SlicerAutomatedDentalTools

Slicer Automated Dental Tools

The Slicer Automated Dental Tools extension provides automatic dental and craniofacial analysis capabilities. It features a user-friendly graphical interface on 3D Slicer, enabling users to perform complex tasks without any coding expertise.

Compatible with both stable and nightly versions of 3D Slicer. Latest versions supported: 5.7.0 (nightly) and 5.6.1 (stable)

Overview

Slicer automated dental tools is an extension that allows users to perform automatic segmentation, landmark identification and Automatic Orientation on CBCT scans and Intra Oral Scan (IOS) using machine learning tools where the learning mdoels are continously updated.

Features

• Simple: Perform time consuming task with a few clicks.
• Automatic: Automate important Dental and Cranio Facial analysis tasks.
• Flexible: The user can choose which models to use to perform the automated task. New models can be added easily.

Modules

Name	Description
AMASSS	Perform automatic segmentation of CBCT scan. AMASSS is an acronym for Automatic Multi-Anatomical Skull Structure Segmentation.
ALI	Perform automatic landmark identification on either CBCT or IOS scans. ALI is an acronym for Automatic Landmark Identification.
ASO	Perform automatic orientation either on IOS or CBCT files.
AReg	Perform automatic registration on IOS or CBCT files.
AutoCrop3D	Automatically crop a folder of CBCT scans with the same region of interest.
AutoMatrix	Automatically apply one or different matrix to a folder of IOS or CBCT scans.
Medical Data Anonymizer	Automatically anonymize sensitive information in medical documents (DOCX, PDF, TXT, CSV, XML, ODT).
MRI2CBCT	Register TMJ MRIs onto their CBCTs in one run.
FlexReg	Registration of IOS patient per patient with customizable patch creation.
DOCShapeAXI	Automatic classification of 3D Shape. DOC-ShapeAXI is an acronym for Dental Oral and Craniofacial Shape Analysis eXplainability and Interpretability.
BatchDentalSegmentator	DentalSegmentator in batch for mixed or permanent dentition
CLI-C	Classification and Localization of Impacted Canines
VFACE	Visual Facial Asymmetry Classification Engine. Automated classification of facial asymmetry.
GreedyReg	ITK-SNAP style registration of CBCT scans using the Greedy engine (manual, automatic and distant modes, single or batch).
Agent	AI assistant that runs the other modules from a natural-language request, using a local LLM.

These modules provide a convenient user interface, are available through the Automated Dental Tools module category, and share common features :

Input

• All modules can work with one file or a whole sample (folder) as input.
• If the input is a single file already loaded, the result of the predicton will directly show up on the slice views.

Output

• By selecting the “Group output in a folder” checkbox, all the ouput files will be grouped in a single folder for each patient.
• All modules allows the user to save the output in the input folder, or by unchecking the “Save prediction in scan folder” the user can choose a custom output folder.
• The “Prediction ID” field is for the user to choose what will appear on the output file name. (“Pred” by default)

Additionally, the following modules are implemented as python scripted command-line (CLI) modules available in the Automated Dental Tools.Advanced module category and are used internally by the modules described above.

Name	Description
AMASSS_CLI	Perform automatic segmentation of CBCT scans
ALI-CBCT	Perform automatic landmark identification of CBCT scans
ALI-IOS	Perform automatic landmark identification of IOS scans
ASO-CBCT	Perform automatic orientation of CBCT scans
ASO-IOS	Perform automatic orientation of IOS scans
AReg-CBCT	Perform automatic registration of CBCT scans
AReg-IOS	Perform automatic registration of IOS scans
AutoMatrix-CLI	Apply transformation matrices to input volumes or landmarks
Medical Data Anonymizer	Automatically anonymized text files.
MRI2CBCT_PREPARE	Pair the MRI and CBCT folders by patient and prepare each MRI joint
MRI2CBCT_SEGMENT_MRI	Segment the cranial base on the prepared MRI joints (nnU-Net)
MRI2CBCT_REGISTER	Register each MRI joint onto its CBCT from the two cranial base surfaces
MRI2CBCT_RESAMPLE_CBCT_MRI	Resample MRI and CBCT scans (used by AReg and VFACE)
FlexReg_CLI	Perform creation of patch and registration on IOS scans.
DOCShapeAXI	Perform automatic classification of 3D Shape.
BatchDentalSegmentator	DentalSegmentator in batch for mixed or permanent dentition
CLI-C	Classification and Localization of Impacted Canines
VFACE_CLI	Classify automatically facial asymmetry based on linear and angluar measurements
GreedyReg_CLI	Run Greedy-based registration of CBCT scans
Agent_CLI	Route a natural-language request to the right tool, extract its parameters and build the command to run

Requirements

• In addition of the Slicer System requirements, for best performance, 12GB of memory is recommended.
• :warning: Trained networks are required to be manually downloaded. See requirements section specific to each module.

---

AMASSS Module

AMASSS module will allow you to segment CBCT scan using AMASSS algortihm.

Prerequisites

• Download the trained models for AMASSS using the Download latest models button in the module Input section.

Module structure

Input file: The input has to be an oriented CBCT. It can be a single CBCT scan loaded on slicer or a folder containg CBCTs with the following extention:

.nrrd / .nrrd.gz
.nii  / .nii.gz
.gipl / .gipl.gz

Available sample data for testing: MG_test_scan.nii.gz

Load models: The user has to indicate the path of the folder containing the trained models for AMASSS.

Segmentation selection: The user can choose the structure to segment using the selection table. Depending on the type of CBCT to segment, the user can select the “Use small FOV models” checkbox to use on higher definition scans.

Output option: By selecting the “Generate surface files” checkbox. The user will also get a surface model of the segmentation that will be saved in a “VTK files” folder and will be automatically loaded in slicer at the end of the prediction if working on a single file.

Advanced option:

• You can increase/decrease the precision of the segmentation (going above 50 will drastically increase the prediction time and is not necesary worth it, going under 50 will make the prediction much faster but less accurate)
• If the user whant to generate surface files, he can choose the smothness applied on the model.
• Depending on your computer power, you can increase the CPU and GPU usage to increase the predictio speed.

---

ALI Module

ALI module provide a convenient user interface allowing to identify landmarks on different type of scans:

• CBCT scan
• IOS scan

ALI-CBCT

The implementation is based on the ALI-CBCT algortihm originally developed by Maxime Gillot at https://github.com/Maxlo24/ALI_CBCT.

Prerequisites

• Download the trained models for ALI-CBCT using the Download latest models button in the module Input section.

Module structure

Input file: The input has to be an oriented CBCT. It can be a single CBCT scan loaded on slicer or a folder containg CBCTs with the following extention:

.nrrd / .nrrd.gz
.nii  / .nii.gz
.gipl / .gipl.gz

Available sample data for testing: MG_test_scan.nii.gz

Load models: The user has to indicate the path of the folder containing the trained models for ALI-CBCT.

Landmark selection: Once the folder containing the trained models is loaded. The user can choose the landmark he want to identify with the table showing the available landmarks:

---

ALI-IOS

The implementation is based on the ALI-IOS algortihm originally developed by Baptiste Baquero at https://github.com/baptistebaquero/ALIDDM.

Prerequisites

• Download the trained models for ALI-IOS using the Download latest models button in the module Input section.

Module structure

Input file: The input has to be an oriented IOS segmented with the Universal Numbering System. This segmentation can be automatically done using the SlicerJawSegmentation extention. The input can be a single IOS loaded on slicer or a folder containg IOS with the following extention: The array name of labels in the vtk surface is: “Universal_ID” or “predictionid” or “PredictionID”

.vtk

Available sample data for testing: T1_01_L_segmented.vtk and T1_01_U_segmented.vtk

Load models: The user has to indicate the path of the folder containing the trained models for ALI-IOS.

Landmark selection: For the IOS landmarks, the user has to choose which tooth he need the landmakrs on by checking the label of the tooth on the left table. Once the folder containing the trained models is loaded. The user can choose the landmark he want to identify with the table on the right showing the available landmarks:

ASO Module

If you want more information and a descriptive tutorial of this tool, take a look at this github page: Automated Standardized Orientation

ASO module provide a convenient user interface allowing to orient different type of scans:

• CBCT scan
• IOS scan

How the module works?

2 Modes Available (Semi or Fully Automated)

• Semi-Automated (to only run the landmark-based registration with landmark and scans as input)
• Fully-Automated (to perform Pre Orientation steps, landmark Identification and ASO with only scans as input)

Mode	Input
Semi-Automated	Scans, Landmark files
Fully-Automated	Scans, ALI Models, Pre ASO Models (for CBCT files), Segmentation Models (for IOS files)

Input file:

Input Type	Input Extension Type
CBCT	.nii, .nii.gz, .gipl.gz, .nrrd, .nrrd.gz
IOS	.vtk .stl .vtp .off .obj

Test Files Available: You can either download them using the link or by using the Test Files button.

Module Selected	Download Link to Test Files	Information
Semi-CBCT	Test Files	Scan and Fiducial List for this Reference
Fully-CBCT	Test File	Only Scan
Semi-IOS	Test Files	Mesh and Fiducial List Reference
Fully-IOS	Test Files	Only Mesh Reference

Reference:

The user has to choose a folder containing a Reference Gold File with an oriented scan with landmarks. You can either use your own files or download ours using the Download Reference button in the module Input section.

Input Type	Reference Gold Files
CBCT	CBCT Reference Files
IOS	IOS Reference Files

Landmark selection

The user has to decide which landmarks he will use to run ASO.

Input Type	Landmarks Available
CBCT	Cranial Base, Lower Bones, Upper Bones, Lower and Upper Teeth
IOS	Upper and Lower Jaw

The landmark selection is handled in the Landmark Reference Section:

Models Selection

For the Fully-Automated Mode, models are required as input, use the Select Button to automatically download, extract and select the selected models.

AReg Module

AReg module provide a convenient user interface allowing to orient different type of scans:

• CBCT scan
• IOS scan

How the module works?

3 Modes Available

• Semi-Automated (to only run the automatic registration with masks segmentation and scans as input - only for CBCT)
• Fully-Automated (to perform the automatic registration with only scans as input)
• Orientation and Registration (to perform the automatic orientation using ASO process and the automatic registration with only scans as input)

Mode	Input
Semi-Automated (only for CBCT)	Scans, Masks segmentation
Fully-Automated	Scans, Segmentation Models
Orientation and Registration	Scans, ALI Models (for CBCT files), Segmentation Models

For CBCT, a final step is added and consist of segmenting different skeletal structures using AMASSS process within each mode.

Input file:

Input Type	Input Extension Type
CBCT	.nii, .nii.gz, .gipl.gz, .nrrd, .nrrd.gz
IOS	.vtk .stl .vtp .off .obj

Test Files Available: You can either download them using the link or by using the Test Files button.

Module Selected	Download Link to Test Files	Information
Semi-CBCT	Test Files	Scan and Masks segmentation
Fully-CBCT	Test File	Only Oriented Scan
Orientation and Registration for CBCT	Test File	Only Scan
Fully-IOS	Test Files	Only Mesh )
Orientation and Registration for IOS	Test Files	Only Mesh

Pipelines

CBCT - AReg Pipeline

IOS - Patch prediction

Some Results

Models Selection

For the Fully-Automated Mode, models are required as input, use the Select Button to automatically download, extract and select the selected models.

AutoCrop3D Module

AutoCrop3D stands for “Scans automatically cropped” according to a Region Of Interest (ROI).

Useful Information

This module has the same utility as “Crop Volume”, which offers more interactive feedbacks to the user, but “AutoCrop3D” allows you to easily trim the ROI of all the cases in your study. Then you can load your files into Slicer to view them. More Information about Crop Volume

It solves the issue of files that may be too heavy for “Crop Volume Sequence” More Information about Crop volume Sequence.

How does the module work?

Input:

Input	Input Type	Input Extension Type	Files to Test
File or Folder	CBCT	.nii, .nii.gz, .gipl.gz, .nrrd, .nrrd.gz	Segmentation.zip
Region of Interest*	Volume	.json	ROI.mrk.zip

You can choose a suffix which will be added to the existing name of your file(s).

If you want to generate VTK files from your Segmentations, the Segmentation files must have “Seg” somewhere in the name. For example: patientName_Seg.nii.gz

How to create a Region Of Interest?

Use the module “Volume Rendering” to create your ROI and then save it as a .json file.

1. Upload a scan file
2. Choose the option Display ROI
3. Change the ROI as you like
4. Save it

AutoMatrix Module

The AutoMatrix module provides a user interface to apply a matrix to a folder for different types of scans or segmentations:

• CBCT
• IOS

How the module works?

How to name the new files and where they are stored

The module will replicate the folder structure from your input folder inside the output folder. Each new file will keep the original filename, followed by the suffix you specify. If the “Add matrix name to filename” option is enabled, the matrix filename will also be appended.

Example:

With matrix name	Without matrix name
Input file: patient1_T1_MA.nii.gz	Input file: patient1_T1_MA.nii.gz
Input matrix: patient1_matrix1.tfm	Input matrix: Matrix_mirror.tfm
Suffix: _apply	Suffix: _mir
Output file: patient1_T1_MA_apply_matrix1.nii.gz	Output file: patient1_T1_MA_mir.nii.gz

4 Modes available

Sometimes a LinkName is required. See the information below.

Mode	Download Link to Test Files	Information
File patient and file matrix	Test Files	LinkName is not required. The matrix is applied to the input file. This new file is save in the output folder.
File patient and folder matrix	Test Files	LinkName required. For each matrix applied, a new file is created.
Folder patient and file matrix	Test Files	LinkName is not required. The matrix will be applied to all files in the input folder.
Folder patient and folder matrix	Test Files	LinkName required. For each matrix applied, a new file is created. You can have more than one file with the same patient name.

When a LinkName is required :

The name of the matrix files and the name of the seg/scan files must begin with the same patient name. After the patient name, you must put and underscore before any other information. Example : Name file patient : patient1_T1_MA.nii.gz Name matrix files : patient1_left_MA.tfm

Mirror :

There is button “Mirror” that will automatically download the matrix mirror and put in input.

Input file:

Input Type	Input Extension Type	Input Matrix Extension
CBCT	.nii.gz	.tfm .npy .h5 .mat .txt
IOS	.vtk .stl .vtp .off .obj	.tfm .npy .h5 .mat .txt
Landmark	.mrk.json	.tfm .npy .h5 .mat .txt

Medical Data Anonymizer Module

The Medical Data Anonymizer module provides automated anonymization of sensitive information in medical documents using AI-powered text analysis. It helps ensure HIPAA compliance and patient privacy protection by identifying and removing or replacing personally identifiable information (PII) and protected health information (PHI) from various document formats.

Features

Supported File Formats:

• PDF documents
• Word documents (.docx)
• Text files (.txt)
• CSV files (.csv)
• XML files (.xml)
• OpenDocument Text (.odt)

Anonymization Options:

• Names (patients, doctors, healthcare providers)
• Phone Numbers
• Email Addresses
• Dates and Times
• Addresses and Locations
• Social Security Numbers
• Medical License Numbers
• Driver’s License Numbers (optional)
• Credit Card Numbers (optional)
• Bank Account Numbers (optional)
• IP Addresses (optional)
• URLs/Websites (optional)

Anonymization Methods:

• Replace with Label: Replaces sensitive data with generic labels (e.g., <PERSON>, <DATE>)
• Redact: Completely removes sensitive information
• Hash: Replaces with cryptographic hashes for consistent anonymization
• Mask: Replaces with asterisks or other masking characters

How does the module work?

1. Install Dependencies:

   • Click “Install Dependencies” to automatically install required libraries (Presidio, file processing tools)
   • Restart 3D Slicer after installation

2. Select Input and Output:

   • Input Directory: Folder containing documents to be anonymized
   • Output Directory: Folder where anonymized documents will be saved

3. Configure Anonymization:

   • Choose Entity Types: Select which types of sensitive information to anonymize
   • Advanced Options:
     • Anonymization Method: Choose how to handle detected sensitive data
     • Confidence Threshold: Set detection sensitivity (0.0-1.0, default: 0.5)

4. Process Files:

   • Click “Anonymize Files” to start processing
   • Progress bar shows real-time status
   • Anonymized files are saved with “_anonymized” suffix

Output

• Anonymized Documents: Each input file generates an anonymized version in the selected format
• File Mapping CSV: file_mappings.csv tracks original and anonymized filenames with unique identifiers
• Processing Log: Detailed logging of the anonymization process

Use Cases

• Research Data Preparation: Remove PHI from clinical notes and research documents
• Data Sharing: Prepare medical documents for sharing with external collaborators
• Compliance: Ensure HIPAA compliance when processing medical documents
• Archive Management: Create anonymized versions of medical records for long-term storage

Prerequisites

• Dependencies: Automatically installed via the “Install Dependencies” button
• Language Models: English language model is automatically downloaded
• Supported Systems: Windows, macOS, Linux

MRI2CBCT Module

The MRI2CBCT module provides a user interface to perform the registration between MRI and CBCT scans.

How does the module work?

The whole registration runs from one button. Choose an MRI folder, a CBCT folder and an output folder; patients are paired by the ID at the start of the file name, with one CBCT and one or two MRIs per patient.

1. Set up this computer (once): installs the Python packages and downloads the models the pipeline needs.
2. MRI preparation: each MRI is reoriented and, for a bilateral acquisition, split into a left and a right joint using the acquisition slice spacing. It can be resampled to the CBCT resolution (0.3 mm).
3. CBCT orientation: the CBCT is oriented with ASO (PRE_ASO_CBCT, ALI_CBCT, SEMI_ASO_CBCT).
4. Cranial base segmentation: on the CBCT with AMASSS, on each MRI joint with a dedicated nnU-Net model.
5. Registration: each MRI joint is registered onto its CBCT by matching the two cranial base surfaces (FPFH, RANSAC, ICP), with an automatic recovery from flipped solutions.
6. Review: browse the results patient by patient and swap left and right when a side was assigned wrongly.

For each patient and side the output folder holds the registered MRI, its mask and the transform, plus the CBCT mask and a summary.csv with a quality check per joint.

Use test files downloads a sample patient (one CBCT, two MRIs) once and fills in the three folders, so pressing Run is all that is left; the expected result is next to it, in reference/.

FlexReg Module

FlexReg is a module that allows you to register patient-specific Intra Oral Scans. It lets you create custom patches for registration.

How does the module work?

1. Load your files:

   • Choose your path by clicking on “Select” and load it by clicking on “View”.

2. Draw your own patch:

   • Butterfly Patch:
     • You can draw a butterfly patch using the parameters and select your own teeth. Note that this patch, with the default values, is called “ButterflyPatch” and is created when you run AREG_IOS:

       

   • Custom Patch:
     • You can draw the patch using landmarks and customize it:

     • 

   • You can combine multiple patches by creating them one by one. All your patches will be used as a single patch for the registration.

   • 

3. Registration:

   • After creating your patch for both the fixed and moving scans, choose the output folder and the suffix you would like. If you have a lower arch to move, you can select it.
   • Register them by clicking on “Registration”.

DOCShapeAXI Module

This integrates advanced deep learning algorithms enabling the classification of nasopharynx airways obstruction, mandibular condyles, and alveolar bone defect in cleft. The tool facilitates the automated classification of medical shapes, assisting in clinical diagnosis and decision-making. The extension was originally developped https://github.com/lucieDLE/DOC-ShapeAXI/

Prerequisites

• This extension works with Linux and Windows.
• This extension requires the installation of SlicerConda. Follows the instructions for more information.

Supported Dataset

Currently, the extension supports only three types of datasets:

• Nasopharynx Airway Obstruction: Three models are available for this dataset.
  • Severity model: classifies 4 levels of airway obstruction ranging from 0 to 3, where 0 is the lowest level and 3 is the greatest level.
  • Binary model: classifies between 0 and 1, where 0 is the absence and 1 is the presence of adenoid hypertrophy-related obstruction.
  • Regression model: directly predicts the ratio of obstruction ranging between 0 and 100%.
• Mandibular Condyle:
  • Severity model: classifies the condylar changes into 4 levels of severity ranging from 0 to 3.
• Alveolar Bone Defect in Cleft:
  • Severity model: classifies the defect into 4 levels of severity ranging from 0 to 3.

How does the module work?

1. Select a Data Type: nasopharynx airways obstruction, mandibular condyles, and alveolar bone defect in cleft.

2. Select an Input folder: a folder that contains 3D models stored as VTK files.
3. Select an Output directory: a folder that will contain all the outputs (model, predictions results stored in a CSV file, and explainability visualizations).
4. Run the module: Click on the Run Prediction button.

Viewing the Results

All the results are saved in the Output directory specified by the user. you can find:

A. the csv file containing the prediction made by the network.

B. a folder named Explainability containing the visualization results:

1. Open the folder and drag and drop a shape in Slicer
2. Go to the module Models
3. In the Display drop-down menu, click on the Scalar section
4. check the visibility box and change the color table to ColdToHotRainbow
5. Set the Data Scalar Range to Auto if not set
6. Change the Active Scalar to the class you want to visualize

BatchDentalSeg

3D Slicer extension for fully-automatic segmentation of CT and CBCT dental volumes in batch using Dentalsegmentator model (for adults scans), PediatricDentalsegmentator model (for children scans), UniversalLabDentalSegmentator (for labeling all the teeth), NasoMaxillaDentalSegmentator (for segment separatly the upper skull and the Naso Maxillary Complex.

After selecting the following:

1. Input folder (containing volumes to process)
2. Output folder (where segmentations will be saved)
3. Model (the pretrained network that best fits your scans)

this module generates the following labeled regions for each volume:

PediatricDentalSegmentator

(Identical to DentalSegmentator)

• Maxilla & Upper Skull
• Mandible
• Upper Teeth
• Lower Teeth
• Mandibular canal

DentalSegmentator

• Maxilla & Upper Skull
• Mandible
• Upper Teeth
• Lower Teeth
• Mandibular canal

NasoMaxillaDentalSegmentator

This model splits the upper skull into two regions and also segments teeth and mandible:

• Upper Skull (excluding nasal/maxillary complex)
• Naso-Maxilla Complex
• Mandible
• Upper Teeth
• Lower Teeth
• Mandibular canal

UniversalLabeling

Same as DentalSegmentator, but each tooth receives its own unique label:

• Maxilla & Upper Skull
• Mandible
• Upper Teeth (each tooth labeled separately)
• Lower Teeth (each tooth labeled separately)
• Mandibular canal

DentalSegmentator model

DentalSegmentator is based on nnU-Net framework. It has been trained on 470 dento-maxillo-facial CT and CBCT scans from 7 institutions by Dot G, et al, here is their paper and nnUnet paper:

Dot G, et al. DentalSegmentator: robust open source deep learning-based CT and CBCT image segmentation. Journal of Dentistry (2024) doi:10.1016/j.jdent.2024.105130

Isensee F, et al. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods. 2021;18(2):203-211. doi:10.1038/s41592-020-01008-z

PediatricDentalSegmentator model

As DentalSegmentator, PediatricDentalSegmentator is also based on nnUnet. It has been trained on 513 dento-maxillo-facial CBCT scans including scans withs primary teeth.

UniversalLabDentalSegmentator model

As PediatricDentalSegmentator, UniversalLabDentalSegmentator is also based on nnUnet. It has been trained on 513 dento-maxillo-facial CBCT scans including scans withs primary teeth

NasoMaxillayDentalSegmentator model

As DentalSegmentator, NasoMaxillaryDentalSegmentator is also based on nnUnet. It has been trained on 135 dento-maxillo-facial CBCT scans including scans withs primary teeth.

Using the extension

This extension is compatible with 3D Slicer Stable Release (version 5.8.0, or later), downloadable from the official website.

The plugin can be installed in Slicer using the extension manager. It’s a module in the Automated Dental Tools extension so it can be can be found using the search bar by typing “Automated Dental Tools”.

After the installation process and restart of Slicer, the extension can be found in the module file explorer under Automated Dental Tools>BatchDentalSegmentator. It can also be found by using the find module button and searching for the keyword BatchDentalSegmentator.

You will also need to install the NNUNet extension, you can find it by using the find module button and searching for the keyword NNUNet

To use the extension, click on Select Folder to select the folder where your scans are and click on Select Output Folder to select the folder where your segmentations will be save.

Then choose the model that you want to use :

Finally click on the Apply button to start the segmentations.

If your device doesn’t include CUDA, the processing may be very long and a dialog will ask for confirmation before starting the segmentation process.

During execution, the processing can be canceled using the Stop button. The progress will be reported in the console logs.

After the segmentation process has run, the segmentation will be loaded into the application. The segmentation results can be modified using the Segment Editor tools.

The segmentation can be exported using the Export segmentation menu and selecting the export format to use.

The Surface smoothing slider allows to change the 3D view surface smoothing algorithm.

Exemple of using

CLI-C Module

CLI-C (Classification and Localization of Impacted Canines) is a 3D Slicer extension designed for automated segmentation of dental Cone-Beam Computed Tomography (CBCT) images using a deep learning Mask R-CNN model. It streamlines dental segmentation workflows for intra-osseous teeth, facilitating accurate identification and visualization of impacted canines for orthodontic and dental surgical planning.

---

Key Features

• Automatic Segmentation: Precisely segments and classify impacted teeth, determininng the intra-osseous localization including buccal, bicortical, and palatal regions from CBCT scans.
• Mask R-CNN Model Integration: Utilizes advanced deep learning architectures to provide accurate and robust segmentation.
• Progress Monitoring: Real-time progress bar and log updates during segmentation tasks.
• User-Friendly Interface: Simple GUI integrated seamlessly within 3D Slicer.
• Batch Processing: Supports segmentation of single files or batch processing of entire directories of CBCT scans.

---

Installation

Prerequisites

• 3D Slicer 5.6+
• Python packages: torch, torchvision, nibabel, numpy, scipy, requests

The module checks and automatically installs these dependencies upon first usage.

Setup

Load the Module in 3D Slicer

• Open 3D Slicer.
• Navigate to Module → Slicer Automated Dental Tools → CLIC.

---

Usage

Step 1: Load CBCT Data

• Use the file dialog to select your CBCT image or directory containing multiple CBCT images.
• Available sample data for testing: MN138.nii , UM06.nii

Step 2: Download/Select Model

• Click “Download Model” if using for the first time, or specify your existing model directory.

Step 3: Configure your output

• Click “Choose save folder” if you want to a specific output folder or clic on save in input folder if you want the outputs in the same folder than the input
• write the suffix you want for th outputs

Step 3: Run Segmentation

• Click Predict to begin segmentation, if it s the first time you run a pop up will maybe ask you to install some dependency, clic on ‘yes’, the process will take a little time to start.
• Observe the progress bar and log output for real-time feedback.

Step 4: Visualize Results

• Segmentations automatically load into Slicer’s viewer with clear labeling.
• A color-coded legend (Buccal, Bicortical, Palatal) appears in slice views.
• Be carefull, if you have scans with differents field of view in your input folder and you try to navigate through the scans when th process is end the views can turn black, in this case you have to clic on the button reset field of view.

VFACE Module

VFACE (Visual Facial Asymmetry Classification Engine) is a 3D Slicer extension designed for automated classification of facial asymmetry. It provides advanced measurement extraction and machine learning-based classification of facial structures including the mandible and maxilla.

---

Key Features

• Automated Asymmetry Classification: Automatically classifies facial asymmetry patterns based on facial structure analysis.
• Facial Measurement Extraction: Extracts key measurements from facial structures for quantitative analysis.
• Machine Learning Based: Uses trained deep learning models for accurate asymmetry classification.
• User-Friendly Interface: Simple GUI integrated seamlessly within 3D Slicer for easy access and use.
• Batch Processing: Supports analysis of single files or batch processing of multiple datasets.

---

Installation

Prerequisites

• Required Python packages will be automatically installed upon first usage.

Setup

Load the Module in 3D Slicer

• Open 3D Slicer.
• Navigate to Module → Slicer Automated Dental Tools → VFACE.

---

Usage

The module provides an intuitive interface for facial asymmetry classification. Simply load your facial data, select the appropriate models and parameters, and run the analysis to obtain automated classification results and measurements.

For detailed usage instructions and tutorials, please refer to the module’s built-in help documentation.

GreedyReg Module

GreedyReg performs ITK-SNAP style registration of CBCT scans using the Greedy registration engine. It combines manual approximation tools with automatic registration, and can process a single pair of scans or a whole folder in batch.

Prerequisites

• The Greedy binary is required. If it is not found, the module shows a Download Greedy button that installs it automatically (~60MB).
• The Distant Registration mode additionally relies on the ALI-CBCT Python libraries. If they are missing, the module offers to install them.

How does the module work?

1. Input Volumes: select the fixed (T1) and moving (T2) CBCT scans to register.
2. Create Mask (optional): paint a region to focus the registration on a specific area.
3. Manual Alignment: roughly pre-align the moving scan using rotation/translation sliders and the manual approximation tools.
4. Automatic Registration: run the Greedy registration, save the registered volume, or use Run Batch Registration to process an entire folder of scans.
5. Distant Registration (Large Misalignment): when the two scans start far apart, this mode uses ALI-CBCT landmarks to first bring them close together before registering. It is also available in batch.

Agent Module

The Agent module is the graphical front-end of the AI Agent extension. Instead of opening each module by hand, you describe what you want to do with your dental/orthodontic imaging data in natural language, and the agent selects the right Automated Dental Tools module, extracts its parameters and runs it for you.

The agent runs fully locally: it is powered by the qwen3:8b model through Ollama, so no data leaves your machine.

Prerequisites

• Ollama must be installed.
• Click the Check button once to install the required Python dependencies and pull the qwen3:8b model. This step is only needed the first time.

How does the module work?

1. Describe your task: Type your request in plain language (e.g. “orient these CBCT scans” or “segment this folder”).
2. Add your data: Drag & drop one or more files or folders into the drop zone.
3. Choose a mode:
   • Agent (Automated): the agent picks the matching tool, fills in its parameters and runs it automatically.
   • Consultant: the agent recommends the most appropriate tool(s) for your request without running them.
4. Run: the agent routes your request to the corresponding module. If a tool execution fails, the agent automatically tries to fix the parameters from the error and retries (up to a few attempts).

The routing, parameter extraction and command building are handled by the internal Agent_CLI module.

Acknowledgements

Authors: Maxime Gillot (University of Michigan), Baptiste Baquero (UoM), Luc Anchling (UoM), Nathan Hutin (UoM), Alexandre Buisson (UNC), Paul Dumont (UNC), Enzo Tulissi (UoM), Alban Gaydamour (UoM), Jeanne Claret (UoM),Gaelle Leroux (UoM), Lucia Cevidanes (UoM), Juan Carlos Prieto (UNC), David Allemang (Kitware), Jean-Christophe Fillion-Robin (Kitware), Connor Bowley (Kitware), James Butler (Kitware).

Supported by NIDCR R01DE 024450, AA0F Grabber Family Teaching and Research Award and by Research Enhancement Award Activity 141 from the University of the Pacific, Arthur A. Dugoni School of Dentistry.

License

This software is licensed under the terms of the Apache Licence Version 2.0.