Classification Of Benign And Malignant Thyroid Nodules In Ultrasound Images

In the world,cancer is the second leading cause of the greatest number of deaths after the heart disease.Image processing methods are being widely used for medical purposes.Recently,for improving the-initial-detection-and the treatment stages,especially in different cancer’s nodules including brain cancer,lung cancer and breast cancer etc.The common malignancy in endocrine system of human body is Thyroid cancer.Ultrasound images are generally monochrome,and they have a lower resolution.The regions,which show the cancer in ultrasound images are usually blurred,vague in margin and are irregular in shape.Moreover,the structures of cancer region are very similar to benign or normal tissues.The medical image processing assists the radiologists to diagnose a problem.In that way,it saves the time to detect the cancer nodules.The vital problem in medical science is the classification of the thyroid which is directly related to the health condition of human body.By nature,mostly the thyroid nodules are benign in which only less than 5% of these are malignant.Majority of the existing techniques have some limitations,because these are examined and evaluated under limited dataset,also they have not well-thought-out for external environmental changes.Thus,for satisfactory classification of thyroids Convolutional Neural Network(CNN)are applied to the raining ultrasound images.For the better accuracy of the detection,CNN VGG-16 model can extract the features in which the cancer regions are existing.In this study,VGG 16 is used for the classification of thyroid nodules.We evaluated our model in complex dataset,which mostly meet all real-world challenges to make our model robust.The main purpose of this model is to develop correct,a fast and reliable classification system for thyroid cancer,that would be very useful in medical field.