#
Changelog
#
v4.14.0
/v4.14
Released: 2025-04-14
This is the latest stable version, recommended version for production use.
NOTE: When upgrading from a previous version, please go through the changelog below as there can be occasional breaking changes.
- Segmentation of the root canals on CBCT (e.g.
rootcanal_11
mesh structure). - Support for intra-oral scans in
.dcm
format.
- Improved IOS landmark segmentation.
- More robust handling of mallformed DICOM scans.
- More robust handling of
.nii
CBCT scans. - Better generation of simulated IOS roots.
- Bug fix related to the voxel_maps output file, preventing the client from selecting
all_teeth
as a structure. - Bug fixes related to the modelled IOS output files.
#
v4.8.0
/v4.8
Released: 2024-12-04
- Removed the
lower_closed_crown_gingiva
andupper_closed_crown_gingiva
structures (in the mesh output files).
- Support for deciduous fused teeth (e.g. newly added structures:
fused_tooth51
,fused_tooth52
, ..). - Automatically generated realistic roots for crown models (e.g.
root_prediction_crown_11
). - Automatic removal of buccal brackets and lingual wires on crown and fused tooth models (e.g.
cleaned_closed_crown_11
andcleaned_fused_tooth_11
). - New IOS modelling features, offering fine-grained shaping of the base with optional text engraving, adjustable gingiva margin, base height, and more.
- Added
fallbackToCbctTooth
andfallbackToIosCrown
options (underfusedMeshOptions
), allowing to fall back in case something went wrong in the fusion. - Added new options to customize the panoramic arch output file (see API docs for the changes).
- Improved the segmentation of the dentition on CBCT, especially on supernumerary cases and cases with impacted teeth.
- Improved computation of IOS landmarks (more accurate and robust).
- Improved the computation of closed crown output files.
- Improved the segmentation of the skull on CBCT.
- Improved the segmentation of the mandibular nerve canal on CBCT.
- Improved the segmentation of the incisive nerve and lingual foramen on CBCT.
- Improved the gingiva detection and IOS orientation detection for IOS scans.
- More robust reading of CBCT data with
.nii
extension. - Support for uploading result files to Azure Blob Storage, adding a mandatory header to the request.
- More robust handling of edentulous cases.
- Various improvements to the handling of exotic DICOM files (more robust parsing).
- Fixed parsing of options passed to the panoramic arch output file.
#
v4.0.0
/24.06
Released: 2024-06-28
This is the previous stable version. We recommend moving to the newest stable version for the latest features and improvements. Version 24.06
will still be supported for a few months.
- The per-structure (mandible, dentition, etc.) options
decimationFactor
andsmoothingFactor
have been removed, and replaced bysmoothingIterations
anddecimateMaxError
. This is due to the switching to new meshing algorithms which provide better meshes, more efficiently and more accurately.- The
smoothingIterations
option controls the additional amount of smoothing to apply to the mesh. The default value of 0 for small structures (teeth, nerves, etc.) and 1 for big structures (skull, mandible, etc.) should give good and accurate results in most cases, but if more smoothness is desired, higher values can be used; we encourage you to explore what works best for you. - The
decimateMaxError
option determines the maximum allowed error (in millimeters) when decimating the mesh. The default value of 0.03 for small structures (teeth, nerves, etc.) and 0.1 for big structures (skull, mandible, etc.) should result in a decent mesh size reduction with very minimal accuracy loss, but larger and smaller values can be used if desired. Setting this to zero will disable decimation completely (maximizing accuracy, but potentially increasing the mesh size significantly).
- The
- The makeManifold option for CBCT structures (cleaning the mesh and making it watertight) has now been removed. Almost all meshes will be manifold by default now, but if you want to be 100% sure that your meshes are manifold, you should set the
smoothingIterations
to 0. - The deprecated routes
/implant-job
,/cmf-job
,/orthodontics-job
, and/segmentation-job
have been removed; please use/job
(equivalent to/generic-job
).
- The
/generic-job
route will be deprecated in favor of the/job
route. Both routes have exactly the same functionality, the goal is to simplify the naming. - The
success
status of the job has been updated; it will now only befalse
only if all output files failed. If at least one output file succeeded, the job success will be marked astrue
. Note that it is still possible to check the status of each output file individually (see the section on handling job errors). /webhookJobResults
will be deprecated in favor of/pollJobResults
.- The
capSize
option underoptions.dicom
is nowfalse
by default. By default, DICOM files will thus be processed in their original resolution, regardless of how big they are. - The
individualRegistrationFusion
option is nowfalse
by default. It makes for more accurate results. - The
externalJobId
field in the job body is now marked as deprecated and will be completely removed in the next release. We recommend leaving this field empty and instead relying on the unique identifier generated by the Engine.
New
ios_landmarks_upper
andios_landmarks_lower
output files, returning a JSON with the automatically computed landmarks (mesial-distal points, long axis, etc.) of the intra-oral scan.New
cleaned_ios_upper
andcleaned_ios_lower
output files, returning the intra-oral scan with an added base and with rough edges removed.Added the option to automatically generate a gingiva with added base and/or closed emergent profile. These can be requested using
addBase
andcloseProfile
under the newly addedgingivaOptions
.Added the outputs
ios_local_orientation_upper
andios_local_orientation_lower
, which allow to automatically detect the orientation of the IOS, which allows to ensure that it always faces the user.Added new variations to the segmented gingiva meshes, which match better when displayed in combination with fused teeth or closed IOS crowns. Available through the new mesh structures:
upper_fused_teeth_gingiva
,lower_fused_teeth_gingiva
,upper_closed_crown_gingiva
andlower_closed_crown_gingiva
.Fused teeth meshes can now be decimated and smoothed.
New
keepPatientId
option for thedicom
output file, which will preserve the patient ID in the DICOM tags.
Greatly improved automated fusion of CBCT roots with IOS crowns. There is now a smoother transition between the IOS crown and CBCT root and better closing of the crown in the interproximal region. Moreover, the overall robustness and speed is improved.
Greatly improved AI segmentation of IOS dentition. It is now faster (~1.6x speedup) with improved accuracy and robustness (less under-segmentation and better outlining of the tooth border).
Greatly improved AI segmentation of (CB)CT dentition, especially on harder cases like the one shown below. Overall, there is a better separation of teeth in crowding scenarios, better classification and detection of teeth, and improved robustness and accuracy. Moreover, the speed got improved (~1.7x speedup).
Improved automated closing of IOS crowns. It is now almost 10x faster, has a smoother transition, and is more robust.
Improved AI segmentation of (CB)CT skull and mandibular canals.
Improved documentation.
- ZIP output has been made more robust, allowing it to still return partial results if one of the outputs inside the ZIP fails.
- More robust handling of sliced DICOM files with missing position and instanceNumber tags.
- More robust handling of DICOM files with empty group tags.
- More robust IOS to CBCT registration.
More info about the benchmark can be found at the benchmark page
#
Segmentations
#
Algorithms
#
v3.13.0
/v2
Released: 2024-02-23
This is an outdated stable version. We recommend moving to the newest stable version for the latest features and improvements. Version v2
will soon get deactivated.
No older version to compare to, checkout version 24.06
for the latest changes available.
More info about the benchmark can be found at the benchmark page