A Concise Review for Exploring Deep Learning's Potential in Cervical Cancer Prediction from Medical Images

Abstract

Cervical cancer originates in the cervix situated between the vagina and the bottom end of the uterus. It evolves gradually which begins with the appearance of aberrant cells in the cervical tissue. These aberrant cells might develop into cancer cells and migrate more into the cervix and adjacent tissues if they are not treated. Therefore, a patient's survival depends on rapid identification of cervical cancer. Various imaging modalities are widely used to identify cervical nodules as pre-cancer or cancer cells. But, limited results were determined and takes more time and needs many skilled radiologists. To solve this problem, many Deep Learning (DL) frameworks have emerged in these decades for automatic cervical cancer detection and categorization. These algorithms can detect suspicious nodules early, improving patient outcomes and aiding physicians in decision-making, thereby reducing fatality risk. This study provides an in-depth analysis of many DL frameworks developed to recognize and categorize cervical cancer from various imaging modalities. Initially, different cervical cancer categorization systems designed by many researchers based on DL algorithms are briefly examined. A comparison research is carried out to comprehend the shortcomings of those algorithms and recommend an alternative method for accurately classifying cervical cancer in order to regulate worldwide morality rates.