Feat: Add MRI upload validation and safer preprocessing handling

[Sujan075]

closes#14 — Add MRI Upload Validation & Better Error Handling for Invalid Medical Images

Overview

This PR improves the safety and stability of the MRI upload and preprocessing pipeline by introducing layered validation, safer preprocessing checks, and standardized API error handling.

The goal was to prevent invalid or corrupted uploads from reaching inference while keeping the Flask API stable and user-friendly during failures.

---

What Changed

Added validators.py

Created a dedicated validation module to centralize upload and preprocessing validation logic.

Implemented:

• ValidationError
• validate_file_present()
• validate_file_extension()
• validate_file_size()
• validate_mime_type()
• validate_image_loadable()
• validate_tensor_shape()

Validation now combines:

• extension allowlist checks
• magic-byte/MIME validation
• PIL verification
• OpenCV load verification

This prevents invalid or corrupted files from reaching preprocessing or inference.

---

Hardened /api/predict pipeline in app.py

Reworked the upload flow into a sequential validation pipeline:

1. file presence validation
2. extension validation
3. upload size validation
4. MIME/magic-byte validation
5. secure file save
6. image integrity verification
7. preprocessing safety checks
8. inference safety handling
9. tensor shape validation

All failure paths now return structured JSON responses instead of silent failures or raw exceptions.

---

Improved Error Handling & Logging

• Replaced bare except: blocks with explicit exception handling and logging
• Removed raw exception leakage from API responses
• Standardized API error envelopes across handlers
• Updated 404/413/500 handlers for consistency

Example response format:

{
  "error": true,
  "code": "INVALID_MIME",
  "message": "File content does not match a valid MRI image format"
}

---

Added Validation Tests

Expanded test_tumor_detection.py with validation-focused unit tests covering:

• invalid extensions
• wrong MIME types
• corrupted images
• oversized uploads
• empty uploads
• invalid tensor shapes
• valid PNG/TIFF uploads

All tests pass successfully.

---

Environment Configuration

Added configurable upload limit support via:

MAX_UPLOAD_MB=16

Flask upload limits and validator limits now stay synchronized.

---

Testing

Verified with:

• pytest
• manual upload testing
• corrupted file testing
• renamed .txt -> .jpg validation
• oversized upload testing
• valid image inference flow

Result:

• 25 tests passed successfully

---

Outcome

This PR improves:

• MRI upload safety
• preprocessing stability
• API reliability
• validation clarity
• protection against corrupted uploads
• consistency of error responses

while keeping the implementation scoped specifically to the inference upload pipeline.

[Sujan075]

Hi @Brijeshthummar02

I’ve completed the implementation for Issue #14 — “Add MRI Upload Validation & Better Error Handling for Invalid Medical Images”.

Implemented Improvements

• Added layered MRI upload validation
• Added strict extension + MIME/magic-byte verification
• Added configurable upload size validation
• Added corrupted image detection using PIL + OpenCV checks
• Hardened /api/predict preprocessing and inference flow
• Standardized JSON API error responses
• Replaced silent exception handling with structured logging
• Added validation-focused unit tests for invalid/corrupted uploads

Validation & Testing

Tested scenarios include:

• invalid file extensions
• wrong MIME types
• corrupted images
• oversized uploads
• empty uploads
• valid PNG/TIFF uploads

All tests are passing successfully.

This PR keeps the changes scoped specifically to upload validation and inference stability without modifying unrelated training/model logic.

Looking forward to your review and feedback. Thanks!

[Brijeshthummar02]

@Sujan075 show me before / after test logs and changes.

[Sujan075]

Image 1:

Image 2:

Image 3:

Hey @Brijeshthummar02,
Ive shared the images you shall go through them and review it to me. Looking forward for the review. Thank you.

[Brijeshthummar02]

@Sujan075 i am still not able to figure out entire sequential validation pipeline, explain me in clean sketch how you did it with what was there earlier. also try to clean up ai code a bit. it doesn't show multiple implementations you have mentioned.

[Sujan075]

Hi @Brijeshthummar02,

I cleaned up and simplified the validation flow to make the sequential pipeline clearer and remove mixed validation approaches.

Earlier Flow

The previous implementation had:

• validation logic partially inside app.py
• duplicate image loading paths
• older validation remnants mixed with the new validator flow
• repeated OpenCV/PIL handling in multiple places
• overly verbose comments/documentation

This made the upload pipeline harder to follow during review.

---

Refactored Sequential Validation Flow

Current flow is now:

1. Validate file presence
2. Validate file extension
3. Validate file size
4. Validate MIME type
5. Validate image integrity (stream validation)
6. Save file securely
7. Validate saved image integrity
8. Load image + generate preview
9. Preprocess + validate tensor shape
10. Run inference

---

Cleanup Done

• centralized validation flow through validators.py
• removed duplicate OpenCV/PIL loading logic
• removed legacy validation remnants
• simplified comments/documentation
• reduced verbose/AI-looking code structure
• preserved existing API responses and behavior

---

Validation Status

• all 25 unit tests still passing
• all integration validation scenarios still passing
• HTTP status codes and JSON responses unchanged

The goal of this cleanup was to make the pipeline easier to review and maintain without changing functionality.
Thanks for the review feedback.

[Brijeshthummar02]

have you added this to our existing dashboard on main web?

[Sujan075]

Hii @Brijeshthummar02,

Yess,
The validation pipeline was integrated directly into the existing "/api/predict" flow used by the main web dashboard upload flow.

So the current dashboard/frontend now uses:

• file presence validation
• extension validation
• MIME validation
• corrupted image detection
• upload size validation
• structured JSON error handling

The changes were not implemented as isolated utility checks — they are part of the active inference upload pipeline used by the existing web interface.

I also verified this through runtime testing from the actual prediction flow while validating the integration scenarios.

[Brijeshthummar02]

resolve conflict

[Sujan075]

Hi @Brijeshthummar02

I've addressed the review feedback:

• cleaned up and simplified the validation pipeline
• centralized validation flow for better readability
• removed duplicate/legacy validation paths
• explained the sequential validation flow in the previous comment
• resolved the merge conflict and synced the branch with the latest master changes

Please let me know if there are any other changes you'd like me to make. Thanks!