random_datetime)
+
+
+def generate_random_examination_date() -> date:
+    """Generates a random date within the last 20 years."""
+    today = date.today()
+    start_date = today - timedelta(days=20 * 365)  # Approximate 20 years back
+    time_between_dates = today - start_date
+    days_between_dates = time_between_dates.days
+    random_number_of_days = random.randrange(days_between_dates)
+    random_date = start_date + timedelta(days=random_number_of_days)
+    return random_date
+
+
+def update_name_db(first_name: Optional[str], last_name: Optional[str]):
+    """Adds first and last names to the respective lookup tables if they don't exist."""
+    if first_name:
+        FirstName.objects.get_or_create(name=first_name)
+    if last_name:
+        LastName.objects.get_or_create(name=last_name)
+
+
+def calculate_patient_hash(instance: "SensitiveMeta", salt: str = SECRET_SALT) -> str:
+    """Calculates the patient hash for the instance."""
+    dob = instance.patient_dob
+    first_name = instance.patient_first_name
+    last_name = instance.patient_last_name
+    center = instance.center
+
+    if not dob:
+        raise ValueError("Patient DOB is required to calculate patient hash.")
+    if not center:
+        raise ValueError("Center is required to calculate patient hash.")
+
+    assert first_name is not None, "First name is required to calculate patient hash."
+    assert last_name is not None, "Last name is required to calculate patient hash."
+
+    hash_str = get_patient_hash(
+        first_name=first_name,
+        last_name=last_name,
+        dob=dob,
+        center=center.name,  # Use center name
+        salt=salt,
+    )
+    return sha256(hash_str.encode()).hexdigest()
+
+
+def calculate_examination_hash(
+    instance: "SensitiveMeta", salt: str = SECRET_SALT
+) -> str:
+    """Calculates the examination hash for the instance."""
+    dob = instance.patient_dob
+    first_name = instance.patient_first_name
+    last_name = instance.patient_last_name
+    examination_date = instance.examination_date
+    center = instance.center
+
+    if not dob:
+        raise ValueError("Patient DOB is required to calculate examination hash.")
+    if not examination_date:
+        raise ValueError("Examination date is required to calculate examination hash.")
+    if not center:
+        raise ValueError("Center is required to calculate examination hash.")
+
+    if not first_name:
+        raise ValueError("First name is required to calculate examination hash.")
+    if not last_name:
+        raise ValueError("Last name is required to calculate examination hash.")
+
+    hash_str = get_patient_examination_hash(
+        first_name=first_name,
+        last_name=last_name,
+        dob=dob,
+        examination_date=examination_date,
+        center=center.name,  # Use center name
+        salt=salt,
+    )
+    return sha256(hash_str.encode()).hexdigest()
+
+
+def create_pseudo_examiner_logic(instance: "SensitiveMeta") -> "Examiner":
+    """Creates or retrieves the pseudo examiner based on instance data."""
+    first_name = instance.examiner_first_name
+    last_name = instance.examiner_last_name
+    center = instance.center  # Should be set before calling save
+
+    if not first_name or not last_name or not center:
+        logger.warning(
+            f"Incomplete examiner info for SensitiveMeta (pk={instance.pk or 'new'}). Using default examiner."
+        )
+        # Ensure default center exists or handle appropriately
+        try:
+            default_center = Center.objects.get(name="endoreg_db_demo")
+        except Center.DoesNotExist:
+            logger.error(
+                "Default center 'endoreg_db_demo' not found. Cannot create default examiner."
+            )
+            raise ValueError("Default center 'endoreg_db_demo' not found.")
+
+        examiner, _created = Examiner.custom_get_or_create(
+            first_name="Unknown", last_name="Unknown", center=default_center
+        )
+    else:
+        examiner, _created = Examiner.custom_get_or_create(
+            first_name=first_name, last_name=last_name, center=center
+        )
+
+    return examiner
+
+
+def get_or_create_pseudo_patient_logic(instance: "SensitiveMeta"):
+    """Gets or creates the pseudo patient based on instance data."""
+    # Ensure necessary fields are set
+    if not instance.patient_hash:
+        instance.patient_hash = calculate_patient_hash(instance)
+    if not instance.center:
+        raise ValueError("Center must be set before creating pseudo patient.")
+    if not instance.patient_gender:
+        raise ValueError("Patient gender must be set before creating pseudo patient.")
+    if not instance.patient_dob:
+        raise ValueError("Patient DOB must be set before creating pseudo patient.")
+
+    dob = instance.patient_dob
+    year = dob.year
+    month = dob.month
+
+    patient, _created = Patient.get_or_create_pseudo_patient_by_hash(
+        patient_hash=instance.patient_hash,
+        center=instance.center,
+        gender=instance.patient_gender,
+        birth_year=year,
+        birth_month=month,
+    )
+    return patient, _created
+
+
+def get_or_create_pseudo_patient_examination_logic(
+    instance: "SensitiveMeta",
+):
+    """Gets or creates the pseudo patient examination based on instance data."""
+    # Ensure necessary fields are set
+    if not instance.patient_hash:
+        instance.patient_hash = calculate_patient_hash(instance)
+    if not instance.examination_hash:
+        instance.examination_hash = calculate_examination_hash(instance)
+
+    # Ensure the pseudo patient exists first, as PatientExamination might depend on it
+    if not instance.pseudo_patient:
+        pseudo_patient, _created = get_or_create_pseudo_patient_logic(instance)
+        instance.pseudo_patient = pseudo_patient  # Assign FK directly
+
+    patient_examination, _created = (
+        PatientExamination.get_or_create_pseudo_patient_examination_by_hash(
+            patient_hash=instance.patient_hash,
+            examination_hash=instance.examination_hash,
+            # Optionally pass pseudo_patient if the method requires it
+            # pseudo_patient=instance.pseudo_patient
+        )
+    )
+    return patient_examination, _created
+
+
+@transaction.atomic  # Ensure all operations within save succeed or fail together
+def perform_save_logic(instance: "SensitiveMeta") -> "Examiner":
+    """
+    Contains the core logic for preparing a SensitiveMeta instance for saving.
+    Handles data generation (dates), hash calculation, and linking pseudo-entities.
+
+    This function is called on every save() operation and implements a two-phase approach:
+
+    **Phase 1: Initial Creation (with defaults)**
+    - When a SensitiveMeta is first created (e.g., via create_from_dict),
+      it may have missing patient data (names, DOB, etc.)
+    - Default values are set to prevent hash calculation errors:
+      * patient_first_name: "unknown"
+      * patient_last_name: "unknown"
+      * patient_dob: random date (1920-2000)
+    - A temporary hash is calculated using these defaults
+    - Temporary pseudo-entities (Patient, Examination) are created
+
+    **Phase 2: Update (with extracted data)**
+    - When real patient data is extracted (e.g., from video OCR via lx_anonymizer),
+      update_from_dict() is called with actual values
+    - The instance fields are updated with real data (names, DOB, etc.)
+    - save() is called again, triggering this function
+    - Default-setting logic is skipped (fields are no longer empty)
+    - Hash is RECALCULATED with real data
+    - New pseudo-entities are created/retrieved based on new hash
+
+    **Example Flow:**
+    ```
+    # Initial creation
+    sm = SensitiveMeta.create_from_dict({"center": center})
+    # → patient_first_name = "unknown", patient_last_name = "unknown"
+    # → hash = sha256("unknown unknown 1990-01-01 ...")
+    # → pseudo_patient_temp created
+
+    # Later update with extracted data
+    sm.update_from_dict({"patient_first_name": "Max", "patient_last_name": "Mustermann"})
+    # → patient_first_name = "Max", patient_last_name = "Mustermann" (overwrites)
+    # → save() triggered → perform_save_logic() called again
+    # → Default-setting skipped (names already exist)
+    # → hash = sha256("Max Mustermann 1985-03-15 ...") (RECALCULATED)
+    # → pseudo_patient_real created/retrieved with new hash
+    ```
+
+    Args:
+        instance: The SensitiveMeta instance being saved
+
+    Returns:
+        Examiner: The pseudo examiner instance to be linked via M2M after save
+
+    Raises:
+        ValueError: If required fields (center, gender) cannot be determined
+    """
+
+    # --- Pre-Save Checks and Data Generation ---
+
+    # 1. Ensure DOB and Examination Date exist
+    if not instance.patient_dob:
+        logger.debug(
+            f"SensitiveMeta (pk={instance.pk or 'new'}): Patient DOB missing, generating random."
+        )
+        instance.patient_dob = generate_random_dob()
+    if not instance.examination_date:
+        logger.debug(
+            f"SensitiveMeta (pk={instance.pk or 'new'}): Examination date missing, generating random."
+        )
+        instance.examination_date = generate_random_examination_date()
+
+    # 2. Ensure Center exists (should be set before calling save)
+    if not instance.center:
+        raise ValueError("Center must be set before saving SensitiveMeta.")
+
+    # 2.5 CRITICAL: Set default patient names BEFORE hash calculation
+    #
+    # **Why this is necessary:**
+    # Hash calculation (step 4) requires first_name and last_name to be non-None.
+    # However, on initial creation (e.g., via get_or_create_sensitive_meta()), these
+    # fields may be empty because real patient data hasn't been extracted yet.
+    #
+    # **Two-phase approach:**
+    # - Phase 1 (Initial): Set defaults if names are missing
+    #   → Allows hash calculation to succeed without errors
+    #   → Creates temporary pseudo-entities with default hash
+    #
+    # - Phase 2 (Update): Real data extraction (OCR, manual input)
+    #   → update_from_dict() sets real names ("Max", "Mustermann")
+    #   → save() is called again
+    #   → This block is SKIPPED (names already exist)
+    #   → Hash is recalculated with real data (step 4)
+    #   → New pseudo-entities created with correct hash
+    #
+    # **Example:**
+    # Initial:  patient_first_name = "unknown" → hash = sha256("unknown unknown...")
+    # Updated:  patient_first_name = "Max"     → hash = sha256("Max Mustermann...")
+    #
+    if not instance.patient_first_name:
+        instance.patient_first_name = DEFAULT_UNKNOWN
+        logger.debug(
+            "SensitiveMeta (pk=%s): Patient first name missing, set to default '%s'.",
+            instance.pk or "new",
+            DEFAULT_UNKNOWN,
+        )
+
+    if not instance.patient_last_name:
+        instance.patient_last_name = DEFAULT_UNKNOWN
+        logger.debug(
+            "SensitiveMeta (pk=%s): Patient last name missing, set to default '%s'.",
+            instance.pk or "new",
+            DEFAULT_UNKNOWN,
+        )
+
+    # 3. Ensure Gender exists (should be set before calling save, e.g., during creation/update)
+    if not instance.patient_gender:
+        # Use the now-guaranteed first_name for gender guessing
+        first_name = instance.patient_first_name
+        gender_str = guess_name_gender(first_name)
+        if not gender_str:
+            raise ValueError(
+                "Patient gender could not be determined and must be set before saving."
+            )
+        # Convert string to Gender object
+        try:
+            gender_obj = Gender.objects.get(name=gender_str)
+            instance.patient_gender = gender_obj
+        except Gender.DoesNotExist:
+            # If the gender is 'unknown' (likely because name was DEFAULT_UNKNOWN),
+            # we should auto-create it rather than crashing.
+            if gender_str == "unknown" or instance.patient_first_name == DEFAULT_UNKNOWN:
+                logger.warning(
+                    f"Gender '{gender_str}' not found in DB. Auto-creating default entry."
+                )
+                gender_obj, _ = Gender.objects.get_or_create(
+                    name="unknown",
+                    defaults={
+                        "abbreviation": "?", 
+                        "description": "Auto-created default gender"
+                    }
+                )
+                instance.patient_gender = gender_obj
+            else:
+                # If it's a specific gender (e.g., 'male') that is missing, 
+                # that is a configuration error we should raise.
+                raise ValueError(f"Gender '{gender_str}' not found in database.")
+    # 4. Calculate Hashes (depends on DOB, Exam Date, Center, Names)
+    #
+    # **IMPORTANT: Hashes are RECALCULATED on every save!**
+    # This enables the two-phase update pattern:
+    # - Initial save: Hash based on default "unknown unknown" names
+    # - Updated save: Hash based on real extracted names ("Max Mustermann")
+    #
+    # The new hash will link to different pseudo-entities, ensuring proper
+    # anonymization while maintaining referential integrity.
+    instance.patient_hash = calculate_patient_hash(instance)
+    instance.examination_hash = calculate_examination_hash(instance)
+
+    # 5. Get or Create Pseudo Patient (depends on hash, center, gender, dob)
+    # Assign directly to the FK field to avoid premature saving issues
+    pseudo_patient, _created = get_or_create_pseudo_patient_logic(instance)
+    instance.pseudo_patient = pseudo_patient
+
+    # 6. Get or Create Pseudo Examination (depends on hashes)
+    # Assign directly to the FK field
+    pseudo_examination, _created = get_or_create_pseudo_patient_examination_logic(
+        instance
+    )
+    instance.pseudo_examination = pseudo_examination
+
+    # 7. Get or Create Pseudo Examiner (depends on names, center)
+    # This needs to happen *after* the main instance has a PK for M2M linking.
+    # We create/get it here and return it to the main save method.
+    examiner_instance = create_pseudo_examiner_logic(instance)
+
+    # 8. Ensure SensitiveMetaState exists (will be checked/created *after* main save)
+
+    # Return the examiner instance so the model's save method can handle M2M linking
+    return examiner_instance
+
+
+def create_sensitive_meta_from_dict(
+    cls: Type["SensitiveMeta"], data: Dict[str, Any]
+) -> "SensitiveMeta":
+    """
+    Create a SensitiveMeta instance from a dictionary.
+
+    **Center handling:**
+    This function accepts TWO ways to specify the center:
+    1. `center` (Center object) - Directly pass a Center instance
+    2. `center_name` (string) - Pass the center name as a string (will be resolved to Center object)
+
+    At least ONE of these must be provided.
+
+    **Example usage:**
+    ```python
+    # Option 1: With Center object
+    data = {
+        "patient_first_name": "Patient",
+        "patient_last_name": "Unknown",
+        "patient_dob": date(1990, 1, 1),
+        "examination_date": date.today(),
+        "center": center_obj,  # ← Center object
+        "text": text #from extraction
+
+    }
+    sm = SensitiveMeta.create_from_dict(data)
+
+    # Option 2: With center name string
+    data = {
+        "patient_first_name": "Patient",
+        "patient_last_name": "Unknown",
+        "patient_dob": date(1990, 1, 1),
+        "examination_date": date.today(),
+        "center_name": "university_hospital_wuerzburg",  # ← String
+        "anonymized_text": "anonymized text"
+    }
+    sm = SensitiveMeta.create_from_dict(data)
+    ```
+
+    Args:
+        cls: The SensitiveMeta class
+        data: Dictionary containing field values
+
+    Returns:
+        SensitiveMeta: The created instance
+
+    Raises:
+        ValueError: If neither center nor center_name is provided
+        ValueError: If center_name does not match any Center in database
+    """
+
+    field_names = {
+        f.name
+        for f in cls._meta.get_fields()
+        if not f.is_relation or f.one_to_one or (f.many_to_one and f.related_model)
+    }
+    selected_data = {k: v for k, v in data.items() if k in field_names}
+
+    # --- Convert patient_dob if it's a date object ---
+    dob = selected_data.get("patient_dob")
+    if isinstance(dob, date) and not isinstance(dob, datetime):
+        # Convert date to datetime at the start of the day and make it timezone-aware
+        aware_dob = timezone.make_aware(datetime.combine(dob, datetime.min.time()))
+        selected_data["patient_dob"] = aware_dob
+        logger.debug("Converted patient_dob from date to aware datetime: %s", aware_dob)
+    elif isinstance(dob, str):
+        # Handle string DOB - check if it's a field name or actual date
+        if dob == "patient_dob" or dob in [
+            "patient_first_name",
+            "patient_last_name",
+            "examination_date",
+        ]:
+            logger.warning(
+                "Skipping invalid patient_dob value '%s' - appears to be field name",
+                dob,
+            )
+            selected_data.pop("patient_dob", None)  # Remove invalid value
+        else:
+            # Try to parse as date string
+            try:
+                import dateparser
+
+                parsed_dob = dateparser.parse(
+                    dob, languages=["de"], settings={"DATE_ORDER": "DMY"}
+                )
+                if parsed_dob:
+                    aware_dob = timezone.make_aware(
+                        parsed_dob.replace(hour=0, minute=0, second=0, microsecond=0)
+                    )
+                    selected_data["patient_dob"] = aware_dob
+                    logger.debug(
+                        "Parsed string patient_dob '%s' to aware datetime: %s",
+                        dob,
+                        aware_dob,
+                    )
+                else:
+                    logger.warning(
+                        "Could not parse patient_dob string '%s', removing from data",
+                        dob,
+                    )
+                    selected_data.pop("patient_dob", None)
+            except Exception as e:
+                logger.warning(
+                    "Error parsing patient_dob string '%s': %s, removing from data",
+                    dob,
+                    e,
+                )
+                selected_data.pop("patient_dob", None)
+    # --- End Conversion ---
+
+    # Similar validation for examination_date
+    exam_date = selected_data.get("examination_date")
+    if isinstance(exam_date, str):
+        if exam_date == "examination_date" or exam_date in [
+            "patient_first_name",
+            "patient_last_name",
+            "patient_dob",
+        ]:
+            logger.warning(
+                "Skipping invalid examination_date value '%s' - appears to be field name",
+                exam_date,
+            )
+            selected_data.pop("examination_date", None)
+        else:
+            # Try to parse as date string
+            try:
+                # First try simple ISO format for YYYY-MM-DD
+                if len(exam_date) == 10 and exam_date.count("-") == 2:
+                    try:
+                        from datetime import datetime as dt
+
+                        parsed_date = dt.strptime(exam_date, "%Y-%m-%d").date()
+                        selected_data["examination_date"] = parsed_date
+                        logger.debug(
+                            "Parsed ISO examination_date '%s' to date: %s",
+                            exam_date,
+                            parsed_date,
+                        )
+                    except ValueError:
+                        # Fall back to dateparser for complex formats
+                        import dateparser
+
+                        parsed_date = dateparser.parse(
+                            exam_date, languages=["de"], settings={"DATE_ORDER": "DMY"}
+                        )
+                        if parsed_date:
+                            selected_data["examination_date"] = parsed_date.date()
+                            logger.debug(
+                                "Parsed string examination_date '%s' to date: %s",
+                                exam_date,
+                                parsed_date.date(),
+                            )
+                        else:
+                            logger.warning(
+                                "Could not parse examination_date string '%s', removing from data",
+                                exam_date,
+                            )
+                            selected_data.pop("examination_date", None)
+                else:
+                    # Use dateparser for non-ISO formats
+                    import dateparser
+
+                    parsed_date = dateparser.parse(
+                        exam_date, languages=["de"], settings={"DATE_ORDER": "DMY"}
+                    )
+                    if parsed_date:
+                        selected_data["examination_date"] = parsed_date.date()
+                        logger.debug(
+                            "Parsed string examination_date '%s' to date: %s",
+                            exam_date,
+                            parsed_date.date(),
+                        )
+                    else:
+                        logger.warning(
+                            "Could not parse examination_date string '%s', removing from data",
+                            exam_date,
+                        )
+                        selected_data.pop("examination_date", None)
+
+            except Exception as e:
+                logger.warning(
+                    "Error parsing examination_date string '%s': %s, removing from data",
+                    exam_date,
+                    e,
+                )
+                selected_data.pop("examination_date", None)
+
+    # Handle Center - accept both center_name (string) and center (object)
+    from ..administration import Center
+
+    center = data.get("center")  # First try direct Center object
+    center_name = data.get("center_name")
+
+    if center is not None:
+        # Center object provided directly - validate it's a Center instance
+        if not isinstance(center, Center):
+            raise ValueError(f"'center' must be a Center instance, got {type(center)}")
+        selected_data["center"] = center
+    elif center_name:
+        # center_name string provided - resolve to Center object
+        try:
+            center = Center.objects.get(name=center_name)
+            selected_data["center"] = center
+        except Center.DoesNotExist:
+            raise ValueError(f"Center with name '{center_name}' does not exist.")
+    else:
+        # Neither center nor center_name provided
+        raise ValueError(
+            "Either 'center' (Center object) or 'center_name' (string) is required in data dictionary."
+        )
+
+    # Handle Names and Gender
+    first_name = selected_data.get("patient_first_name") or DEFAULT_UNKNOWN
+    last_name = selected_data.get("patient_last_name") or DEFAULT_UNKNOWN
+    selected_data["patient_first_name"] = first_name  # Ensure defaults are set
+    selected_data["patient_last_name"] = last_name
+
+    patient_gender_input = selected_data.get("patient_gender")
+
+    if isinstance(patient_gender_input, Gender):
+        # Already a Gender object, nothing to do
+        pass
+    elif isinstance(patient_gender_input, str):
+        # Input is a string (gender name)
+        try:
+            selected_data["patient_gender"] = Gender.objects.get(
+                name=patient_gender_input
+            )
+        except Gender.DoesNotExist:
+            logger.warning(
+                f"Gender with name '{patient_gender_input}' provided but not found. Attempting to guess or use default."
+            )
+            # Fall through to guessing logic if provided string name is invalid
+            normalized = (patient_gender_input or "").lower()
+            if normalized in {"male", "female", "unknown"}:
+                gender_obj, _ = Gender.objects.get_or_create(
+                    name=normalized,
+                    defaults={
+                        "abbreviation": normalized[:1].upper() or None,
+                        "description": "Auto-created default gender entry",
+                    },
+                )
+                selected_data["patient_gender"] = gender_obj
+            else:
+                patient_gender_input = None  # Reset to trigger guessing
+
+    if not isinstance(
+        selected_data.get("patient_gender"), Gender
+    ):  # If not already a Gender object (e.g. was None, or string lookup failed)
+        gender_name_to_use = guess_name_gender(first_name)
+        if not gender_name_to_use:
+            logger.warning(
+                f"Could not guess gender for name '{first_name}'. Setting Gender to unknown."
+            )
+            gender_name_to_use = "unknown"
+        try:
+            selected_data["patient_gender"] = Gender.objects.get(
+                name=gender_name_to_use
+            )
+        except Gender.DoesNotExist:
+            gender_obj, _ = Gender.objects.get_or_create(
+                name=gender_name_to_use,
+                defaults={
+                    "abbreviation": gender_name_to_use[:1].upper() or None,
+                    "description": "Auto-created default gender entry",
+                },
+            )
+            selected_data["patient_gender"] = gender_obj
+
+    # Handle Text
+    selected_data["text"] = data.get("text") or DEFAULT_UNKNOWN
+
+    # --- Add missing optional fields safely ---
+    file_path = data.get("file_path")
+    if file_path:
+        selected_data["file_path"] = str(file_path)
+        logger.debug(f"Set file_path: {file_path}")
+
+    casenumber = data.get("casenumber")
+    if casenumber:
+        selected_data["casenumber"] = str(casenumber).strip()
+        logger.debug(f"Set casenumber: {casenumber}")
+
+    exam_time = data.get("examination_time")
+    if exam_time:
+        try:
+            from datetime import time as dt_time
+
+            # Accepts strings like "14:35" or full datetime
+            if isinstance(exam_time, str):
+                h, m = exam_time.strip().split(":")[:2]
+                selected_data["examination_time"] = dt_time(int(h), int(m))
+            elif isinstance(exam_time, datetime):
+                selected_data["examination_time"] = exam_time.time()
+            elif isinstance(exam_time, date):
+                # no time info — ignore
+                logger.debug(
+                    f"examination_time value {exam_time} has no time component; skipping"
+                )
+            else:
+                selected_data["examination_time"] = exam_time
+        except Exception as e:
+            logger.warning(f"Invalid examination_time '{exam_time}': {e}")
+
+    anonymized_text = data.get("anonymized_text") or data.get("anonym_text")
+    if anonymized_text:
+        if isinstance(anonymized_text, (str, bytes)):
+            selected_data["anonymized_text"] = (
+                anonymized_text.decode()
+                if isinstance(anonymized_text, bytes)
+                else anonymized_text
+            )
+        else:
+            selected_data["anonymized_text"] = str(anonymized_text)
+        logger.debug(
+            "Set anonymized_text (length=%d)", len(selected_data["anonymized_text"])
+        )
+
+    # Update name DB
+    update_name_db(first_name, last_name)
+
+    # Instantiate without saving yet
+    sensitive_meta = cls(**selected_data)
+
+    # Call save once at the end. This triggers the custom save logic.
+    sensitive_meta.save()  # This will call perform_save_logic internally
+
+    return sensitive_meta
+
+
+def update_sensitive_meta_from_dict(
+    instance: "SensitiveMeta", data: Dict[str, Any]
+) -> "SensitiveMeta":
+    """
+    Updates a SensitiveMeta instance from a dictionary of new values.
+
+    **Integration with two-phase save pattern:**
+    This function is typically called after initial SensitiveMeta creation when real
+    patient data becomes available (e.g., extracted from video OCR, report parsing, or
+    manual annotation).
+
+    **Example workflow:**
+    ```python
+    # Phase 1: Initial creation with defaults
+    sm = SensitiveMeta.create_from_dict({"center": center})
+    # → patient_first_name = "unknown", hash = sha256("unknown...")
+
+    # Phase 2: Update with extracted data
+    extracted = {
+        "patient_first_name": "Max",
+        "patient_last_name": "Mustermann",
+        "patient_dob": date(1985, 3, 15)
+    }
+    update_sensitive_meta_from_dict(sm, extracted)
+    # → Sets: sm.patient_first_name = "Max", sm.patient_last_name = "Mustermann"
+    # → Calls: sm.save()
+    # → Triggers: perform_save_logic() again
+    # → Result: Hash recalculated with real data, new pseudo-entities created
+    ```
+
+    **Key behaviors:**
+    - Updates instance attributes from provided dictionary
+    - Handles type conversions (date strings → date objects, gender strings → Gender objects)
+    - Tracks patient name changes to update name database
+    - Calls save() at the end, triggering full save logic including hash recalculation
+    - Default-setting in perform_save_logic() is skipped (fields already populated)
+
+    Args:
+        instance: The existing SensitiveMeta instance to update
+        data: Dictionary of field names and new values
+
+    Returns:
+        The updated SensitiveMeta instance
+
+    Raises:
+        Exception: If save fails or required conversions fail
+    """
+    field_names = {
+        f.name
+        for f in instance._meta.get_fields()
+        if not f.is_relation or f.one_to_one or (f.many_to_one and f.related_model)
+    }
+    # Exclude FKs that should not be updated directly from dict keys (handled separately or via save logic)
+    excluded_fields = {"pseudo_patient", "pseudo_examination"}
+    selected_data = {
+        k: v for k, v in data.items() if k in field_names and k not in excluded_fields
+    }
+
+    # Handle potential Center update - accept both center_name (string) and center (object)
+    from ..administration import Center
+
+    center = data.get("center")  # First try direct Center object
+    center_name = data.get("center_name")
+
+    if center is not None:
+        # Center object provided directly - validate and update
+        if isinstance(center, Center):
+            instance.center = center
+            logger.debug(f"Updated center from Center object: {center.name}")
+        else:
+            logger.warning(
+                f"Invalid center type {type(center)}, expected Center instance. Ignoring."
+            )
+        # Remove from selected_data to prevent override
+        selected_data.pop("center", None)
+    elif center_name:
+        # center_name string provided - resolve to Center object
+        try:
+            center_obj = Center.objects.get(name=center_name)
+            instance.center = center_obj
+            logger.debug(f"Updated center from center_name string: {center_name}")
+        except Center.DoesNotExist:
+            logger.warning(
+                f"Center '{center_name}' not found during update. Keeping existing center."
+            )
+    else:
+        # Both are None/missing - remove 'center' from selected_data to preserve existing value
+        selected_data.pop("center", None)
+    # If both are None/missing, keep existing center (no update needed)
+
+    # Set examiner names if provided, before calling save
+    examiner_first_name = data.get("examiner_first_name")
+    examiner_last_name = data.get("examiner_last_name")
+    if examiner_first_name is not None:  # Allow setting empty strings
+        instance.examiner_first_name = examiner_first_name
+    if examiner_last_name is not None:
+        instance.examiner_last_name = examiner_last_name
+
+    # Handle patient_gender specially with graceful error handling
+    patient_gender_input = data.get("patient_gender")
+    if patient_gender_input is not None:
+        try:
+            if isinstance(patient_gender_input, Gender):
+                selected_data["patient_gender"] = patient_gender_input
+            elif isinstance(patient_gender_input, str):
+                gender_input_clean = patient_gender_input.strip()
+                # Try direct case-insensitive DB lookup first
+                gender_obj = Gender.objects.filter(
+                    name__iexact=gender_input_clean
+                ).first()
+                if gender_obj:
+                    selected_data["patient_gender"] = gender_obj
+                    logger.debug(
+                        f"Successfully matched gender string '{patient_gender_input}' to Gender object via iexact lookup"
+                    )
+                else:
+                    # Use mapping helper for fallback
+                    mapped = _map_gender_string_to_standard(gender_input_clean)
+                    if mapped:
+                        gender_obj = Gender.objects.filter(name__iexact=mapped).first()
+                        if gender_obj:
+                            selected_data["patient_gender"] = gender_obj
+                            logger.info(
+                                f"Mapped gender '{patient_gender_input}' to '{mapped}' via fallback mapping"
+                            )
+                        else:
+                            logger.warning(
+                                f"Mapped gender '{patient_gender_input}' to '{mapped}', but no such Gender in DB. Trying 'unknown'."
+                            )
+                            unknown_gender = Gender.objects.filter(
+                                name__iexact="unknown"
+                            ).first()
+                            if unknown_gender:
+                                selected_data["patient_gender"] = unknown_gender
+                                logger.warning(
+                                    f"Using 'unknown' gender as fallback for '{patient_gender_input}'"
+                                )
+                            else:
+                                logger.error(
+                                    f"No 'unknown' gender found in database. Cannot handle gender '{patient_gender_input}'. Skipping gender update."
+                                )
+                                selected_data.pop("patient_gender", None)
+                    else:
+                        # Last resort: try to get 'unknown' gender
+                        unknown_gender = Gender.objects.filter(
+                            name__iexact="unknown"
+                        ).first()
+                        if unknown_gender:
+                            selected_data["patient_gender"] = unknown_gender
+                            logger.warning(
+                                f"Using 'unknown' gender as fallback for '{patient_gender_input}' (no mapping)"
+                            )
+                        else:
+                            logger.error(
+                                f"No 'unknown' gender found in database. Cannot handle gender '{patient_gender_input}'. Skipping gender update."
+                            )
+                            selected_data.pop("patient_gender", None)
+            else:
+                logger.warning(
+                    f"Unexpected patient_gender type {type(patient_gender_input)}: {patient_gender_input}. Skipping gender update."
+                )
+                selected_data.pop("patient_gender", None)
+        except Exception as e:
+            logger.exception(
+                f"Error handling patient_gender '{patient_gender_input}': {e}. Skipping gender update."
+            )
+            selected_data.pop("patient_gender", None)
+
+    # TODO Review: Handle new optional fields on update
+    for key in (
+        "file_path",
+        "casenumber",
+        "examination_time",
+        "anonymized_text",
+        "anonym_text",
+    ):
+        if key in data and data[key] is not None:
+            val = data[key]
+            if key in ("file_path", "casenumber"):
+                setattr(instance, key, str(val))
+            elif key in ("anonymized_text", "anonym_text"):
+                setattr(
+                    instance,
+                    "anonymized_text",
+                    val if isinstance(val, str) else str(val),
+                )
+            elif key == "examination_time":
+                try:
+                    from datetime import time as dt_time
+
+                    if isinstance(val, str) and ":" in val:
+                        h, m = val.strip().split(":")[:2]
+                        setattr(instance, "examination_time", dt_time(int(h), int(m)))
+                    elif isinstance(val, datetime):
+                        setattr(instance, "examination_time", val.time())
+                except Exception as e:
+                    logger.warning(f"Skipping invalid examination_time '{val}': {e}")
+
+    # Update other attributes from selected_data
+    patient_name_changed = False
+    for k, v in selected_data.items():
+        # Skip None values to avoid overwriting existing data
+        if v is None:
+            logger.debug(f"Skipping field '{k}' during update because value is None")
+            continue
+
+        # Avoid overwriting examiner names if they were just explicitly set
+        if (
+            k not in ["examiner_first_name", "examiner_last_name"]
+            or (k == "examiner_first_name" and examiner_first_name is None)
+            or (k == "examiner_last_name" and examiner_last_name is None)
+        ):
+            try:
+                # --- Convert patient_dob if it's a date object ---
+                value_to_set = v
+                if k == "patient_dob":
+                    if isinstance(v, date) and not isinstance(v, datetime):
+                        aware_dob = timezone.make_aware(
+                            datetime.combine(v, datetime.min.time())
+                        )
+                        value_to_set = aware_dob
+                        logger.debug(
+                            "Converted patient_dob from date to aware datetime during update: %s",
+                            aware_dob,
+                        )
+                    elif isinstance(v, str):
+                        parsed = parse_any_date(v)
+                        if parsed:
+                            aware_dob = timezone.make_aware(
+                                datetime.combine(parsed, datetime.min.time())
+                            )
+                            value_to_set = aware_dob
+                            logger.debug(
+                                "Parsed string patient_dob '%s' during update to aware datetime: %s",
+                                v,
+                                aware_dob,
+                            )
+                        else:
+                            logger.warning(
+                                "Could not parse patient_dob string '%s' during update, skipping",
+                                v,
+                            )
+                            continue
+                elif k == "examination_date":
+                    if isinstance(v, str):
+                        parsed = parse_any_date(v)
+                        if parsed:
+                            value_to_set = (
+                                parsed  # field is DateField, so keep it as date
+                            )
+                            logger.debug(
+                                "Parsed string examination_date '%s' during update to date: %s",
+                                v,
+                                value_to_set,
+                            )
+                        else:
+                            logger.warning(
+                                "Could not parse examination_date string '%s' during update, skipping",
+                                v,
+                            )
+                            continue
+                    elif isinstance(v, date):
+                        value_to_set = v
+
+                        # --- End Conversion ---
+
+                # Check if patient name is changing
+                if (
+                    k in ["patient_first_name", "patient_last_name"]
+                    and getattr(instance, k) != value_to_set
+                ):
+                    patient_name_changed = True
+
+                setattr(instance, k, value_to_set)  # Use value_to_set
+
+            except Exception as e:
+                logger.error(
+                    f"Error setting attribute '{k}' to '{v}': {e}. Skipping this field."
+                )
+                continue
+
+    # Update name DB if patient names changed
+    if patient_name_changed:
+        try:
+            update_name_db(instance.patient_first_name, instance.patient_last_name)
+        except Exception as e:
+            logger.warning(f"Error updating name database: {e}")
+
+    # Call save - this will trigger the full save logic including hash recalculation etc.
+    try:
+        instance.save()
+    except Exception as e:
+        logger.error(f"Error saving SensitiveMeta instance: {e}")
+        raise
+
+    return instance
+
+
+def update_or_create_sensitive_meta_from_dict(
+    cls: Type["SensitiveMeta"],
+    data: Dict[str, Any],
+    instance: Optional["SensitiveMeta"] = None,
+):
+    """Logic to update or create a SensitiveMeta instance from a dictionary."""
+    # Check if the instance already exists based on unique fields
+    sensitive_meta: "SensitiveMeta"
+    _created: bool
+    if instance:
+        # Update the existing instance
+        sensitive_meta = update_sensitive_meta_from_dict(instance, data)
+        _created = False
+
+    else:
+        # Create a new instance
+        sensitive_meta = create_sensitive_meta_from_dict(cls, data)
+        _created = True
+    return sensitive_meta, _created
+
+
+def _map_gender_string_to_standard(gender_str: str) -> Optional[str]:
+    """Maps various gender string inputs to standard gender names used in the DB."""
+    mapping = {
+        "male": ["male", "m", "männlich", "man"],
+        "female": ["female", "f", "weiblich", "woman"],
+        "unknown": ["unknown", "unbekannt", "other", "diverse", ""],
+    }
+    gender_lower = gender_str.strip().lower()
+    for standard, variants in mapping.items():
+        if gender_lower in variants:
+            return standard
+    return None
+
+
+def _create_anonymized_record(
+    instance: "SensitiveMeta",
+    DEFAULT_ANONYMIZED=None,
+    DEFAULT_ANONYMIZED_DATE=timezone.make_aware(datetime(1900, 1, 1)),
+) -> None:
+    """
+    Create a SensitiveMeta instance with all sensitive fields set to anonymized defaults.
+    This is only called after anonymization and will delete all data that can identify a patient from the database.
+    What is left will only be the patient hash.
+
+    Args:
+        instance: The existing SensitiveMeta instance to anonymize
+        DEFAULT_ANONYMIZED: Usually None, The default string to use for anonymized fields (e.g., "anonymized,")
+    """
+
+    instance.refresh_from_db()
+    instance.get_patient_hash()
+    instance.get_patient_examination_hash()
+
+    anonymized_data = {
+        "patient_first_name": DEFAULT_ANONYMIZED,
+        "patient_last_name": DEFAULT_ANONYMIZED,
+        "patient_dob": DEFAULT_ANONYMIZED_DATE,
+        "examination_date": DEFAULT_ANONYMIZED_DATE,
+    }
+    sensitive_meta = update_sensitive_meta_from_dict(instance, anonymized_data)
+
+    sensitive_meta.save()