Abstract
Thyroid nodule lesions are one of the most common lesions of the thyroid; the incidence rate has been the highest in the past thirty years. X-ray computed tomography (CT) plays an increasingly important role in the diagnosis of thyroid diseases. Nonetheless, as a result of the artifact and high complexity of thyroid CT image, the traditional machine learning method cannot be applied to CT image processing. In this paper, an end-to-end thyroid nodule automatic recognition and classification system is designed based on CNN. An improved Eff-Unet segmentation network is used to segment thyroid nodules as ROI. The image processing algorithm optimizes the ROI region and divides the nodules. A low-level and high-level feature fusion classification network CNN-F is proposed to classify the benign and malignant nodules. After each module is connected in series with the algorithm, the automatic classification of each nodule can be realized. Experimental results demonstrate that the proposed end-to-end thyroid nodule automatic recognition and classification system has excellent performance in diagnosing thyroid diseases. In the test set, the segmentation IOU reaches 0.855, and the classification output accuracy reaches 85.92%.
Highlights
In order to better solve this problem, many scholars have done a lot of research using computer-aided diagnosis technology
A single multiclass segmentation network cannot achieve high accuracy and the classification effect of benign and malignant is poor. erefore, in this research, a CNN-based semantic segmentation network plus an image classification network is designed, and image processing algorithms are used to make the two connected in a reasonable manner, so that the segmentation network focuses on the segmentation of the thyroid nodules and the background. e classification network focuses on extracting image features of thyroid nodules to achieve benign and malignant classification
We designed a set of end-to-end automatic recognition and classification system for thyroid nodules, which can realize the benign and malignant classification of nodules without any marking on computed tomography (CT) images and can achieve accurate diagnosis of a single nodule, meeting the need of multinodule CT images
Summary
In order to better solve this problem, many scholars have done a lot of research using computer-aided diagnosis technology. In the study of thyroid nodules based on CT, Peng used the first-order texture features in nonenhanced CT images and support vector machine analysis to classify the Computational Intelligence and Neuroscience nodules, and the accuracy reached 88% [7]. The above researches have reached a high level in the study of thyroid classification, they require the prerequisite for the nodule information to be labeled, and it is impossible to achieve high-efficiency data processing. On this basis, Zhao et al designed an improved Unet architecture, Dense-Unet, to achieve the region of interest (ROI) segmentation of thyroid nodules and used multidimensional input fusion CNN model to achieve the classification of benign and malignant nodules. For multinodule CT images, nodules are segmented to achieve separate classification of each nodule
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call