Understanding And Mitigating Noise In Orthodontic Imaging
Noise refers to unclear, grainy areas present in a radiograph that hides important details. X-rays can prove to be an aid in the diagnosis of multiple deformities. However, exposure to high doses or persistent exposure can lead to radiation related infirmities. Therefore, technicians and clinicians try to obtain the best radiographs with minimal radiation dose. Repeating X-rays is generally discouraged and only done when the resultant image fails to provide required anatomical details.
Repeating X-rays due to noise is a thing of the past. AI is capable of identifying regions of noise and mitigating it. The reproduction of details by artificial intelligence algorithms has significantly improved the field of orthodontic imagery.
How AI Understands Noise In Orthodontic Imaging?
Machine Learning Of Important Points
By referring to the buzz word “Artificial intelligence” we are actually explaining the combined role of machine learning and artificial intelligence algorithms. AI’s learning process is what makes it capable of understanding noise in radiographic images. There are several ways in which AI makes sense of the provided data.
Supervised learning utilizes segmentation of anatomical structures (provided by medical experts). In this method, seasoned clinicians segment anatomical landmarks on radiographs and feed the data to AI learning. However, this method is costly and dependent on the expertise of specialized consultants.
On the other hand, unsupervised learning requires a larger amount of data pouring into the system. A big dataset (extracted from different patients) is fed into the algorithm. The intelligent system learns the shape, design and position of anatomical landmarks on orthodontic images. A further sub-class of machine learning is deep learning that enables more robust identification of imaging points.
Identifying The Cause Of Noise
A common issue faced during computed tomography (CT) is erroneous positioning of the patient. This can be attributed to inaccurate vertical centring of the patient within the CT scanner which leads to poor image quality, noise and consequently repeated examinations.
After noticing this repeated issue, Booj et al. (2019) tried AI to rescue patients from repeated scans. The team used an AI enabled 3D camera that recreated the patient’s body mesh. This program provided a perfect center for patient position. The automated procedure evidently improved image quality. With advancements, AI algorithms can now identify the noise source and counter the issue by reconstructing accurate body representations from 2D images.
AI To Mitigate Orthodontic Image Noise
The good thing about Artificial intelligence is that it is continuously evolving and improving with every iteration. The programs aim to minimize humanoid issues while also keeping a track of the new challenges brought about by automation.
To counter problems of noise in health imaging, AI comes with denoising algorithms. AI denoising algorithm enhances diagnostic accuracy and also reduces radiological workflow. According to a study, the incorporation of AI significantly minimizes imaging noise and improves image quality. The result of such a change is quicker diagnosis without any compromises in diagnostic confidence.
CT scan has gradually become a medical necessity. Computed tomography scans accurately provide pictures of bodily structures at different angles. Noise in CT can jeopardize orthodontic diagnosis. There are several ways by which AI denoises a CT scan.
Filtered Back Projection
The primary method of imaging reconstruction is the filtered back projection. It was designed as an analytic reconstruction algorithm to overcome the issues and limitations of scans. It reduces blurring and noise from images. Research shows that AI-based filtered back projection improves the signal-to-noise ratio in CBCT scans.
Iterative Reconstruction
The latest and most effective way of denoising is Iterative reconstruction. In this method, an AI reconstruction algorithm begins with image assumption and then compares it to the present (real-time) values. The algorithm simultaneously adjusts the contrast such that the assumed image and the presented image are in agreement.
Advanced Modeled Iterative Reconstructive (ADMIRE) is the best way of reducing noise. When compared with weighted Filtered back projection (wFBP), ADMIRE did a better job. According to studies, deep-learning based reconstruction algorithms improve image quality by improving the contrast-to-noise ratio.
Benefits Of Reducing Noise In Orthodontics Using AI
There are multiple benefits of employing AI to minimize radiographic noise. AI saves a lot of precious time. The time taken by filtered back projection and iterative reconstruction is less. It also increases diagnostic confidence and saves patients from under-diagnosis. Repeated exposure to X-rays is discouraged due to the negative impact of radiation. The use of AI to efficiently reduce noise from the provided image saves repeated X-ray imaging of patients.
Final Word
Imaging noise is a frequent issue of orthodontic practice. Inaccurate and noisy images can interfere with cephalometric landmarks leading to poor orthodontic diagnosis. Scientific advancements have let AI do a wonderful job at reducing the noise without the need to repeat radiographs/CT scans.
The deep learning models of AI paired with neural networks compare the provided image with assumed image and reconstruct the areas masked by noise. Different methods of AI incorporation in noise reduction include filtered back projection and iterative reconstruction. Both methods have shown to significantly improve signal-to-noise ratio (SNR) in images. However, Advanced Model Iterative Reconstructive (ADMIRE) showed better results in less time. Reduction in noise allows precise diagnosis and saves time and prevents inaccurate orthodontic treatment plans.
References
- Booij, R., Budde, R. P., Dijkshoorn, M. L., & van Straten, M. (2019). Accuracy of automated patient positioning in CT using a 3D camera for body contour detection. European Radiology, 29, 2079-2088.
- Brendlin, A. S., Schmid, U., Plajer, D., Chaika, M., Mader, M., Wrazidlo, R., … & Tsiflikas, I. (2022). AI denoising improves image quality and radiological workflows in pediatric ultra-low-dose thorax computed tomography scans. Tomography, 8(4), 1678-1689.
- Brendlin, A. S., Estler, A., Plajer, D., Lutz, A., Grözinger, G., Bongers, M. N., … & Artzner, C. P. (2022). AI denoising significantly enhances image quality and diagnostic confidence in interventional cone-beam computed tomography. Tomography, 8(2), 933-947.
- Brendlin, A. S., Plajer, D., Chaika, M., Wrazidlo, R., Estler, A., Tsiflikas, I., … & Bongers, M. N. (2022). Ai denoising significantly improves image quality in whole-body low-dose computed tomography staging. Diagnostics, 12(1), 225.
- Kazimierczak, W., Wajer, R., Komisarek, O., Wajer, A., Kazimierczak, N., Janiszewska-Olszowska, J., & Serafin, Z. (2023). Enhanced Cone-Beam Computed Tomography Imaging through Deep Learning Model Reconstruction: Noise Reduction and Image Quality Optimization in Dental Diagnostics.