Brain Tumor Detection Using YOLOv8n

Overview

This project focuses on the detection of brain tumors from MRI images using the YOLOv8n (You Only Look Once) object detection model. The aim is to provide a reliable and efficient method for identifying tumors in MRI scans, aiding early diagnosis and treatment. The model was trained on a dataset of 19,000 images and achieves high performance in precision and accuracy. An interactive interface allows users to upload images for real-time tumor detection.

🚨 Project Video Demo 🚨

---

🎥 Watch the full video walkthrough of the project here. 🎬

---

Project Motivation

This project was undertaken as a part of a Drone Lab assignment at the Scaler School of Technology. The lab emphasizes innovative solutions, and this challenge presented an opportunity to apply deep learning to a significant real-world problem: brain tumor detection in MRI scans.

Brain tumors can be life-threatening if not detected early. Manual analysis of MRI images by radiologists is time-consuming and prone to human error. By utilizing deep learning, we can assist in speeding up the detection process while improving accuracy. This project represents both a technical challenge and an opportunity to contribute to healthcare advancements, which aligns with the Drone Lab’s focus on innovation and impact.

Data Source

The dataset used in this project was sourced from Kaggle, containing 19,000 MRI brain images annotated for tumor detection. The dataset was structured to work with the YOLOv8n model.

• Image Format: JPEG
• Annotations: Provided in YOLO format (bounding boxes for tumor regions)
• Dataset Split:
  • Training Set: Used to train the model.
  • Validation Set: Used during training to fine-tune the model.
  • Test Set: Used for final evaluation and performance metrics.

Environment Setup

Step 1: Python Installation

The project was built using Python 3.x. Make sure Python is installed on your system.

Step 2: Required Libraries

Install the following Python libraries to manage the dataset, train the model, visualize results, and create the user interface:

pip install ultralytics
pip install opencv-python
pip install matplotlib
pip install pandas
pip install ipywidgets

• Ultralytics: For YOLOv8 model training and inference.
• OpenCV: For image processing and displaying results.
• Matplotlib: For plotting graphs and visualizing model performance.
• Pandas: To structure data and generate tables.
• Ipywidgets: For creating interactive widgets like file upload buttons.

Step 3: Setting Up Jupyter Notebook

To train the model and visualize the results, Jupyter Notebook was used. Install it as follows:

pip install notebook

Once installed, launch Jupyter Notebook to train the model and monitor progress.

Model Training

Why YOLOv8n?

The YOLOv8n model was selected for its speed and accuracy, offering an efficient real-time object detection solution, especially suitable for identifying brain tumors in MRI images.

Training Process

The model was trained on 19,000 MRI images using an NVIDIA RTX 4050 GPU to speed up the process. After many hours of training, the model was optimized and achieved excellent results in tumor detection.

Model Performance Evaluation

After training, the model was evaluated on a custom test set. The following metrics were recorded:

• Accuracy: 0.9202
• Precision: 1.0
• Recall: 0.9202
• F1-Score: 0.9584

Performance Visualization

To visualize the model’s performance, a table and bar chart were created to display the evaluation metrics.

Table of Metrics:

Metric	Value
Accuracy	0.9202
Precision	1.0
Recall	0.9202
F1-Score	0.9584

Bar Chart:

import matplotlib.pyplot as plt

metrics = ['Accuracy', 'Precision', 'Recall', 'F1-Score']
values = [0.9202, 1.0, 0.9202, 0.9584]

plt.bar(metrics, values, color=['blue', 'green', 'red', 'purple'])
plt.xlabel('Metrics')
plt.ylabel('Values')
plt.title('YOLOv8n Model Performance')
plt.show()

This chart visually demonstrates the model's high accuracy and precision.

User Interface for Tumor Detection

A user-friendly upload button was created, allowing users to upload their own MRI images. The YOLOv8n model processes the images in real-time, detects any tumors present, and draws a bounding box around the detected tumor. The confidence score is displayed above the bounding box.

Upload Functionality:

import cv2
import matplotlib.pyplot as plt
from ultralytics import YOLO
import numpy as np
from tkinter import Tk, filedialog, Button, Label

# Load the pre-trained YOLO model globally
model = YOLO(r'C:\Users\Mohit\BrainTumor_NewYolov8n\runs\detect\train\weights\best.pt')  # Ensure the correct path to your model

# Function to perform inference on the uploaded image
def detect_tumor(image_path):
    print(f"Processing the image: {image_path}")
    
    # Read image using OpenCV
    image = cv2.imread(image_path)
    
    # Perform inference using YOLOv8 model
    results = model(image)
    print("Inference completed.")

    # Get detections (bounding boxes and confidence)
    detections = results[0].boxes.xyxy.cpu().numpy()  # Bounding boxes
    confidences = results[0].boxes.conf.cpu().numpy()  # Confidence scores

    # Check for detections
    if len(detections) == 0:
        print("No tumor detected.")
        status_label.config(text="No tumor detected.")
        return

    # Draw bounding boxes and display confidence
    print("Tumor(s) detected. Drawing bounding boxes...")
    for i in range(len(detections)):
        box = detections[i]
        confidence = confidences[i]
        if confidence >= 0.25:  # You can adjust this threshold if needed
            x1, y1, x2, y2 = map(int, box)  # Convert to integer coordinates
            cv2.rectangle(image, (x1, y1), (x2, y2), (0, 255, 0), 2)  # Draw bounding box
            label = f"Conf: {confidence:.2f}"
            cv2.putText(image, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 87, 51), 1)  # Display confidence

    # Convert image from BGR to RGB for displaying
    detected_image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

    # Display the image with bounding boxes and confidence
    plt.figure(figsize=(8, 8))
    plt.imshow(detected_image_rgb)
    plt.axis('off')
    plt.title("Detected Tumor(s)")
    plt.show()
    
    print("Image processed successfully!")
    status_label.config(text="Tumor detection completed.")

# Function to select an image using a file dialog
def upload_image():
    # Open a file dialog and return the selected file path
    file_path = filedialog.askopenfilename(title="Select an Image", filetypes=[("Image Files", "*.jpg;*.jpeg;*.png")])
    if file_path:
        print("Image selected for processing.")
        status_label.config(text="Image selected for processing.")
        detect_tumor(file_path)
    else:
        print("No image selected.")
        status_label.config(text="No image selected.")

# Initialize the main Tkinter window
root = Tk()
root.title("Brain Tumor Detection")
root.geometry("300x150")  # Set the size of the window

# Create an upload button
upload_button = Button(root, text="Upload Image", command=upload_image, width=20, height=2)
upload_button.pack(pady=20)

# Create a status label
status_label = Label(root, text="Upload an image to detect brain tumors.", wraplength=250)
status_label.pack(pady=10)

# Run the Tkinter event loop
root.mainloop()

How It Works:

• Upload Image: Users can upload MRI images through the widget.
• Real-Time Detection: The model processes the image, detects tumors, and draws bounding boxes around them.
• Confidence Score: The confidence score for detection is displayed above the bounding box.

Conclusion

The YOLOv8n model provides an effective solution for detecting brain tumors in MRI scans. With 92% accuracy and 100% precision, it demonstrates the potential to assist radiologists by quickly identifying tumor regions. The user-friendly interface further adds value, allowing for real-time tumor detection.

Future Improvements:

• Model Fine-Tuning: Training the model with larger datasets could improve its performance even further.
• Multi-Abnormality Detection: Expanding the model to detect other brain conditions could make it a comprehensive diagnostic tool.
• Web Application: Future steps may include deploying this project as a web application to increase accessibility for medical professionals.