feat(lab): Lab Report Analysis - OCR + Groq AI

.env.example

@@ -34,5 +34,8 @@ MAIL_USERNAME=
 MAIL_PASSWORD=
 MAIL_DEFAULT_SENDER=
 
+# Lab report AI (Groq API-only, no local model)
+GROQ_API_KEY=
+
 # Set ALLOW_DROP_TABLES=true ONLY for a deliberate local reset. Never on prod.
 ALLOW_DROP_TABLES=false

app.py

@@ -562,6 +562,41 @@ def create_app():
         except Exception as e:
             return jsonify({'success': False, 'error': str(e)}), 500
 
+    @app.route('/api/extract_lab_report', methods=['POST'])
+    def extract_lab_report():
+        try:
+            if 'file' not in request.files:
+                return jsonify({'success': False, 'error': 'No file uploaded.'})
+            f = request.files['file']
+            if not f.filename:
+                return jsonify({'success': False, 'error': 'No file selected.'})
+            import uuid
+            from werkzeug.utils import secure_filename
+            fname = f"{uuid.uuid4().hex}_{secure_filename(f.filename)}"
+            fpath = os.path.join(app.config['UPLOAD_FOLDER'], fname)
+            f.save(fpath)
+            from lab_prediction import process_lab_image
+            return jsonify(process_lab_image(fpath))
+        except Exception as e:
+            return jsonify({'success': False, 'error': str(e)}), 500
+
+    @app.route('/api/predict_lab_report', methods=['POST'])
+    def predict_lab_report():
+        try:
+            data = request.json or {}
+            from lab_prediction import predict_lab_disease, FEATURES
+            values = {k: (data.get(k) if data.get(k) != '' else None) for k in FEATURES}
+            for k in list(values):
+                v = values[k]
+                if k != 'Gender' and v is not None:
+                    try:
+                        values[k] = float(v)
+                    except (ValueError, TypeError):
+                        pass
+            return jsonify(predict_lab_disease(values))
+        except Exception as e:
+            return jsonify({'success': False, 'error': str(e)}), 500
+
     @app.route('/api/save_gastro_audit', methods=['POST'])
     def save_gastro_audit():
         try:

lab_prediction.py

@@ -0,0 +1,127 @@
+"""Lab report disease prediction — Groq API only (no local .pkl model).
+
+Pipeline:
+  process_lab_image(path) -> OCR (ocr_utils) -> 13-field dict + raw text
+  predict_lab_disease(values) -> Groq chat-completions -> disease + confidence
+
+Predictable diseases: Anemia, Asthma, Heart Attack, Liver Disease,
+Infection, Diabetics, Cardiovascular Disease, Pancreatitis, Kidney Disease.
+"""
+import json
+import os
+import re
+
+import requests
+
+FEATURES = [
+    "Gender", "Age", "Hemoglobin", "RBC", "WBC", "AST", "ALT",
+    "Cholesterol", "Spirometry", "Creatinine", "Glucose", "Lipase", "Troponin",
+]
+
+DISEASES = [
+    "Anemia", "Asthma", "Heart Attack", "Liver Disease", "Infection",
+    "Diabetics", "Cardiovascular Disease", "Pancreatitis", "Kidney Disease",
+]
+
+MODEL_CANDIDATES = [
+    "openai/gpt-oss-120b",
+    "openai/gpt-oss-20b",
+    "qwen/qwen3-32b",
+    "llama-3.3-70b-versatile",
+    "llama-3.1-8b-instant",
+]
+
+GROQ_URL = "https://api.groq.com/openai/v1/chat/completions"
+
+
+def _groq_key() -> str:
+    return os.environ.get("GROQ_API_KEY", "")
+
+
+def process_lab_image(image_path: str) -> dict:
+    """OCR a lab report image -> {success, extracted_values, raw_text}."""
+    from ocr_utils import ocr_image, extract_lab_values
+    try:
+        raw = ocr_image(image_path)
+    except Exception as e:
+        return {"success": False, "error": f"OCR failed: {e}"}
+    return {
+        "success": True,
+        "extracted_values": extract_lab_values(raw),
+        "raw_text": raw[:2000],
+    }
+
+
+def _prompt(values: dict) -> str:
+    lines = "\n".join(f"- {k}: {values.get(k)}" for k in FEATURES)
+    return (
+        "You are a clinical lab assistant. Given these lab values:\n"
+        f"{lines}\n\n"
+        f"Possible diseases: {', '.join(DISEASES)}.\n"
+        "Reply with ONLY valid JSON, no markdown, no extra text, "
+        'in this shape: {"disease": "<one of the list>", '
+        '"confidence": <0-100 number>, '
+        '"top3": [{"disease": "<name>", "confidence": <0-100>}]}'
+    )
+
+
+def _parse_reply(text: str) -> dict:
+    cleaned = text.strip()
+    if cleaned.startswith("```"):
+        cleaned = re_sub_fences(cleaned)
+    data = json.loads(cleaned)
+    disease = str(data.get("disease", "")).strip()
+    conf = float(data.get("confidence", 0))
+    top3 = data.get("top3") or [{"disease": disease, "confidence": conf}]
+    return {"disease": disease, "confidence": round(conf, 1), "top3": top3}
+
+
+def re_sub_fences(s: str) -> str:
+    s = re.sub(r"^```[a-zA-Z]*\n?", "", s)
+    s = re.sub(r"\n?```$", "", s)
+    return s.strip()
+
+
+def predict_lab_disease(values: dict) -> dict:
+    """Call Groq (model fallback chain) -> disease prediction dict."""
+    api_key = _groq_key()
+    if not api_key:
+        return {"success": False, "error": "GROQ_API_KEY is not set on the server."}
+
+    missing = [k for k in FEATURES if values.get(k) in (None, "")]
+    prompt = _prompt(values)
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+    last_err = "unknown error"
+
+    for model in MODEL_CANDIDATES:
+        try:
+            r = requests.post(
+                GROQ_URL,
+                headers=headers,
+                json={
+                    "model": model,
+                    "messages": [{"role": "user", "content": prompt}],
+                    "temperature": 0.2,
+                    "max_tokens": 512,
+                },
+                timeout=60,
+            )
+            if r.status_code != 200:
+                last_err = f"{model}: HTTP {r.status_code} {r.text[:200]}"
+                continue
+            content = r.json()["choices"][0]["message"]["content"]
+            parsed = _parse_reply(content)
+            if parsed["disease"] not in DISEASES:
+                # Accept model answer anyway but flag it
+                pass
+            parsed.update({
+                "success": True,
+                "missing": missing,
+                "accuracy": f"Groq AI ({model})",
+            })
+            return parsed
+        except Exception as e:
+            last_err = f"{model}: {e}"
+            continue
+
+    return {"success": False, "error": f"AI Prediction Failed: {last_err}"}

ocr_utils.py

@@ -0,0 +1,67 @@
+"""OCR helpers for lab report images (EasyOCR + OpenCV + regex).
+
+13 extracted fields: Gender, Age, Hemoglobin, RBC, WBC, AST, ALT,
+Cholesterol, Spirometry, Creatinine, Glucose, Lipase, Troponin.
+Missing values are returned as None.
+"""
+import re
+
+_reader = None
+
+
+def get_reader():
+    """Lazy singleton EasyOCR reader (downloads ~100MB weights on first call)."""
+    global _reader
+    if _reader is None:
+        import easyocr
+        _reader = easyocr.Reader(['en'], gpu=False)
+    return _reader
+
+
+def ocr_image(image_path: str) -> str:
+    """Run OCR on an image file and return raw text."""
+    import cv2
+    reader = get_reader()
+    img = cv2.imread(image_path)
+    if img is None:
+        raise ValueError("Could not read image. Use JPG/PNG.")
+    # EasyOCR handles numpy arrays (BGR) directly.
+    results = reader.readtext(img)
+    return " ".join([t[1] for t in results])
+
+
+def _num(patterns, text, cast=float):
+    for pat in patterns:
+        m = re.search(pat, text, re.IGNORECASE)
+        if m:
+            try:
+                return cast(m.group(1).replace(',', ''))
+            except (ValueError, IndexError):
+                continue
+    return None
+
+
+def extract_lab_values(text: str) -> dict:
+    """Regex-extract the 13 lab fields from OCR text (None if not found)."""
+    t = text or ""
+    gender = None
+    if re.search(r'\bfemale\b', t, re.IGNORECASE):
+        gender = "Female"
+    elif re.search(r'\bmale\b', t, re.IGNORECASE):
+        gender = "Male"
+
+    return {
+        "Gender": gender,
+        "Age": _num([r'age\D{0,10}(\d{1,3})'], t, int),
+        "Hemoglobin": _num([r'h(?:ae)?moglobin\D{0,15}(\d+\.?\d*)', r'\bHb\D{0,10}(\d+\.?\d*)'], t),
+        "RBC": _num([r'\bRBC\D{0,10}(\d+\.?\d*)'], t),
+        "WBC": _num([r'\bWBC\D{0,10}([\d,]+\.?\d*)', r'W\.?B\.?C\.?\D{0,10}([\d,]+\.?\d*)'], t),
+        "AST": _num([r'\bAST\D{0,10}(\d+\.?\d*)', r'\bSGOT\D{0,10}(\d+\.?\d*)'], t),
+        "ALT": _num([r'\bALT\D{0,10}(\d+\.?\d*)', r'\bSGPT\D{0,10}(\d+\.?\d*)'], t),
+        "Cholesterol": _num([r'cholesterol\D{0,15}(\d+\.?\d*)'], t),
+        "Spirometry": _num([r'spirometry\D{0,15}(\d+\.?\d*)', r'\bFEV1?\D{0,10}(\d+\.?\d*)'], t),
+        "Creatinine": _num([r'creatinine\D{0,15}(\d+\.?\d*)'], t),
+        "Glucose": _num([r'glucose\D{0,15}(\d+\.?\d*)', r'blood sugar\D{0,15}(\d+\.?\d*)'], t),
+        "Lipase": _num([r'lipase\D{0,15}(\d+\.?\d*)'], t),
+        "Troponin": _num([r'troponin\D{0,15}(\d+\.?\d*)'], t),
+    }

requirements.txt

@@ -17,5 +17,8 @@ python-dotenv
 # --- AI image model (CPU builds recommended, see Dockerfile) ---
 torch
 torchvision
+# --- Lab report OCR ---
+opencv-python==4.8.1.78
+easyocr==1.7.1
 # --- Optional: GCP Cloud SQL Python Connector (only needed if USE_CLOUD_SQL_CONNECTOR=true) ---
 # cloud-sql-python-connector[pg8000]

routes/doctor.py

@@ -281,3 +281,8 @@ def gastro_analysis():
 @login_required
 def urology_analysis():
     return render_template('doctor/urology_analysis.html')
+
+@doctor_bp.route('/lab_analysis')
+@login_required
+def lab_analysis():
+    return render_template('doctor/lab_analysis.html')

templates/doctor/lab_analysis.html

@@ -0,0 +1,124 @@
+{% extends 'layout.html' %}
+{% block content %}
+<div class="container-fluid py-4">
+    <div class="row mb-4">
+        <div class="col-12">
+            <h2>Lab Report Analysis</h2>
+            <p class="text-muted">Upload a lab report image to auto-fill values, verify them, then predict disease (Groq AI).</p>
+        </div>
+    </div>
+    <div class="row justify-content-center">
+        <div class="col-12 col-xl-10">
+            <div class="card shadow-sm mb-4">
+                <div class="card-body p-4">
+                    <h5 class="card-title mb-3"><i data-lucide="upload" class="me-2"></i>1. Upload Lab Report (JPG/PNG)</h5>
+                    <div class="d-flex gap-2 align-items-center flex-wrap">
+                        <input type="file" id="lab-file" class="form-control" accept=".jpg,.jpeg,.png,image/*" style="max-width: 400px;">
+                        <button type="button" class="btn btn-secondary" id="btn-extract">
+                            <i data-lucide="scan" class="me-2"></i>Extract Values
+                        </button>
+                        <span class="text-muted small" id="extract-status">No file processed.</span>
+                    </div>
+                </div>
+            </div>
+
+            <form id="labForm">
+                <div class="card shadow-sm mb-4">
+                    <div class="card-body p-4">
+                        <h5 class="card-title mb-3"><i data-lucide="clipboard-check" class="me-2"></i>2. Verify / Enter Values</h5>
+                        <div class="row g-2">
+                            <div class="col-md-3"><label class="form-label">Gender</label><select class="form-select" id="Gender"><option value="">--</option><option>Male</option><option>Female</option></select></div>
+                            <div class="col-md-3"><label class="form-label">Age</label><input type="number" class="form-control" id="Age" step="1"></div>
+                            <div class="col-md-3"><label class="form-label">Hemoglobin</label><input type="number" class="form-control" id="Hemoglobin" step="0.1"></div>
+                            <div class="col-md-3"><label class="form-label">RBC</label><input type="number" class="form-control" id="RBC" step="0.01"></div>
+                            <div class="col-md-3"><label class="form-label">WBC</label><input type="number" class="form-control" id="WBC" step="1"></div>
+                            <div class="col-md-3"><label class="form-label">AST</label><input type="number" class="form-control" id="AST" step="0.1"></div>
+                            <div class="col-md-3"><label class="form-label">ALT</label><input type="number" class="form-control" id="ALT" step="0.1"></div>
+                            <div class="col-md-3"><label class="form-label">Cholesterol</label><input type="number" class="form-control" id="Cholesterol" step="1"></div>
+                            <div class="col-md-3"><label class="form-label">Spirometry</label><input type="number" class="form-control" id="Spirometry" step="0.01"></div>
+                            <div class="col-md-3"><label class="form-label">Creatinine</label><input type="number" class="form-control" id="Creatinine" step="0.01"></div>
+                            <div class="col-md-3"><label class="form-label">Glucose</label><input type="number" class="form-control" id="Glucose" step="1"></div>
+                            <div class="col-md-3"><label class="form-label">Lipase</label><input type="number" class="form-control" id="Lipase" step="1"></div>
+                            <div class="col-md-3"><label class="form-label">Troponin</label><input type="number" class="form-control" id="Troponin" step="0.001"></div>
+                        </div>
+                        <div class="d-flex justify-content-end mt-3">
+                            <button type="button" class="btn btn-primary px-4 py-2 fw-bold shadow-sm" id="btn-predict-lab">
+                                <i data-lucide="cpu" class="me-2"></i>Predict Disease
+                            </button>
+                        </div>
+                    </div>
+                </div>
+            </form>
+
+            <div id="lab-result" style="display:none;" class="card shadow-sm border-0">
+                <div class="card-body p-4">
+                    <h5 class="card-title text-primary border-bottom pb-2 mb-3"><i data-lucide="activity" class="me-2"></i>Prediction Result</h5>
+                    <div class="alert alert-info mb-3">
+                        <div class="small text-muted text-uppercase fw-bold">Predicted Disease</div>
+                        <h4 id="lab-disease" class="mb-1"></h4>
+                        <small class="text-muted">Confidence: <span id="lab-confidence"></span>% (<span id="lab-model"></span>)</small>
+                    </div>
+                    <h6>Top predictions</h6>
+                    <ul id="lab-top3" class="list-group mb-0"></ul>
+                </div>
+            </div>
+        </div>
+    </div>
+</div>
+
+<script>
+const LAB_FIELDS = ["Gender","Age","Hemoglobin","RBC","WBC","AST","ALT","Cholesterol","Spirometry","Creatinine","Glucose","Lipase","Troponin"];
+
+document.getElementById('btn-extract').addEventListener('click', function () {
+    const fileInput = document.getElementById('lab-file');
+    const status = document.getElementById('extract-status');
+    if (!fileInput.files.length) { alert('Select an image first.'); return; }
+    const btn = this; btn.disabled = true; status.innerText = 'Running OCR...';
+    const fd = new FormData(); fd.append('file', fileInput.files[0]);
+    fetch('/api/extract_lab_report', { method: 'POST', body: fd })
+        .then(r => r.json())
+        .then(data => {
+            btn.disabled = false;
+            if (data.success) {
+                const v = data.extracted_values || {};
+                LAB_FIELDS.forEach(f => {
+                    const el = document.getElementById(f);
+                    if (el && v[f] !== null && v[f] !== undefined) el.value = v[f];
+                });
+                status.innerText = 'Values extracted — please verify.';
+            } else {
+                status.innerText = 'Extract failed: ' + data.error;
+            }
+        })
+        .catch(e => { btn.disabled = false; status.innerText = 'Error: ' + e; });
+});
+
+document.getElementById('btn-predict-lab').addEventListener('click', function () {
+    const payload = {};
+    LAB_FIELDS.forEach(f => { payload[f] = document.getElementById(f).value; });
+    const btn = this; btn.disabled = true; btn.innerText = 'Predicting...';
+    fetch('/api/predict_lab_report', { method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify(payload) })
+        .then(r => r.json())
+        .then(data => {
+            btn.disabled = false; btn.innerHTML = 'Predict Disease';
+            if (data.success) {
+                document.getElementById('lab-result').style.display = 'block';
+                document.getElementById('lab-disease').innerText = data.disease;
+                document.getElementById('lab-confidence').innerText = data.confidence;
+                document.getElementById('lab-model').innerText = data.accuracy || '';
+                const ul = document.getElementById('lab-top3'); ul.innerHTML = '';
+                (data.top3 || []).forEach(t => {
+                    const li = document.createElement('li');
+                    li.className = 'list-group-item d-flex justify-content-between';
+                    li.innerHTML = '<span>' + t.disease + '</span><strong>' + t.confidence + '%</strong>';
+                    ul.appendChild(li);
+                });
+                document.getElementById('lab-result').scrollIntoView({ behavior: 'smooth' });
+            } else {
+                alert('AI Prediction Failed: ' + data.error);
+            }
+        })
+        .catch(e => { btn.disabled = false; alert('Error: ' + e); });
+});
+</script>
+{% endblock %}

templates/layout.html

@@ -49,6 +49,12 @@
                         <span>Urology AI Form</span>
                     </a>
                 </li>
+                <li class="sidebar-menu-item {% if request.endpoint == 'doctor.lab_analysis' %}active{% endif %}">
+                    <a href="{{ url_for('doctor.lab_analysis') }}">
+                        <i data-lucide="flask-conical"></i>
+                        <span>Lab Report</span>
+                    </a>
+                </li>
                 {% elif current_user.role == 'patient' %}
                 <li class="sidebar-menu-item {% if request.endpoint == 'patient.dashboard' %}active{% endif %}">
                     <a href="{{ url_for('patient.dashboard') }}">