Effective Method for Cancer Diagnosis Employing Deep Learning Algorithms

Abstract

Leading sources of death and morbidity, lung and colon cancers sometimes develop simultaneously and pose serious risks to people's health. Early discovery is essential for successful therapy, but the procedure for conducting a histopathological analysis may be laborious and complicated. Deep learning (DL) approaches have become important technologies that allow for more efficient examination of a greater number of patients. Previous studies have typically used individual DL models, which require more computation for feature extraction and analysis. However, based on several mild DL models, this study provides a modification framework for the early detection of lung and colorectal cancer. This innovative framework combines several reduction techniques that improve data representation. The reduced features generated by FHWT are fused to the three DL models using a discrete wavelet transform (DWT). In addition, all three DL models combine features of PCA. Finally, four different machine learning techniques are applied to feature reductions from PCA and FHWT-DWT algorithms, yielding an impressive accuracy rate of 99.6%.The results of these algorithms, based on DL on lighter models, shows that this is more effective than the previous methods In complex cases. This framework demonstrates the power of reductionist and adaptive approaches, resulting in better data interpretation and diagnostic methods.