Intelligent Medical Image Classifier by CNN

How to design a customized CNN architecture to classify HRCT lung image patches of ILD patterns?

In this project, we have designed a customized Convolutional Neural Networks (CNN) with two convolution layers and two max pool layers to classify lung image patches with lung disease. It was a part of medical image classification project in the research center. The customized CNN framework could efficiently learn the image features from lung image patches, which are the most suitable features for the classification purpose. The classification problem is implemented in PyThon using PyThorch library.

PyTorch is a Python-based library that provides functionalities such as TorchScript for creating serializable and optimizable models. Tensors in PyTorch are similar to NumPy’s n-dimensional arrays which can also be used with GPUs.

We can consider Convolutional Neural Networks as feature extractors that help to extract features from images. We use filters to extract features from the images and Pooling techniques to reduce the number of learnable parameters.

We use pooling layers in order to select the largest values on the feature maps. In theory, any type of operation can be done in pooling layers, but in practice, only max pooling is used because we want to find the outliers. Fully connected layers are placed before the classification output of a CNN and are used to flatten the results before classification.

We can play around with the hyperparameters of the CNN model and try to improve accuracy even further. Some of the hyperparameters to tune can be the number of convolutional layers, number of filters in each convolutional layer, number of epochs, number of dense layers, number of hidden units in each dense layer, etc.