Abstract
Image processing is an essential component of the clinical workflow and is used at several stages to improve the diagnostic process and facilitate treatment planning. It can be broadly defined as the application of transformations to digital images, typically with the aim of enhancing the image or extracting relevant information from it. Common examples of medical image processing techniques are (1) image segmentation, i.e., the partitioning of images into several distinct regions; (2) image enhancement, e.g., noise reduction, sharpness enhancement, and artifact correction; (3) tomographic reconstruction, e.g., in computed tomography or magnetic resonance imaging; and (4) registration, i.e., spatial matching of multiple images. Recent innovations in the field of artificial intelligence, especially deep learning (DL), have resulted in a wide variety of potential applications in medical image processing. DL models, mostly trained using supervised learning, can offer several benefits over alternative (semi-)automated or manual approaches, owing to their accuracy, robustness, and speed. This chapter will provide an overview of neural networks used in image-to-image processing and will describe principles and applications of DL and unsupervised clustering in image segmentation. The next chapter will focus on DL applications in image enhancement, reconstruction, and registration.
| Original language | English |
|---|---|
| Title of host publication | Artificial Intelligence in Dentistry |
| Place of publication | Cham |
| Publisher | Springer |
| Publication date | Jan 2024 |
| Pages | 283-316 |
| Chapter | 17 |
| ISBN (Print) | 978-3-031-43826-4, 978-3-031-43829-5 |
| ISBN (Electronic) | 978-3-031-43827-1 |
| DOIs | |
| Publication status | Published - Jan 2024 |