Early Detection of Ovarian Cancer using Deep Learning

Abstract

Ovarian cancer is one of the deadliest gynecological malignancies, and early diagnosis through histopathological image analysis can significantly improve patient outcomes. This study proposes a convolutional neural network (CNN)-based framework for the classification of ovarian cancer using histopathological images. Greyscale conversion, normalization, and Contrast Limited Adaptive Histogram Equalization (CLAHE) were included in a multi-stage preprocessing pipeline to improve image quality. Furthermore, photometric data augmentation methods raise model generalization and adaptation capacity. An attention module included in the proposed CNN model lets the network concentrate on important areas of the image, thereby improving classification performance. Five well-known transfer learning models—MobileNet, ResNet50, VGG16, DenseNet 201, and VGG19—were assessed against the proposed method's efficacy. Moreover, k-fold cross-validation was used to guarantee the dependability and strength of the model over several data splits. Experimental results show that the attention-based CNN model beats the comparison models, therefore proving its potential as a strong tool for the automatic ovarian cancer classification in histological pictures.