Automated Osteosarcoma Detection and Classification Using Advanced Deep Learning with Remora Optimization Algorithm

Osteosarcoma is considered a primary malignant bone tumor which affects young people and adults. Manual osteosarcoma identification is time consuming and necessitates expert knowledge. Since earlier detection of osteosarcoma reduces death rate, computer aided diagnosis (CAD) models can be developed to examine the medical images for decision making. The latest breakthroughs in machine learning (ML) and deep learning (DL) methods find to be useful for enhancing detection performance and diagnostic time. In this view, this article introduces an Automated Osteosarcoma Detection and Classification using metaheuristics with Advanced Deep Learning (AODC-MADL) algorithm. The proposed AODC-MADL model makes use of recent DL models with hyperparameter optimization algorithms to detect and classify osteosarcoma. For achieving that, the presented AODC-MADL algorithm pre-processes the input images to optimize the image quality. Moreover, the Dense-EfficientNet architecture is utilized to form a set of feature vectors. Besides, attention based bidirectional recurrent neural network (ABiRNN) model receives the feature vectors and performs classification process. At last, the remora optimization algorithm (ROA) is used to optimally choose the hyperparameters related to the ABiRNN model. A series of experiments have been conducted to depict the outstanding performance of the AODC-MADL algorithm. The experimental outcome emphasized that the AODC-MADL algorithm has obtained higher performance over other approaches.