CancerAI: A Deep Learning Framework for Ovarian Cancer Prediction

• Main Authors: Revathy, G., Maheswari, P. Uma, Madhavi, S., P, Vijay Anand, Begum, R. Vajubunnisa, Jasmin, H.
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; Convolution neural network; Convolutional neural networks; Deep learning; Histopathology; Observers; Ovarian cancer; Partitioning algorithms; Prediction; Prediction algorithms; Training; Transfer learning
• ISSN: 2996-5357
• DOI: 10.1109/ICESC60852.2024.10689884

Ovarian cancer is the fifth most common cause of death from cancer in women. This is mainly because it is often diagnosed at a late stage, as the earliest symptoms are unclear and inconsistent. Existing diagnostic techniques, such as biomarkers, biopsies, and imaging tests, have notable drawbacks such as reliance on subjective interpretation, inconsistency across different observers, and time-consuming testing procedures. This research study introduces an innovative deep learning framework that employs a convolutional neural network (CNN) algorithm to accurately predict and diagnose ovarian cancer, thereby overcoming the existing constraints. The CNN was trained using a dataset of histopathology images. The dataset was partitioned into test and training subsets to enhance the model's performance. The algorithm demonstrated an impressive accuracy rate of 96%, successfully detecting 97.12% of malignant instances and precisely categorizing 95.02% of healthy cells. This method effectively mitigates the challenges related to human expert evaluation, including elevated rates of misclassification and variability across different observers, while also reducing the time required for analysis. The results underscore the capability of this CNN-based technique to offer a more precise, effective, and dependable strategy for forecasting and detecting ovarian cancer. Subsequent investigations will prioritize the integration of new breakthroughs in deep learning to further amplify the efficacy of the suggested approach.