 |
| Title |
Artificial Intelligence May Aid In The Diagnosis Of Carotid Artery Calcifications |
| Clinical Question |
For patients undergoing dental radiographic examination, can artificial intelligence be used to increase a dentist’s diagnostic accuracy of carotid artery calcifications? |
| Clinical Bottom Line |
Artificial intelligence systems alone cannot yet be relied upon for increasing the diagnostic accuracy of carotid artery calcification detection. Dentists have a professional responsibility to detect these calcifications on any radiograph they order. Failure to do so may result in the misdiagnosis of life-threatening cardiovascular diseases and stroke. Given the variability in dentists' educational backgrounds and experiences, incidental radiographic findings are often overlooked. While AI has the potential to assist in preventing such oversights in the future, for now dentists must ensure they possess adequate radiographic competency or know when to refer their radiographs to specialists as necessary. |
| Best Evidence |
(you may view more info by clicking on the PubMed ID link) |
| PubMed ID |
Author / Year |
Patient Group |
Study type
(level of evidence) |
| #1) 36292226 | Ajami/2022 | 56 CBCT scans (15,257 axial slices) | Diagnostic accuracy study | | Key results | 56 CBCT scans (15,257 axial slices) were used to train, validate, and test a deep learning artificial intelligence model for detection of cervical carotid artery calcifications. 40 CBCT scans (11,252 axial slices) were utilized to train the deep learning model via a two-step process to first detect the correct region that carotid calcifications would be found (below C2-C3 vertebrae), and secondly to correctly identify carotid artery calcifications. 16 CBCT scans (4,005 axial slices) were used to test the deep learning model. Half of the training and test sets were positive for carotid artery calcifications as confirmed by a board-certified Oral and Maxillofacial Radiologist. The test group resulted in 94.2% sensitivity and 96.5% specificity. The positive predictive value was 56.9% and the negative predictive value was 99.7%. The diagnostic accuracy was 96.35%. The investigators concluded that reliable deep learning models can be used as an effective tool in detecting carotid calcifications in CBCT imaging. | | #2) 36352043 | Song/2022 | 352 Panoramic Radiographs | Comparative study of diagnostic accuracy | | Key results | The sensitivity and specificity of artificial intelligence alone was 0.774 and 0.717. The sensitivity and specificity of the general dentist group during the first reading was 0.219 and 0.645, and for the second reading with AI assistance it was 0.74 and 0.78. The sensitivity and specificity of the oral radiologist group during the first reading was 0.9 and 0.83, and for the second reading with AI assistance it was 0.81 and 0.883. The diagnostic accuracy of the general dentists increased from 0.645 to 0.759. The diagnostic accuracy of the oral radiologists increased from 0.829 to 0.875. This study shows that artificial intelligence increased the diagnostic accuracy of general dentists and oral radiologists in the diagnosis of carotid artery calcifications. | |
| Evidence Search |
(Artificial Intelligence OR Deep Learning) AND (Carotid Calcifications OR Atherosclerosis) |
Comments on
The Evidence |
Ajami/2022 utilized CBCT scans from 2009-2019 with confirmed diagnosis of cervical artery calcifications via the gold standard, a board-certified oral and maxillofacial radiologist, to train and test deep learning artificial intelligence accuracy. Human level supervision was used to locate the calcifications in training and utilized image segmentation to identify the regions of interest for the deep learning model. Testing limitations included a small sample size from one CBCT machine, no power analysis to determine the appropriate sample size for model testing, and exclusion of other calcifications found in the neck to train the model in differentiation. Testing metrics of the deep learning model are based on sensitivity, specificity, positive and negative predictive values, and diagnostic accuracy.
In Song/2022, 352 panoramic radiographs were reviewed by a general dentist group and an oral radiologist group. Two weeks later, the same radiographs were reviewed by the same groups, but with the assistance of FAST-RCNN (fast-region-based convergence neural network with ResNet Back-bone). This article evaluates a single artificial intelligence model; there are numerous other models. Additionally, the sample size of participating dentists is small and may not adequately represent the entire population.
|
| Applicability |
It is the responsibility of every dentist to diagnose all incidental findings on any radiograph they capture. With the increasing use of CBCT machines, many dentists may not have received education on these technologies during dental school. While studies demonstrate that artificial intelligence alone may not achieve satisfactory diagnostic accuracy, it significantly enhances diagnostic accuracy when combined with a dentist's clinical judgment. However, as of now, there is no AI model available to aid dentists in radiologic diagnosis. Further research and development are necessary. |
| Specialty/Discipline |
(Oral Medicine/Pathology/Radiology) |
| Keywords |
Artificial Intelligence, Carotid Artery Calcification, radiology
|
| ID# |
3574 |
| Date of submission: |
10/22/2024 |
| E-mail |
ceolla@livemail.uthscsa.edu |
| Author |
Stephen Ceolla |
| Co-author(s) |
Ashleigh Denny |
| Co-author(s) e-mail |
dennya2@livemail.uthscsa.edu |
| Faculty mentor/Co-author |
Dr. Kyumin Whang |
| Faculty mentor/Co-author e-mail |
Whang@uthscsa.edu |
Basic Science Rationale
(Mechanisms that may account for and/or explain the clinical question, i.e. is the answer to the clinical question consistent with basic biological, physical and/or behavioral science principles, laws and research?) |
post a rationale |
| None available | |
|
Comments and Evidence-Based Updates on the CAT
(FOR PRACTICING DENTISTS', FACULTY, RESIDENTS and/or STUDENTS COMMENTS ON PUBLISHED CATs) |
post a comment |
| None available | |
|
|
