| --- |
| license: mit |
| datasets: |
| - masoudnickparvar/brain-tumor-mri-dataset |
| metrics: |
| - accuracy |
| pipeline_tag: image-classification |
| library_name: keras |
| tags: |
| - cnn |
| - keras |
| - brain-tumor |
| - medical-imaging |
| - tensor-flow |
| - image-classification |
| language: |
| - en |
| --- |
| |
| Brain Tumor Detection CNN Model |
|
|
| This model was trained using a Convolutional Neural Network (CNN) to classify brain MRI images as either having a tumor or not. It uses Keras with TensorFlow backend and was trained on the publicly available [Brain Tumor MRI Dataset](https://www.kaggle.com/datasets/masoudnickparvar/brain-tumor-mri-dataset) from Kaggle. |
| Dataset |
|
|
| The dataset contains 3,762 T1-weighted contrast-enhanced MRI images, labeled as: |
|
|
| - **Yes** – Images with a brain tumor |
| - **No** – Images without a brain tumor |
|
|
| The data is balanced and preprocessed into two folders: `yes/` and `no/`. |
|
|
| Train Accuracy: ~98% |
| Validation Accuracy: ~96% |
|
|
| ## 🧠 Model Architecture |
|
|
| - Type: CNN |
| - Framework: Keras (TensorFlow backend) |
| - Input shape: `(150, 150, 3)` |
| - Final Activation: `sigmoid` |
| - Loss: `binary_crossentropy` |
| - Optimizer: `Adam` |
|
|
| Code to Run: |
|
|
| ```python |
| import streamlit as st |
| import tensorflow as tf |
| import numpy as np |
| from tensorflow.keras.preprocessing import image |
| from PIL import Image |
| |
| # Load the trained model |
| @st.cache_resource |
| def load_model(): |
| return tf.keras.models.load_model('models/brain_tumor_model.h5') # Update path if needed |
| |
| model = load_model() |
| |
| # Define class labels |
| class_names = ['glioma_tumor', 'meningioma_tumor', 'no_tumor', 'pituitary_tumor'] |
| |
| # UI |
| st.title("🧠 Brain Tumor Detection from MRI") |
| st.write("Upload an MRI image to detect the type of brain tumor.") |
| |
| # Upload image |
| uploaded_file = st.file_uploader("Choose an MRI image", type=["jpg", "jpeg", "png"]) |
| |
| if uploaded_file is not None: |
| # Show image |
| img = Image.open(uploaded_file) |
| st.image(img, caption="🖼️ Uploaded Image", use_container_width=True) |
| |
| # Preprocessing |
| img = img.resize((224, 224)) # ✅ Make sure it matches your model's input size |
| img_array = image.img_to_array(img) |
| img_array = np.expand_dims(img_array, axis=0) / 255.0 |
| |
| # Prediction |
| predictions = model.predict(img_array) |
| confidence = float(np.max(predictions)) * 100 |
| predicted_class = class_names[np.argmax(predictions)] |
| |
| # Output |
| st.success(f"🎯 Predicted Tumor Type: **{predicted_class}**") |
| st.info(f"📊 Model Confidence: **{confidence:.2f}%**") |
| |
