Refactor deployment workflow and clean up unused dependencies

.github/workflows/deploy.yml

@@ -6,7 +6,7 @@ on:
   workflow_dispatch:
     inputs:
       force_build:
-        description: 'Force rebuild all services (backend/rag/model)'
+        description: 'Force rebuild all services'
         required: false
         default: 'false'
 

client/src/features/dashboard/Help&Support/hooks.ts

@@ -6,8 +6,7 @@ export interface SupportContactPayload {
   message: string
 }
 
-// Send a support request to the TumorLens team (contact@tumorlens.health).
-// The sender's name and email are resolved server-side from the logged-in user.
+// Submit Message to Support
 export const useSubmitSupport = () =>
   useMutation({
     mutationFn: async (data: SupportContactPayload) => {

client/src/features/dashboard/layout.tsx

@@ -267,7 +267,6 @@ export const DashboardLayout = ({ children }: { children?: React.ReactNode } = {
     const currentPath = useRouterState({ select: (s) => s.location.pathname });
     const allNavItems = NAV.flatMap((s) => s.items);
 
-    // Exact match first, then prefix match for sub-routes (e.g. /results/PAT-00042)
     const activeNavItem =
         allNavItems.find((i) => i.id === currentPath) ??
         allNavItems.find((i) => i.id !== "/" && currentPath.startsWith(i.id + "/"));
@@ -292,6 +291,9 @@ export const DashboardLayout = ({ children }: { children?: React.ReactNode } = {
         return () => window.removeEventListener("keydown", handler);
     }, []);
 
+    // Org Member
+    const isOrgMember = user?.org?.role === "member";
+
     return (
         <div
             className="flex h-screen overflow-hidden"

scan_model/app.py

@@ -1,5 +1,5 @@
 import io
-from fastapi import FastAPI, File, UploadFile, HTTPException
+from fastapi import FastAPI, File, UploadFile
 from torchvision import transforms
 import torch
 from PIL import Image
@@ -11,66 +11,80 @@
 model = timm.create_model('efficientnet_b0', pretrained=False, num_classes=4)
 model.load_state_dict(torch.load("model.pth", map_location="cpu"))
 model.eval()
+
 CLASSES = ['glioma', 'meningioma', 'notumor', 'pituitary']
 transform = transforms.Compose([
     transforms.Resize((224, 224)),
     transforms.ToTensor(),
     transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
 ])
 
-# Prediction Endpoint
+
+def heuristic_mri_check(image: Image.Image) -> tuple[bool, str]:
+    """Reject obvious non-MRI input using cheap image-statistics checks."""
+    arr = np.asarray(image.resize((224, 224)), dtype=np.float32)
+    r, g, b = arr[..., 0], arr[..., 1], arr[..., 2]
+
+    # 1. Grayscale: MRI scans have R≈G≈B
+    chan_diff = (np.abs(r - g).mean() + np.abs(r - b).mean() + np.abs(g - b).mean()) / 3
+    if chan_diff > 12.0:
+        return False, "Image is colored; MRI scans are grayscale."
+
+    gray = arr.mean(axis=2)
+
+    # 2. Dark background: borders should be mostly black
+    border = np.concatenate([
+        gray[:20, :].ravel(), gray[-20:, :].ravel(),
+        gray[:, :20].ravel(), gray[:, -20:].ravel(),
+    ])
+    if (border < 30).mean() < 0.55:
+        return False, "Borders are not dark; no MRI background detected."
+
+    # 3. Central mass: center brighter than edges
+    h, w = gray.shape
+    center = gray[h // 4:3 * h // 4, w // 4:3 * w // 4].mean()
+    if center < 35 or center <= border.mean() + 10:
+        return False, "No central tissue mass detected."
+
+    # 4. Bimodal histogram: dark background peak + tissue peak
+    hist, _ = np.histogram(gray, bins=32, range=(0, 255))
+    if hist[:6].sum() == 0 or hist[6:].sum() == 0 or hist[:6].sum() / gray.size < 0.20:
+        return False, "Intensity distribution inconsistent with MRI."
+
+    return True, "Passed heuristic MRI checks."
+
+
 @app.post("/predict")
 async def predict(file: UploadFile = File(...)):
-
-    # Layer 1 — file type
-    if file.content_type not in ['image/jpeg', 'image/png']:
-        return {
-            "success": False,
-            "message": "Unsupported file type — only JPEG and PNG are allowed"
-        }
+    # Layer 1 — declared type (cheap filter, not trusted)
+    if file.content_type not in ('image/jpeg', 'image/png'):
+        return {"success": False, "message": "Unsupported file type — only JPEG and PNG are allowed"}
 
     contents = await file.read()
 
-    # Layer 2 — empty or corrupted
-    if len(contents) == 0:
-        return {
-            "success": False,
-            "message": "Empty file — no data to process"
-        }
-
+    # Layer 2 — empty / corrupted
+    if not contents:
+        return {"success": False, "message": "Empty file — no data to process"}
     try:
+        Image.open(io.BytesIO(contents)).verify()        # validate integrity
         image = Image.open(io.BytesIO(contents)).convert("RGB")
     except Exception:
-        return {
-            "success": False,
-            "message": "Corrupted image — unable to open"
-        }
-
-    # Check if image is grayscale (MRI scans are grayscale)
-    image_array = np.array(image)
-    r, g, b = image_array[:,:,0], image_array[:,:,1], image_array[:,:,2]
-    is_grayscale = np.mean(np.abs(r.astype(int) - g.astype(int))) < 10 and \
-                np.mean(np.abs(g.astype(int) - b.astype(int))) < 10
-
-    if not is_grayscale:
-        return {
-            "success": False,
-            "message": "Invalid image — expected a grayscale MRI scan"
-        }
+        return {"success": False, "message": "Corrupted image — unable to open"}
 
-    tensor = transform(image).unsqueeze(0)
+    # Layer 3 — classical heuristic pre-check
+    ok, reason = heuristic_mri_check(image)
+    if not ok:
+        return {"success": False, "message": f"Invalid MRI - {reason}"}
 
+    # Inference
+    tensor = transform(image).unsqueeze(0)
     with torch.no_grad():
-        outputs = model(tensor)
-        probs = torch.softmax(outputs, dim=1)[0]
+        probs = torch.softmax(model(tensor), dim=1)[0]
         confidence, predicted = torch.max(probs, 0)
 
-    # Layer 3 — confidence threshold
+    # Layer 4 — confidence threshold
     if confidence.item() < 0.70:
-        return {
-            "success": False,
-            "message": "Low confidence — not a valid MRI scan"
-        }
+        return {"success": False, "message": "Low confidence — prediction unreliable"}
 
     return {
         "success": True,
@@ -79,12 +93,7 @@ async def predict(file: UploadFile = File(...)):
         "all_scores": {c: round(probs[i].item(), 4) for i, c in enumerate(CLASSES)},
     }
 
-# Health Check Endpoint
+
 @app.get("/")
 def health():
-    return {
-        "status": "ok",
-        "message": "Scan model service is running."
-    }
-
-# uv run uvicorn app:app --reload
+    return {"status": "ok", "message": "Scan model service is running."}

server/package.json

@@ -39,7 +39,6 @@
     "bullmq": "^5.76.6",
     "class-transformer": "^0.5.1",
     "class-validator": "^0.15.1",
-    "cloudinary": "^2.9.0",
     "cookie-parser": "^1.4.7",
     "helmet": "^8.1.0",
     "otplib": "^13.4.0",

server/src/Organization/organization.service.ts

@@ -427,6 +427,4 @@ export class OrganizationService {
 
         return { success: true };
     }
-
-    //---------------- AUDIT LOGS ----------------//
 }