AI and Imaging. Podcast Topic Summary AAO 2022 Episode 154. September 21st 2022

9 months ago

Hosted by

Farooq Ahmed

AI and Imaging. Podcast Topic Summary AAO 2022

Join me for a summary of this years American Association Meeting from the summer looking at AI and imaging. The topics covered at CBCT the facts, use in transverse and incidental findings. Other topics of facial imaging or 4D as a diagnostic tool. Finally looking at AI, its application currently and potential use of blockchain technology in orthodontics.

CBCT Shaza Mardini

Myths

It is new developed 1990s, used 2001
Dose is too high – now just over the dosage of cephalogram and OPG, as low as 46uSv Buckley 2018 – 5 hour flight
- Children sensitive due to growth
- ALARA, ALADA to bear in mind

Accuracy

Small deviation of true size compared to 2D images Gregory 2004
Panoramic = not accurate and only screening tool Lione 2000

Asymmetry

Accurate measurements for bilateral structure is possible
Degenerative changes in condyle is often responsible for open bite which can be detected by
Example of twins wherein one child with condylar degenerative issue had retro gnathic mandible leading to malocclusion

CBCT and Transverse assessment Onur Kadioglu

PA Ceph should not be utilized for transverse discrepancy. {Cheung et al Aust orthod 2013}

CBCT as gold standard and compared PA ceph to it and has quoted that it has

↓False positives and high degree of sensitivity and specificity
Correctly predicts 88.7% of crossbite; 91.25% no crossbites.

landmark system for transverse measurements using CBCT images. Onur Kadioglu

Maxillary teeth Trifurcation of molar (less likely to change in angulation changes)
mandibular teeth midpoint of root
- 20 mm discrepancy = crossbite

Used transverse discrepancy limit of 20mm to assess the outcomes of cases.

Incidental findings with CBCT Onur Kadioglu

occurrence rate of 4.7 per CBCT
Overall Maxilla Zone 5 Mandible accounted for most common incidental finding.
Most were variation of normal anatomy 62.2%
Findings requiring a referral 11.3%
Checklist produced
- Variation from normal
  - Dilaceration, nasal septal deviation, enlarged incisal canal
- Incidental findings
  - Retained teeth, Mucosal thickening, Idiopathic osteosclorosis
- Incidental finding – refer
  - External RR, Odontogenic pathology

Facial imaging William Harrell

3D facial imaging is ‘sister’ technology to CBCT

3D facial surface image capture and co – registration to CBCT in one patient specific interactive 3D model

2D Vs 3D Vs 4D

2D helps to precisely measure INACCURACY and its reproducible
Accurate in 3D space; one needs to be careful in locating landmarks
4D imaging allows us to have shape analysis of a structure in dynamic state with aid of colour coding

Study’s on facial imaging 3D

Predicting sleep apnea from 3D face photography {Peter eastwood Journal of sleep medicine 2019}

17% to 19% increase in posterior mandibular height;
10% to 13% increase in lateral face height
18% smaller maxillary-mandibular relationship angle
High sensitivity low specificity – screening tool but not diagnostic Positive likelihood ratio 4 (for every 4 people positive only 1 would have the condition)

Influence of snoring mouth breathing and apnea on facial morphology in late childhood – A 3D study {Richmond S et al BMJ sept 2022}

the nasomaxillary complex is depressed sides of the cheeks are narrow and mandible is down and backwards placed- all responsible for reduced airway

AI Veerasathpurush Allareddy

Big data landscape –

With big dataset in healthcare, T test, Chi square test won’t work
- Amplifying bias
- Can’t deal with non- linear data

Whereas Big data analytics can

Discover new patterns
Deductive in nature

Machine learning (subfield of Artificial Intelligence)–

Common ML Algorithm –
- Supervised, unsupervised, reinforced and finally deep learning. Resolve real world problems

Application in orthodontics

Growth prediction and treatment planning

Reseach

21 publications 2018

AI and craniofacial genomics –

Although orthodontic diagnosis is phenotypically based, but a genomic diagnosis is likely to help the orthodontists.
Biological markers of tooth movement Altareddy 2019
- Not just phenotypic diagnosis

Blockchain Technology –