File size: 7,231 Bytes
9686dbe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
45dc8b6
9686dbe
45dc8b6
 
9686dbe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
"""
MamaGuard -- FastAPI Server
Maternal risk prediction API with clinical safety net and explainability.
"""

import torch
import numpy as np
import pickle
import os
from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from fastapi.staticfiles import StaticFiles
from fastapi.responses import FileResponse

from api.schemas import PredictionRequest, PredictionResponse, AlertTier
from api.alert_logic import compute_alert_tier, generate_action_text
from src.model import MamaGuardMamba3
from src.explainability import explain_prediction
from api.extract_report import router as extract_router

# --- App setup ----------------------------------------------------------------
app = FastAPI(
    title="SheGuard -- Maternal Risk API",
    description=(
        "Predicts maternal mortality risk from prenatal visit sequences. "
        "Built with Mamba3 SSM for deployment in low-resource clinics."
    ),
    version="1.0.0",
)

app.include_router(extract_router)

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_methods=["*"],
    allow_headers=["*"],
)

# ─── Model loading ────────────────────────────────────────────────────────────

MODEL_PATH  = "models/mamaguard_mamba3.pt"
SCALER_PATH = "models/scaler.pkl"

model  = None
scaler = None
device = "cuda" if torch.cuda.is_available() else "cpu"

FEATURE_ORDER = ['Age', 'SystolicBP', 'DiastolicBP', 'BS', 'BodyTemp', 'HeartRate']
FEATURE_DEFAULTS = {
    'Age': 30.0, 'SystolicBP': 120.0, 'DiastolicBP': 80.0,
    'BS': 7.5, 'BodyTemp': 36.8, 'HeartRate': 76.0
}


@app.on_event("startup")
async def load_model():
    """Load model and scaler at server startup."""
    global model, scaler

    if not os.path.exists(MODEL_PATH):
        print(f"WARNING: Model not found at {MODEL_PATH}. Run training first.")
        return

    model = MamaGuardMamba3(input_dim=6, d_model=64, n_layers=4, n_classes=3, d_state=32)
    model.load_state_dict(torch.load(MODEL_PATH, map_location=device))
    model.to(device)
    model.eval()

    with open(SCALER_PATH, "rb") as f:
        scaler = pickle.load(f)

    print(f"MamaGuard model loaded on {device}")


# --- Routes -------------------------------------------------------------------

@app.get("/health")
async def health_check():
    """Health check endpoint for load balancers."""
    return {
        "status":       "healthy",
        "model_loaded": model is not None,
        "device":       device,
        "version":      "1.0.0",
    }


@app.post("/predict", response_model=PredictionResponse)
async def predict(request: PredictionRequest):
    """
    Process patient visit data and return risk prediction with explanation.
    """

    if model is None or scaler is None:
        raise HTTPException(
            status_code=503,
            detail="Model not loaded. Contact system administrator."
        )

    # Prepare visit data
    visits = request.visits
    visit_map = {
        'Age': 'age', 'SystolicBP': 'systolic_bp', 'DiastolicBP': 'diastolic_bp',
        'BS': 'blood_sugar', 'BodyTemp': 'body_temp', 'HeartRate': 'heart_rate'
    }

    raw_array = []
    missing_counts = []

    for visit in visits:
        row = []
        missing = 0
        for feat in FEATURE_ORDER:
            attr = visit_map[feat]
            val  = getattr(visit, attr, None)
            if val is None:
                val = FEATURE_DEFAULTS[feat]
                missing += 1
            row.append(val)
        raw_array.append(row)
        missing_counts.append(missing)

    raw_np = np.array(raw_array, dtype=np.float32)

    # Data quality score
    total_fields  = len(visits) * len(FEATURE_ORDER)
    missing_total = sum(missing_counts)
    data_quality  = round(1.0 - missing_total / total_fields, 3)

    # Scale
    scaled_np = scaler.transform(raw_np)

    # Pad or truncate to SEQ_LEN=5
    SEQ_LEN = 5
    n = len(scaled_np)
    if n < SEQ_LEN:
        pad = np.zeros((SEQ_LEN - n, scaled_np.shape[1]), dtype=np.float32)
        scaled_np = np.vstack([pad, scaled_np])
    elif n > SEQ_LEN:
        scaled_np = scaled_np[-SEQ_LEN:]

    # Clinical safety net (hard WHO rules)
    from api.alert_logic import apply_clinical_safety_net, AlertTier

    visits_raw = [v.model_dump() for v in request.visits]
    forced_tier, forced_reason = apply_clinical_safety_net(visits_raw)

    # Run model and explain
    explanation = explain_prediction(model, scaled_np, scaler, device)

    if forced_tier is not None:
        explanation["top_reasons"].insert(0, f"[WHO guideline] {forced_reason}")

    # Alert tier
    alert_tier, suppressed = compute_alert_tier(
        probabilities    = explanation["probabilities"],
        patient_id       = request.patient_id,
        visit_importances= explanation["visit_importance"],
    )

    # Override with clinical rule minimum if needed
    tier_order = {AlertTier.GREEN: 0, AlertTier.AMBER: 1, AlertTier.RED: 2}
    if forced_tier is not None:
        if tier_order[forced_tier] > tier_order[alert_tier]:
            alert_tier = forced_tier
            suppressed = False

    # Action text
    action, transfer_order = generate_action_text(
        tier             = alert_tier,
        top_reasons      = explanation["top_reasons"],
        data_quality     = data_quality,
        staff_available  = request.staff_available,
        blood_units      = request.blood_units,
    )

    return PredictionResponse(
        patient_id        = request.patient_id,
        risk_level        = explanation["risk_level"],
        alert_tier        = alert_tier,
        confidence        = explanation["confidence"],
        data_quality      = data_quality,
        top_reasons       = explanation["top_reasons"],
        feature_importance= explanation["feature_importance"],
        visit_importance  = explanation["visit_importance"],
        action_required   = action,
        transfer_order    = transfer_order,
        suppressed        = suppressed,
        probabilities     = explanation["probabilities"],
    )


@app.get("/stats")
async def get_stats():
    """Returns patient assessment statistics."""
    from api.alert_logic import _alert_history
    total   = len(_alert_history)
    by_tier = {"GREEN": 0, "AMBER": 0, "RED": 0}
    for v in _alert_history.values():
        by_tier[v["tier"].value] += 1

    return {
        "patients_assessed": total,
        "by_tier": by_tier,
        "model_version": "SheGuard-Mamba3-v1",
    }


# --- Dashboard static files ---------------------------------------------------

DASHBOARD_DIR = os.path.join(os.path.dirname(os.path.dirname(__file__)), "dashboard")

if os.path.isdir(DASHBOARD_DIR):
    app.mount("/static", StaticFiles(directory=DASHBOARD_DIR), name="dashboard-static")


@app.get("/")
async def serve_dashboard():
    """Serve the dashboard HTML at root."""
    index_path = os.path.join(DASHBOARD_DIR, "index.html")
    if os.path.exists(index_path):
        return FileResponse(index_path, media_type="text/html")
    return {"status": "ok", "model_loaded": model is not None, "device": device}