2014-11-06 JCEG Meeting Notes

3D less than 2D Pan + Ceph 

Martin Palomo give a presentation showing how CBCT scanners have dropped radiation dosage. 

The technology used to drop 3D dosage apparently might not be applicable to 2D scans, so in 

Toni Magni asked if it's possible to use the same technology used to reduce 3D radiation to reduce 2D radiation. 3D is based on redundant information, while 2D is a one shot deal, so it's harder to improve technology on a system that doesn't have room for improvement. So, the group agreed that it's not possible to reduce 2D further. 

Franco Magni suggests that the CBCT PAN at low dosage is not clinically acceptable, as it's resolution is too low.

Manish shows some real patient data of a QuickScan+ at 4.8s (very low dosage and resolution). Pano does not do justice in assessment to pathology for sure. He's not looking at pano's anymore for pathologies. He uses 3D reconstruction only.

Coffee Break

3D Norms

Can't make 3D models from CBCT because of resolution: not possible to have good resolution enough to build appliances and keep the low dosage. So surface scans are necessary. Right now the surface scanners are therefore developing in parallel with CBCT.

Can we extract stuff from 3dMD and compare/overlap on CBCT. Carla Evans says it's hard to overlap on CBCT cuz there is distortion on the 3dMD. It's easy to keep the subject steady, as they are just insant flashes. They can be setup to make a 360 surface of the head, but the problem is the hair, which gets captured as surface.

At Bolton Brush, there are 3D landmark data from the study thanks to the PA+Ceph combination taken w/o moving the patient.

What raw material is available? 

1. cross-section CBCT age 7-40
2. Longitudinal face scans
3. 3D Bolton landmarks
4. 2D midlines

Lunch Break

The problematic of certain article reviewers rejecting articles solely based on the technique used was discussed. For example, a reviewer that is against CBCT could reject all studies that are CBCT based. 

Intra-oral scanning

Air force wants to identify people without getting close to them. They are interested in what clinicians are using, how accurate they are to be able to identify people quickly in extreme situations. UIC is testing a scanner by Ormco to see how it can be used for this project. Each scan takes about 7 minutes for both upper and lower arch, but when they tried, the first time it took them 1 hour but with practice 14-15 minutes for the entire mouth. Resolution is relatively high, chipping can be detected. This scanner does not require powdering. Can use scanner in wet environment, since it requires little time. It requires multiple scans off of multiple surfaces, so it can reconstruct thenetire surface in post-processing. UIC did some comparisons with iTero for rugae studies, but it's convenience is mostly based on color and size (portability). 

0.1mm is the minimum resolution required for intra-oral scanning. There are two ways to get the digital intra-oral surface: 1. scan the oral cavity 2) scan the cast. This scanner scans oral cavity directly.

ORMCO takes a video scan, and shows result in real time, as opposed to snapshots. It uses laser beams like barcode scanners, using the fringe patterns, and the scanner looks at the reflection of the fringe pattern, thus being able to acquire the image in real time with video. The claim is that ORMCO technology is less accurate. However after a comparison with plaster model, statistically, they have not seen any difference. 

The scan is as accurate than a PVS impression. The scanner avoids the air bubble problematic. The scanner will pickup some defects on the enamel, which alginate cannot show.

3M is the only company that keeps their format completely proprietary. All others allow for STL or other open format exports. Align Technology bought Orthocad. The DVD of the AAO contains a prevention from Anthony Puntillo with a good description of what was available in May for intra-oral scanners.

What would be the ideal characteristics of the intra-oral scanners? Really what the popular interest is whether they are compatible with Invisalign or not. Assistants can take a scan in 8 minutes with iTero for both arches, chair time is not an issue anymore.

dr. Vicente Hernandez projects a table of intraoral scanners on the screen.

X (the statistician) projects some avg growth curves for each measurements. These are mathematical models of growth for growth prediction. Polynomial growth curves have advantages over non-linear as one can fit curves even only with 4 data points. 

3D superimpositions

Cranial superimpositions where shown using Dolphin Imaging. Carla Evans showed palatal rugae surface superimpositions. 

Pretesh showes superimposition of a model scan and an intra-oral scan. The difference between the two was low. Landmark based superimposition. GeoMagic is the software they used, which allows add on modules to be developed in python. Least square can be done, but probably a module needs to be developed for it.

Which software can visualise the cranial base the best?

It's not about the software, but it's about the acquisition.

Is anyone saving the raw images, in case in the future there might be a better reconstruction algorithm? Buzz asks why. 

How to collaborate in the open source software effort?

Harvard, Univ. Michigan are working on this. Send email to Lucy to find out. On the 3D Slicer web page there are links and information to learn how to collaborate.

Regional registration

The tough part is to find a region which is relatively stable. Roof of palate didn't work. The idea was to test the reference that Bjork said was stable. For mandible, they used the lingual part of the symphysis.

How long does it take to superimpose one mandible? With an experienced software user (10 month experience) and a good scan, a superimposition can be done in 8h of hard work.

Legacy collection meeting review

Powerpoint presentation of what happened with the legacy collection has just been posted on the AAO website.

3D CBCT Database

What we did with the legacy collection was not to include information that can be read or derived from the image itself. So age and gender, for example.

Toni Magni brought up the fact that maybe it might be useful for some studies to relate something with some calamities that happen, like war, or flooding, or nuclear fallout. HIPAA is limiting: if we put exact age, we can put scan date. There are 3 values: birthdate, scan date and age, and only one of these can be precise. It seems like the group agrees that year and month would be enough information, but is this still a privacy violation? How much does it have to be de-identified?

• Age/DOB/Date of scan
• Sex
• Race/Ethnicity
• Orthopaedic attempts
• Angle classification
• Dental Anomaly (Impacted/Congenitally missing teeth / topic eruption patterns)
• Craniofacial Anomaly
• Treated Yes/No at T0, at the first scan.
• Trauma

Buzz has a post-mortem card/checklist to go through once the case is being archived, before archiving them, for indexing purposes. This is a way to start.