https://ieeexplore.ieee.org/abstract/document/9432382
Cervical cancer is one of the most prevalent diseases in women ranking fourth in worldwide, mostly occurring in less-developed countries. This is perceived when certain vagaries occur in a woman's cervix. These cancer cells can also spread to other vital organs like lungs, liver and bladder which complicates the problem. Previous discoveries, tests and careful monitoring showed high levels of recovery rates at early detection of cancerous cells. But distinguishing cervical cells in Pap smear is demanding piece of work due to some constraints. Some of the constraint includes entanglement of the morphological changes in the structural parts of the cells. Although there are two methods to obtain the cells which are Colposcopy and pap-test but in reality, Pap-smear test are most favoured due to low cost and pain free diagnosis. Pap Smears are slides of cluster cells in which the cells vary in size, colour and morphology depending upon the degree of abnormality. Earlier studies have utilized both machine learning and deep learning methodologies, the prior one is not the most effective as it requires segmentation and obtaining hand crafted features which consumes crucial time. This paper presents deep learning classification methods applied on the SIPAKMED pap-smear image dataset to establish a reference point for the assessment of forthcoming classification techniques. With this approach highest classification accuracy of 94.89% was obtained with ResNet-152 architecture.