Automatic Breast Cancer Detection Using Inception V3 in Thermography
Mohammed Abdulla Salim Al Husaini, Mohamed Hadi Habaebi, Teddy Surya Gunawan, Md. Rafiqul Islam, Shihab A. Hameed
Abstract
Thermography is a non-invasive, passive imaging technique that is widely used in the medical profession, particularly for breast tumor diagnosis. This research paper suggests the examination of breast thermography through the utilize of Deep convolutional neural network inception v3 including training several times and fine-tuning learning rate. It is hoped that such strategies will enable the development of highly accurate thermography-based diagnostic methods. The proposed technique is training inception v3: Continuous training, training after shutting down MATLAB and after shutting down computer, and the average accuracy of the classification was obtained from epoch 3, 5 and 6, which are 98.104%, 98.712 and 97.816%, respectively. The use best value of learning rate allows the correlation between the variables accuracy, and this is critical because it aids in the selection of the proper variables to be applied in the Deep convolutional neural network's construction. The proposed learning rate has achieved highest accuracy between 1e-3 and 2.5e-3 are 97.816% and 99.928% respectively in identifying breast cancer.