api1.py

import cv2
import time
import threading
import numpy as np
from datetime import datetime
from fastapi import FastAPI, UploadFile, File
from fastapi.staticfiles import StaticFiles
from ultralytics import YOLO
from PIL import Image
import os

# -----------------------------
# 1. Config & Model
# -----------------------------
MODEL_STROKE_PATH = "stroke.pt"
OUTPUT_DIR = "/tmp/outputs"
os.makedirs(OUTPUT_DIR, exist_ok=True)

# Charger YOLO une seule fois
model_stroke = YOLO(MODEL_STROKE_PATH)

BASE_URL = "https://stroke-ia-avc-detect.hf.space"  # ⚠️ à adapter selon ton déploiement

# Mapping des classes vers un rapport médical
CLASS_LABELS = {
    0: "Hémorragie intracrânienne",
    1: "Suspicion de zone ischémique",
    2: "Normale Brain",  # 👉 adapte en fonction des classes de ton modèle
}

# -----------------------------
# 2. Génération de rapport
# -----------------------------
def generate_report(results) -> str:
    boxes = results[0].boxes
    if len(boxes) == 0:
        return "=== RAPPORT AUTOMATIQUE ===\n\nAucune anomalie détectée.\n"

    rapport = "=== RAPPORT AUTOMATIQUE AVC ===\n\n"
    rapport += f"Nombre de lésions détectées : {len(boxes)}\n\n"

    detected_classes = boxes.cls.cpu().numpy().astype(int)
    for i, cls_id in enumerate(detected_classes, 1):
        label = CLASS_LABELS.get(cls_id, f"Classe inconnue {cls_id}")
        rapport += f"- Lésion {i}: {label}\n"

    rapport += "\nRecommandations :\n"
    rapport += "- Vérifier la concordance clinique.\n"
    rapport += "- Considérer un suivi neurologique urgent.\n"

    return rapport

# -----------------------------
# 3. FastAPI
# -----------------------------
app = FastAPI(title="Stroke Detection API")
app.mount("/files", StaticFiles(directory=OUTPUT_DIR), name="files")

@app.post("/predict/")
async def predict_stroke(image_file: UploadFile = File(...), conf: float = 0.8):
    """
    Endpoint qui reçoit une image IRM et renvoie une image annotée + rapport texte
    """
    # Sauvegarde temporaire
    tmp_path = f"/tmp/{image_file.filename}"
    with open(tmp_path, "wb") as f:
        f.write(await image_file.read())

    # Charger image
    image = Image.open(tmp_path).convert("RGB")
    np_img = np.array(image)

    # Conversion en BGR pour OpenCV
    np_img = cv2.cvtColor(np_img, cv2.COLOR_RGB2BGR)

    # Prédiction
    results = model_stroke.predict(source=np_img, conf=conf, verbose=False)

    if len(results[0].boxes) == 0:
        os.remove(tmp_path)
        return {"message": "⚠️ Aucun AVC détecté."}

    # Annoter l’image
    annotated_image = results[0].plot(labels=True)

    # Sauvegarder sortie image
    timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
    out_img_name = f"stroke_result_{timestamp}.png"
    out_img_path = os.path.join(OUTPUT_DIR, out_img_name)
    cv2.imwrite(out_img_path, annotated_image)

    # Sauvegarder rapport
    rapport_text = generate_report(results)
    out_txt_name = f"rapport_{timestamp}.txt"
    out_txt_path = os.path.join(OUTPUT_DIR, out_txt_name)
    with open(out_txt_path, "w", encoding="utf-8") as f:
        f.write(rapport_text)

    # Nettoyage input
    os.remove(tmp_path)

    return {
        "annotated_result_url": f"{BASE_URL}/files/{out_img_name}",
        "rapport_url": f"{BASE_URL}/files/{out_txt_name}",
        "message": "✅ Prédiction réussie avec rapport"
    }

# -----------------------------
# 4. Auto-cleanup toutes les 10 min
# -----------------------------
def auto_cleanup(interval_minutes=10):
    while True:
        time.sleep(interval_minutes * 60)
        for filename in os.listdir(OUTPUT_DIR):
            file_path = os.path.join(OUTPUT_DIR, filename)
            try:
                if os.path.isfile(file_path):
                    os.remove(file_path)
                    print(f"[CLEANUP] Fichier supprimé : {file_path}")
            except Exception as e:
                print(f"[CLEANUP] Erreur suppression {file_path} : {e}")

threading.Thread(target=auto_cleanup, args=(10,), daemon=True).start()