Multi-task Deep Learning Based CT Imaging Analysis For COVID-19: Classification and Segmentation

Amine Amyar; Romain Modzelewski; Su Ruan

Experiment 3: The third experiment was the comparison between our model and convolutional neural network trained to perform classification only. The CNN used is an 8 layer deep neural network with 6 convolutional layers, where each one is followed by a Maxpooling and a Dropout regularization of 25% to prevent the model from overfitting. The feature maps go from 8 to 256 by a factor of 2 between each two layers. We used 3 x 3 filter for convolution and 2 x 2 for Maxpooling. Then a Flatten followed by two Dense layers with 128 neurons and 1 neuron respectively. A Dropout of 50% is also applied to the first layer to reduce and prevent overfitting. The activation function is elu for all layers except the last one which is a sigmoid to generate a probability for each class COVID vs non-COVID. The loss function is the binary cross-entropy and the metric is the accuracy, with the Adam optimizer. The CNN was optimized in order to ensure a fair comparison with our proposed model. The model was trained for 1500 epochs with an early stopping of 35, in the same condition as our model.