Canny Edge and Deep Belief Neural Network-based Mouth Cancer Diagnosis

Mouth cancer is a global health concern thatis faced by almost all countries. Every day lakhs of mouthcancer patients are detected around the world, and it is very challenging to manage such types of issues indeveloping countries where the doctor and patient ratio is very low and do not have a large amount of cancerhospitals to treat all these mouth cancer patients. Mouth cancer requires early and accurate detection and diagnosisof patients to get improved outcomes. Traditional diagnostic methods are effective in diagnosing mouth cancerbut are subjective to the doctors and it leads to variability in the diagnosis of mouth cancer. Deep learning hasachieved great success in medical image-based cancer diagnosis. It shows great results in the diagnosis anddetection of not only mouth cancer but other types of cancer too. There are various types of image processingtechniques i.e. image classification, segmentation, and synthesis are used for the detection and diagnosis of variousclasses of cancer. This research proposes a novel approach that integrates Canny edge detection with a Deep BeliefNeural Network (DBN) for the automated diagnosis of mouth cancer from medical images. Canny edge detection,known for its effectiveness in highlighting the boundaries of structures within an image, is employed to enhancethe critical features relevant to cancerous tissues. These edge-enhanced images are then used as input to a DBN,a deep-learning model capable of extracting complex hierarchical features and performing robust classificationtasks. The proposed system was trained and tested on a dataset of mouth cancer images, demonstrating significantimprovements in detection accuracy, precision, and recall compared to traditional image processing techniquesand other deep learning models. The results were declared with around 73.63 % accuracy. Canny edge detectionwith DBN not only enhances feature extraction but also provides a more accurate tool for early mouth cancerdiagnosis, potentially aiding in clinical decision-making and improving patient prognosis.