A Cervical Cancer Detection System Using Genetic Algorithm Enhanced Context-Aware Feature Selection

Cervical cancer is the fourth highest cause of death among women, but in its early stages, it shows no symptoms. A lack of effective early diagnosis is the fundamental cause of the disease's prevalence and poses the greatest obstacle to researchers. Pap smears, human papillomavirus tests, colposcopy, and biopsies are all invasive procedures that necessitate the assistance of medical professionals; they are time-consuming and do not identify cancer until a later stage. Thus, the absence of early symptoms and lack of invasive diagnostics, highlight the need for utilizing machine learning and deep learning approaches to revolutionize cervical cancer detection. To address this, we propose GCFS-CC(Genetic Algorithm based Context-aware Feature Selection for Cervical Cancer), which is a framework to detect cervical cancer from Pap smear images of cervical cells and classify these using a deep learning method. In our work, we use a genetic algorithm which is a bio-inspired algorithm, to select the best features from the acquired images. For feature selection, we incorporate the context-aware entropy gain method. We evaluated our model on one data set which combined a total of 917 normal and cancerous images, divided into seven classes. Using the selected features we build a deep learning model to validate our proposed framework. The results showed that after feature selection, the performance metrics(accuracy, precision, and recall), all improved. These results pave the way for GCFS-CC to be a valuable tool for early cervical cancer detection. This, however, comes with certain challenges, availability of high-quality images and proper parameter tuning for the genetic algorithm required for feature selection. To successfully implement machine learning-based cervical cancer detection into clinical practice, it will be necessary to address these challenges. Furthermore, any ethical and privacy concerns associated with the use of machine learning algorithms in healthcare settings must be addressed.