Brain Tumor Classification and Segmentation

• 

  GUI for image classification and segmentation

Inspiration

Recognition of automated brain tumors in Magnetic resonance imaging (MRI) is a difficult task due to the complexity of its size and location variability.

In this research statistical analysis of morphological and thresholding techniques are proposed to process the images obtained by MRI for Tumor Detection from Brain MRI Images.

What it does

It takes a huge amount of human effort, however, it's not much efficient as the output predicted by the deep learning model.

How we built it

We have implemented brain tumor classification and segmentation. Classification will classify the type of tumor as a pituitary tumor, glioma tumor, and meningioma tumor. We are using the EfficientNet model for this classification. Segmentation will identify the part of the brain which is affected. We are using the U-net model for Tumor segmentation.

Challenges we ran into

We were stuck at so many parts for selecting the best model and parameters for the model. We also have to put a lot of effort to implement the GUI part, however, we can't deploy it properly during this time of the hackathon, but we will try to complete it soon.

Accomplishments that we're proud of

We explored a lot of things during the implementation of the code and learned a lot about the segmentation and masking of images.

What we learned

We learned how to use a pre-trained model using transfer learning, so many different models using TensorFlow Keras, and so more.

What's next for Untitled

We will try to modify our trained model to make it more accurate and fast. Also, make a responsive GUI such that anyone can access this model and get the output for their brain tumor images.

Built With

• efficientnet-b0
• kaggle
• python
• streamlit
• tensorflow
• u-net