Thyroid Cancer Classification using Transfer Learning

Abstract

The growth of artificial intelligence has been successfully reached remarkable findings in all fields including of medical cases. The powerful performance of artificial intelligence successfully became a second opinion to assist the medical personnel in making diagnosis decision. In the thyroid cancer cases, the growth of artificial intelligence offered valuable benefit since thyroid examination procedures highly depended on the medical personnel skills and experiences. One of the popular findings was that the implementation of transfer learning to find the best network for the targeted problem. In this study, we conducted experiment to develop a classification method for classifying thyroid cancer using transfer learning method. Our experiment was performed in the thyroid public dataset consisting of 348 thyroid ultrasound images divided into two classes (benign and malignant). In this study, we performed DenseNet121 and NasNetLarge as the best networks according to previous similar studies. According to our experiment, NasNetLarge obtained better accuracy than DenseNet121 by 8% of improvement. This result indicated that NasNetLarge was a powerful CNN for classifying the thyroid cancer.

Ähnliche Arbeiten