A deep learning model based on the ResNet152 architecture that detects pneumonia from CT scans of lungs. The dataset used for training this model can be found here.
The model was trained for 100 epochs, with early stopping set to kick in at 10, with a batch size of 32. The model was trained on RTX 3080 Ti Laptop GPU.
Figure 1: A raw image sample of a normal lung |
Figure 2: A masked image sample of a normal lung |
Figure 3: A raw image sample of a pneumonia infected lung |
Figure 4: A masked image sample of pneumonia infected lung |
The model uses a base layer of ResNet152.
Figure 5: PneumoGuard Model Architecture
Residual Networks, or ResNets, learn residual functions with reference to the layer inputs, instead of learning unreferenced functions. Instead of hoping each few stacked layers directly fit a desired underlying mapping, residual nets let these layers fit a residual mapping. They stack residual blocks ontop of each other to form network. ResNet-152, short for "Residual Network 152," is a deep convolutional neural network architecture that belongs to the ResNet family. ResNet-152 is specifically known for its depth, consisting of 152 layers, making it a relatively deep neural network.
PneumoGuard achieved a training accuracy of 95.23%, training loss of 15.59%, validation accuracy of 89.74%, validation loss of 25.33%, testing accuracy of 87.34%, and testing loss of 38.14%. The accuracy can be further improved by training the model for more epochs and modifying the model architecture.
Figure 6: PneumoGuard Accuracy Plot |
Figure 7: PneumoGuard Loss Plot |