An automated cervical cancer diagnosis model using Y-net and ensemble deep learning model

Cervical cancer (CC) is the fourth most common cancer globally, primarily affecting underdeveloped nations. On the other hand, prompt diagnosis might make clinical patient care easier. The issue is that there are less licenced and experienced health cytotechnicians than there are patients in need of diagnosis. To differentiate the cell, the majority of current techniques require accurate image segmentation. Traditional machine learning (ML) diagnostic systems function similarly to cytopathologists, who use manually constructed morphological parameters to assess a cell's malignancy, such as nucleus area, nucleus-cytoplasm perimeter ratio, etc. However, using an ensemble deep learning (DL) model in this work may allow us to eliminate the segmentation and feature selection processes, which are computationally intensive. Adaptive Intellect-Guided Filter (AIGF) was first created to reduce noise in images without adding more guiding images. Second, it predicts a discriminative map for finding significant regions in an image and segments various tissue types in CC images effectively. Finally, advises predicting CC using an ensemble DL model consisting with 2 levels. The initial level encompasses recurrent neural networks namely Modified Deep Maxout Network (MDMN) and Long Short-Term Memory Network (LSTM). The suggested E-LSTM-MDMN is fully connected neural networks at level two. Experimental results of this work show improved performances for values of DSC, JSI, sensitivity, specificity, accuracy, and F-Scores.