A Novel Meta Heuristic Approach with Optimal Deep Learning Neural Network Based Oral Cancer Detection Model
A Subbulakshmi, S. Nagarajan · Indian Journal of Science and Technology · 2023
Objectives: This study proposes a new approach to improve oral cancer detection in medical images by utilizing a Deep Convolutional Neural Network (DCNN) and an optimized Long Short-Term Memory (LSTM) technique. Methods: First, the input oral squamous cell carcinoma images are pre-processed using median filtering as well as CLAHE. Next, feature extraction is performed using the Local Tetra Pattern (LTrP) to extract different features. The HHHLO algorithm is then applied to select the optimal features for the subsequent feature selection process. Finally, the selected features are classified using a hybrid classifier called DCNN-LSTM, which predicts the diagnosis of patients with oral cancer. The investigation of the DCNN-LSTM model involves conducting experiments on a commonly used biomedical image dataset that is readily accessible through the Kaggle repository. Findings: The proposed method was implemented on the MATLAB platform, and its performance was evaluated using various metrics. The results demonstrated the superiority of the DCNN-LSTM model over existing methods, achieving a maximum accuracy of 0.975. Novelty: Oral cancer is a common and formidable type of cancer associated with a significant mortality rate. Keywords: Oral cancer, Improved Squirrel Search Algorithm (ISSA), Deep Convolutional Neural Network (DCNN), Contrast Limited Adaptive Histogram Equalization (CLAHE), Hybrid Horse herd lion optimization (HHHLO)