Integrated Design of Artificial Neural Network with Bald Eagle Search Optimization for Osteosarcoma Classification

Osteosarcoma becomes a most leading factor of bone cancer worldwide and results in high death rates. Initial diagnosis might raise survival rate but the process seems to be time-taking (complexity and reliability indulged for extracting the hand-crafted features) and majorly relies on the experience of pathologists. This work mainly concentrates in the enhancement of diagnosis and recognition of osteosarcoma by making use of computer aided diagnosis (CAD) and computer aided detection (CADx). These tools as convolutional neural network (CNN) could reduce the workload of physicians and offer superior prognosis for patient conditions. CNNs must be well-trained on an enormous volume of data for attaining a more promising performance. The study emphasis on the design of bald eagle search optimization with deep transfer learning (DTL) model for osteosarcoma classification (BESODL-OC) algorithm. The projected BESODL-OC methodology scrutinizes the histopathological images (HSI) for the identification of osteosarcoma. To achieve this, the BESODLOC algorithm exploits Gabor filtering (GF) model for noise removal process. Besides, a deep convolutional neural network (DCNN) based Inceptionv3 methodology is exploited for feature extraction process. Since trial and error based hyperparameter tuning process will be a tedious process, the BESO algorithm is utilized in this study. At last, artificial neural network (ANN) method can be exploited for classifying purposes. A wide range of experiments were made to exhibit the improvised outcome of the BESODL-OC system. Wide-ranging comparison analysis demonstrated the enhancements of the BESODL-OC system over other recent approaches.