Chromosome Abnormality Detection Using Flying Fox Optimization based Deep Convolutional Neural Network

Chromosomes are thread-like structures located in the nucleus of human cells that carry the human genetic information.However, abnormalities in chromosomes are not accurately detected by existing approaches due to similarities between different chromosome classes.To solve this problem, a Flying Fox Optimization-based Deep Convolutional Network (FFO-DCNN) approach is proposed to enhance detection of abnormalities in chromosomes effectively.The exploration of FFO for cold trees is utilized to fine-tune the hyperparameters, which enhances the detection of abnormalities in chromosomes efficiently.Initially, chromosome images are acquired from the Biolmlab dataset, and the raw images are preprocessed to remove noise using a Weiner filter and augmented via a Generative Adversarial Network (GAN).A Deep Convolutional Neural Network (DCNN) model is then employed for feature extraction, followed by effective chromosomes detection.To reduce inaccurate detection results, Flying Fox Optimization is proposed as a hyperparameter tuning model for the softmax layer of the DCNN model.Experimental results of the proposed FFO-DCNN model for chromosome abnormality detection are evaluated using performance metrics, which show an accuracy of 99.41% and 99.13 % for the Biolmlab and hospital datasets, respectively.These results are superior to existing approaches such as the Hybrid CNN -Support Vector Machine (Hybrid CNN-SVM) approach.