Thyroid Ultrasound Image Database and Marker Mask Inpainting Method for Research and Development

Abstract

Objective: The main objective of this study is to build a rich and high-quality thyroid ultrasound image database (TUD) for computer-aided diagnosis (CAD) systems to support accurate diagnosis and prognostic modeling of thyroid disorders. In addition, most of the raw thyroid ultrasound images contain artificial markers, which seriously affect the robustness of CAD systems due to their strong prior location information. Therefore, a Marker Mask Inpainting (MMI) method is proposed to erase artificial markers and improve image quality. Methods: First, a set of thyroid ultrasound images are collected from the General Hospital of the Northern Theater Command. Then, two modules are designed in MMI, namely Marker Detection Module (MD) and Marker Erasure Module (ME). The MD module detects all markers in the image and stores them into a binary mask. According to the binary mask, the ME module erases the markers and generates a unmarked image. Finally, we build a new thyroid ultrasound image database (TUD) based on the marked images and the unmarked images. TUD is carefully annotated and statistically analyzed by professional physicians to ensure accuracy and consistency. Moreover, several normal thyroid gland images and some ancillary information about benign and malignant nodules are provided. Results: Several typical segmentation models are evaluated on TUD. The experimental results show that our TUD can facilitate the development of more accurate CAD systems for the analysis of thyroid nodule-related lesions in ultrasound images. Meanwhile, the effectiveness of our MMI is demonstrated by quantitative experiments. Conclusion: A rich and high-quality resource TUD is built, which promotes the development of more effective diagnostic and treatment methods to thyroid diseases. Furthermore, MMI for erasing artificial markers and generating unmarked images is proposed to improve the quality of thyroid ultrasound images. Our TUD are available at https://github.com/NEU-LX/TUD-Datebase.