ABSTRACT
Oral cancer is the most fatal disease among the overall malignant tumours worldwide. As more than 50% of the patients are diagnosed with advanced stages of the disease, the oral cancer prognosis is still bad. Early detection is one of the most crucial factors in lowering cancer fatality rates. But there is still a significant obstacle in the form of delayed diagnosis and developing a model for accurate oral cancer detection. To precisely detect oral cancer, deep learning models are used to reduce the number of deaths from cancer. In addition, the diagnosis system based on cloud-deep learning aids the telehealth services more probable. Therefore, this paper proposes a model, termed HMOCD (Hybrid Model based Oral Cancer Detection via Distributed Cloud Environment) with four stages: Pre-processing, segmentation, feature extraction, and detection. The input image is first pre-processed using the Weiner filter, and the output is then provided to CLAHE for enhancing the filtered image. Second, Modified Deep Joint segmentation is applied to the pre-processed image. Following the segmentation process, features for MTH, LGXP, and M-LBP are extracted. Ultimately, a hybrid classification method that integrates models such as Deep Maxout and enhanced Bi-LSTM is utilised to diagnose the condition.
KEYWORDS:
Nomenclature
Abbreviation | = | Description |
CNN | = | Convolutional Neural Network |
2d MFDFA | = | Two-Dimensional Multi-Fractal Detrended Fluctuation Analysis |
CHMBERT | = | Chinese Medical Bidirectional Encoder Representations from Transformer |
ELM | = | Extreme learning machine |
CLAHE | = | Contrast Limited Adaptive Histogram Equalization |
SVM | = | Support Vector Machine |
R-CNN | = | Regions with Convolutional Neural Networks |
IDL-OSCDC | = | Intelligent Deep Learning enabled Oral Squamous Cell Carcinoma Detection and Classification |
CIDRS | = | Cloud-based intelligent self-diagnosis and department recommendation service |
AEULM+EC | = | Aquila Exploration Updated with Local Movement with Ensemble Classifier |
DBN | = | Deep Belief Network |
DL | = | Deep Learning |
DNN | = | Deep Neural Networks |
ML | = | Machine Learning |
MTH | = | Multi Texton Histogram |
M-LBP | = | Modified significant Local Binary Pattern |
Bi-LSTM | = | Bi-directional Long Short Term Memory |
DT | = | Decision Tree |
LSTM | = | Long Short-Term Memory |
GRU | = | Gated Recurrent Unit |
RF | = | Random Forest |
ANN | = | Artificial Neural Networks |
Disclosure statement
No potential conflict of interest was reported by the author(s).