Classification of Cervix types using Convolution Neural Network (CNN)

The transformation zone is the region in the cervix where precancerous cells develop and later result in cervical cancer. Based on the position of this zone, the cervix is categorized into three classes: Type I, Type II and Type III. Distinguishing these types manually is very challenging for health practitioners. Therefore, it is important to automate the classification of the cervix types using Convolution Neural Network. Various research work has used common CNN architecture such as ResNet, Vgg, Inception which are time-consuming and require a large memory computation. However, this is not suitable for mobile deployment in low-resource regions. Thus, in this work, SqueezeNet a smaller CNN architecture is used and the accuracy obtained is equivalent to AlexNet (a much larger network). It requires a low computation memory and it is suitable for deployment on mobile devices. lt is believed that deploying a cervix type classification system on a mobile device will aid screening by health practitioners and reduce the misdiagnosis of patients.