Brain Tumour Detection using Deep Learning

Published on 06 May 2021. Written by Dineshkumar E

In this machine learning project, we will use deep learning method to detect the brain tumours with the help of MRI (Magnetic Resonance Imaging) images of the brain.

Brain tumours are two types: malignant and benign. Most of the disease will reach the critical stage if not detected earlier. Timely detection of the disease will help a lot in the treatment process.

The major cause of brain tumours is because of the abnormal growth and uncontrolled cell division in the brain. Pituitary, meningioma and Glioma are some of the common types of tumours.

Read more..

SLNOTE
We will use the MR images of the brain to predict the pattern of the tumour with the help of machine learning techniques which makes the process less time consuming with the minimal amount of errors. Basically the MR images are taken in three different directions for brains: axial, coronal and sagittal. These three images are analysed by the model to detect the tumour.

Deep learning: Deep learning is a subset of machine learning. It gives the results with more accuracy sometimes it also exceeds the human-level performance. In deep learning, the models are get trained using a large set of data and neural network architectures to get the desired output with a high level of accuracy. Mostly the models perform the classification directly from text, sound and images.

Following are some of the fields where deep learning is utilized:

Automated driving - self-driving cars to detect the signboards

Aerospace and Defence - using satellites to detect safe zone for troops and locate the area of interest.

Industries - increases worker safety around heavy machinery by detecting the distance between the worker and machines.

Medical - to detect cancer cells

Home automation and AI devices - Google Home, Alexa, Apple home pad, etc. all these devices are powered by deep learning.

SLLATEST
Project Implementation

Feature extraction

If the input data of the algorithm is large, a smaller set of features are created. Features are created from the initial dataset in machine learning and image processing. Here the primary feature set is extracted from the subset this process is called feature extraction.

CNN - It includes several layers such as convolutional layers, input layer, output layer, normalization layers, pooling layers and fully connected layers.

Convolution layer - It is used to classify images after the feature extraction process

Sub-sampling layer - Here the operations are performed to decrease the size of the input image.

By using the dataset which is created with the help of MRI images of patients simulation is performed. CNN is used to detect the brain tumour through brain images to obtain more accurate results clustering algorithms is used for feature extraction.

Alexnet architect is used to distinguish and divide the images. Initially, the images were employed to CNN without any feature extraction methods.

The simulation is performed by the alexnet architect using 5 convolutional layers and 3 layers of normalization layers, sub-sampling layers, lastly layer and fully connected layer.

After the successful simulation, CNN succeeded to classify the images into tumour patient and normal patient with an accuracy of 98 %.

You can also predict the efficiency of the technique by using other classifiers such as RBF classifier, DT classifier, softmax fully connected layer classifier in the CNN architecture.

SLDYK
Kit required to develop Brain Tumour Detection using Deep Learning:

• No kit required

Technologies you will learn by working on Brain Tumour Detection using Deep Learning:

• Machine Learning
• Deep Learning
• Python

Any Questions?