Paper Title
Advancing Serous Carcinoma Diagnosis With Serinet-CNN: A Convolution Neural Network Approach

Abstract
Our research focuses on the categorization of serous carcinoma subtypes, particularly high-grade and low-grade serous carcinoma, which are ovarian cancers that start in the fallopian tube. To distinguish between these subtypes, we created a bespoke Convolutional Neural Network, Serinet-CNN, trained it from scratch and evaluated its performance against a number of transfer learning pre-trained models, such as MobileNet, InceptionV3, Resnet50 and VGG16. Our experimentation show that the bespoke CNN performs better regarding classification accuracy than these pre-trained models. With accuracy, precision, recall and F1-score of 0.98, 0.97, 0.98 and 0.98, respectively, our model's improved performance and interpretability point to its potential for accurate and dependable serous carcinoma subtype classification, which could enable a more precise diagnosis and treatment planning. We also carried out a thorough examination with kernel filters and activation maps, which shed light on the inner workings of the model explaining which part is crucial for the categorization of the image as a certain type of serous carcinoma. Keywords - Serous carcinoma, Convolution Neural Network, Transfer Learning, Attention Maps.