Resources

1. Lab website

Continued development of the lab's new website, in particular improving the section on computer workstation assignments.

2. Lab charter

The lab charter was reviewed in detail and approved. The new version will be put forward for signature by all lab members.

3. PEPR and ERC projects

Discussion of Maxime Taverne's PEPR and ERC projects, which need to be articulated with the Totem project, currently under discussion with Matthieu Ricci and Jean-Baptiste Masson (Institut Pasteur).

4. Attendance table

Discussion of the importance of including an attendance/absence table on the lab's website, to make it easier to schedule appointments.

5. Declarations to the AP-HP's DPO

Discussion of the need to re-declare the lab's various ongoing studies to the AP-HP's Data Protection Officer (DPO), in order to regularize all the lab's health-data-reuse procedures.

Attendees:

• For Helsinki: Junnu Leikola, Miika Tovairi, and their teams
• For Necker / Institut Imagine: Roman Hossein Khonsari

1. (Semi-)automatic bone segmentation from MRI

• Already explored by the Helsinki team, with their engineer Petra Laakkonen (petra.laakkonen@hus.fi).
• First results available: good overall rendering of the cranial volume, but surface resolution still insufficient.
• Ideas raised: additional training data, with specific training on the suture region; possible combination of a statistical model and a generative AI model.
• To do: re-establish the link with Adèle Rohée-Traoré's PhD project, which involves IMT Atlantique (Brest campus) and Laurent Gajny (Arts et Métiers).
• Avenue to explore: MRI-based 3D reconstruction might, by its nature, also allow 3D reconstruction of cartilage — which is missing from CT data.

2. Study of FGFR patients' brains (brain imaging)

• This project follows on from Ombline Delassus's PhD thesis (defended in 2025), in connection with NeuroSpin, David Germanaud and Jean-François Mangin.
• It requires setting up a data transfer between Helsinki Children's Hospital and the laboratory (Necker / Institut Imagine).
• This transfer will be easier once the MoU currently being finalised with Helsinki Children's Hospital (contact: Junnu Leikola) is signed.
• Following this MoU, the possibility of recruiting a PhD student funded by Helsinki, who would work at NeuroSpin on the continuation of this thesis, could be considered.
• This point is to be discussed further with David Germanaud and Jean-François Mangin in the autumn.

Attendees: Julia Boiko, Tetyana Pavlychuk and Roman Hossein Khonsari.

1. Establishing a University Diploma (DU) in microsurgery at Bogomolets University

Proposed strategy:

• Visit of a Ukrainian delegation to the École de chirurgie (Paris), following discussions with Professor Marc Fitoussi and Dr Elliot Pattou, in order to define an action plan and a possible participation of this delegation in a microsurgery training course (from three days to one week).
• On the delegation's return to Ukraine, setting up a microsurgery platform within the Bogomolets University simulation center. The partnership with Enabel (the Belgian agency equivalent to Expertise France) has already funded the purchase of six "watchmaking"-type microscopes as well as six microsurgery sets (kits).
• The simulation center has an animal facility able to house rats. The delegation's visit to the École de chirurgie will also be an opportunity to observe the logistics of using these rats (housing and usage flow).
• Building the DU: to be explored with Université Paris Cité or Sorbonne Université, to set up a DU jointly delivered by Bogomolets University and the French university. The idea is to create a diploma for around ten Ukrainian candidates per year, open to maxillofacial surgeons, orthopedic surgeons, neurosurgeons, plastic surgeons and any other specialist interested in microsurgical reconstruction.
• Funding: the establishment of the DU and, initially, the Ukrainian delegation's stay in Paris could possibly be funded by Expertise France.

2. Developing 3D printing in connection with simulation

• A first project could be setting up a system for printing models used in microsurgery, in partnership with Lynxter and Thomas Batigne, CEO of Lynxter.
• This partnership is being discussed with Delphine Prieur and Juliette Prébot, as well as with Professor Andriy Kopchak's team.
• The development of this 3D printing platform in Kyiv is to be linked to the Superhumans Center project in Kyiv, which is likely to open within the next 24 months.

3. Creating a medical museum at Bogomolets University

• The person in charge is Andriy Iantchichin, an anatomist, currently responsible for the existing museum at the academy.
• Anatomical specimens and various historical objects are currently being moved from the university / simulation center to this anatomy laboratory.
• It would be very valuable to link this project with the French project for a Cité nationale de la santé et du soin (National Centre for Health and Care), currently being structured in Paris — with the Hôtel-Dieu as its envisaged site — and with its heritage dimension led by Assistance Publique – Hôpitaux de Paris (AP-HP). The project's multidisciplinary support committee — physicians, researchers, historians and philosophers of science, artists, heritage specialists —, backed by the École normale supérieure, Université Paris Cité and Sorbonne Université as well as the Academies of Sciences, Medicine and Surgery, could be mobilized to work, together with the Ukrainian team, on the curation of this museum, which holds several exceptional pieces.

Next steps

1. Meeting with the École de chirurgie team in Paris and the Bogomolets team: scheduled for September.
2. Visit of the Bogomolets team to the École de chirurgie to explore prospects for developing the DU: to be organized by December.
3. Meeting with the 3D printing team and the simulation center team (with Thomas Batigne) to discuss the feasibility of the silicone printing partnership with Lynxter.
4. Meeting with Andriy Iantchichin and some members of the medical museum reflection group to discuss the possible structuring of a collaboration.

Attendees:

• Victoria Franchetti-Romero
• John Baxter
• Maxime Taverne

Logistics

• Technical issues with Microsoft Teams: future recurring meetings move to Google Meet.
• Victoria signed her working contract.

Research progress: literature review

Overview of recent readings on the relationship between soft and hard tissues in craniofacial growth.

Ontology and symbolic logic development

Presentation of a preliminary ontology focused on the eye-orbit functional unit, defining anatomical units and initial symbolic logic rules.

Feedback from John Baxter: refine the rules by distinguishing "mechanical rules" from "observations/results" (conclusions the system should infer), and break the rules down as much as possible.

PhD committee (CSI)

Decision to remove the Calgary-based candidate to avoid administrative flags over following the research across time zones; this person could still be asked to serve as a juror for the final thesis defense.

Next meeting

19 February, 3:00 PM French time — shifted due to a 9-hour time difference with John Baxter, who is attending a conference.

Goals for next week

• Separate existing rules into "mechanical rules" (expert system) and "results/observations" (to be logically inferred).
• Incorporate the notion of pressure into the rules (whether volume drives pressure or vice versa, depending on the anatomical entity).
• Formalize the ontology structure (Region → Anatomy type → Specific name), with classes for bones (orbit, neurocranium, skull base, etc.).
• For each anatomical entity, identify measurable properties (volume, pressure, velocity), continuous or categorical.
• Add a category for non-anatomical variables (age).
• Fully formalize the orbital region first, as a template for other modules.

Attendees:

• Victoria Franchetti-Romero
• Olivier Dameron
• Maxime Taverne

OCDM ontology analysis

Olivier Dameron shared his conclusions after exploring the OCDM ontology (Ontology of Craniofacial Development and Malformation): the granularity and level of detail of the relations are of excellent quality, but the description remains purely symbolic, with a notable lack of quantitative aspects and topological data. This is a very positive and solid working base for the rest of the project.

Methodological avenues

Two levels of ontology use were proposed: a descriptive level (repository, simplified use — definitions, bone/muscle hierarchies, basic relations) and an analytical level (reasoning, more complex use for automated inference). An incremental approach is planned, first focusing on a restricted area, the cranial vault, to validate the method.

Ontology/SEM (Structural Equation Modeling) synergy: Maxime Taverne proposes exploring a parallel between ontologies and SEM, using a network of equations to link variables and test model fit against data, in order to extract quantities and structural relations that would feed numeric data into the ontology.

Link with medical imaging

The challenge is to build a bridge between the symbolic world (ontology) and the numeric world (images): labeling segmentations with ontology terms, identifying automatically extractable morphometric features, whose values would fill the quantitative attributes currently missing from OCDM. Adding new structures (such as the brain) is considered a moderate amount of work.

To do

• Extract cranial vault growth rules from the literature and translate them via OCDM's concepts/attributes.
• Research symbolic modeling methods applied to other structures (scoliosis, spine, locomotion).
• Imaging: identify automatically extractable features (dimensions, lengths) usable in the rules, and inventory the numeric data currently available.

Methodological avenues and reflections

Discussion of the possible first contribution: validate all three layers at once, validate each layer one by one, or validate the first two layers together? It was clarified that "validating" the semantic/symbolic layers (the rules) does not mean judging whether a theory is true or false, but validating the knowledge-extraction methodology — for example via a questionnaire checking that rules were correctly formulated, a parallel extraction by a second person, or detecting redundancies and inconsistencies. An approach combining human expertise with automatic/logical techniques is being considered. Proposed steps: validating the knowledge-extraction method; building the "rule engine" and getting it to reason; combining the FEM Toy with the rule engine; in the long term, turning the FEM Toy into a patient-specific model.

An alternative avenue was raised: rather than going from rules to numbers, start from longitudinal patient data to derive the rules (numeric → rules) — a possible additional contribution. For the patient-specific model, two options: build a model from a segmentation, or adapt the FEM Toy ("Sphere brain") to the patient (e.g. closing a suture that is already closed in that patient).

Validating the knowledge-extraction method

How to determine how many rules to include before moving to validation? The model will represent a percentage of the knowledge corpus (the rules extracted from a set of articles), with no minimum required; for articles, stop at saturation (new articles no longer add anything). Weighting the rules by importance is being considered, for example by how often a rule appears in the corpus or by expert consensus gathered through a questionnaire (a Delphi-inspired approach).

Next meeting

15 July, 10:00 AM.

To do

• Read more articles and extract rules from them to complete the table.
• Find a way to categorize these rules.
• Summarize the table (OCDM elements present, to add, etc.), as it will become unreadable in its current form.

Prospective experts

Discussion of the status of the experts approached: working group or co-authors? Current list:

• Pierre Corre (France, maxillofacial surgeon)
• Jean-Michel Salagnac (France, orthodontist and dento-facial orthopedist)
• Jean-Philippe Perrin (France, maxillofacial surgeon)
• Jean-Claude Talmant (France, plastic and maxillofacial surgeon)
• Costanza Meazzini (Italy, orthodontist and cranio-maxillofacial surgery specialist)
• Barry Grayson (USA, craniofacial surgeon / orthodontics)
• Eppo Wolvius (Netherlands, maxillofacial surgeon)
• Hanna Lif (Sweden, plastic and craniofacial surgeon)
• Antoine Balzeau (France, paleoanthropologist)
• Joan Richtsmeier (USA, anthropologist)
• Markus Bastir (Spain, paleoanthropologist)
• Benedikt Hallgrímsson (Canada, biologist/anthropologist)
• Yann Heuzé (France, biologist/anthropologist)

Ideas for completing the list: define selection criteria; aim for at least 20 people to anticipate absences and dropouts in the second round; add Asian experts; address the under-representation of women. Names considered: Giovanna Paternoster, Jesse Taylor, Yang Chai.

Sending invitations

Send invitations by email as soon as possible, to circulate the questionnaire from September, giving a few rule examples, an estimate of the time required, and noting that participants who complete the process will be co-authors of the publication.

Second round

Discussion of how to phrase the second-round questions (avoid free text, use votes only): the approach draws inspiration from the Delphi method without strictly being one — it stops at the second round because the goal is not consensus but gathering enough relevant rules to start a model, deliberately staying limited in scope.

Rule formalism

Two options: leave free text in the questionnaire for new rules and reformulate them afterward using a predefined formalism, or have the experts adopt that formalism directly.

The coming months

• Finish building the knowledge corpus.
• Formulate all the rules from this corpus.
• Send the questionnaire to the experts.
• Once the rules are obtained, check their full coverage in the ontology.
• Use the rules in an expert system able to analyze the rule base (dependencies, precedence...).

Some rules can start being implemented as soon as the first results come in, without waiting for both questionnaire rounds to finish.

Next meeting

2 September, 11:30 AM.

1. Expert questionnaire — organization

Validate all meeting dates with Roman Hossein Khonsari. Communicate an official schedule to the experts, while keeping an informal internal schedule in parallel to manage late replies. For the 6 October date: Pierre Corre is more available on the 5th or 7th — to be adjusted, prioritizing Roman Hossein Khonsari's availability.

2. Analysis of the expert list

Question to dig into: does the panel's gender balance reflect the reality of the field? Check whether reference figures exist, otherwise build a statistic from the authors of the selected articles.

3. Meetings with the experts

Roman Hossein Khonsari leads the meetings (introduction and conclusion), handing over to Victoria Franchetti-Romero for her presentation. Record the meetings for those absent, with additional availability to answer questions, or individual meetings if needed. Schedule meetings around 4–5 PM to better match US/Canada time zones.

4. Empirical / quantitative rules

Set aside for now, to be addressed at implementation time. Open question: integrate them into the OCDM ontology or into the expert system? Olivier Dameron advises against integrating them into the ontology.

5. OCDM

State of the analysis (Olivier Dameron): topological relations are missing from OCDM (e.g. which bone connects to which suture), particularly adjacency relations — an addition judged feasible without a major workload. The calvarium is described more completely than the basicranium from a topological standpoint. The current binary bone-suture relations don't allow direct extraction of bone-bone topological relations; two alternative modeling approaches were proposed.

To do: build a list of bones of interest and their neighbors needed for the rules, to check their coverage in OCDM; fill in the table crossing anatomical entities with rules; make sure OCDM broadly covers the anatomy and topological relations needed.

Possible contribution (following a discussion with Olivier Dameron): assess to what extent the needed model can be generated from OCDM, in three steps — extract the anatomical structures and their relations from OCDM; generalize where needed to a coarser grain than described in OCDM; identify what's missing from the model, then choose between completing the model directly or completing OCDM and re-running inference. Main contribution envisioned: a set of inference rules for these relations; secondary contributions: applying them to the structures of interest, and the corrections/additions made to OCDM.

Next meeting

20 October, 2:00 PM (focused on anatomy / OCDM).

To do

• Validate meeting dates with Roman Hossein Khonsari.
• Send the email to the experts.
• Prepare the slides for the meeting with the experts.
• Fill in the rule-anatomical-structure table (list bones of interest and their neighbors, check their coverage in OCDM).
• Start thinking about the expert system / rule engine.

1. Spirit of the charter

Our lab works with data drawn from the care of children and families. This raw material is a responsibility before it is a resource. This charter sets out the shared rules of conduct: data protection, reporting back results, use of funding, bibliography management, publications, and communication. It applies to everyone, at all times; when in doubt, ask the lab director before acting.

This charter is presented to every lab member — new and existing alike — and signed by each one: by new members on arrival, and retrospectively by members already in post.

2. Clinical data and good conduct

Clinical data (records, imaging, photographs, measurements) never leaves the setting for which it was collected. Three principles:

• Minimization: only the data strictly necessary for the study's objective is collected and used.
• Pseudonymization at source: all data is separated from the patient's identity as soon as it is extracted; no name, no identifying initial, no full date of birth circulates in working files, presentations, or publications.
• Traceability: every study is declared, tied to a methodological reference framework and to an identified lead before any data is collected.

Facial photographs are particularly identifying data: they receive heightened vigilance (secure storage, blurring for any distribution, consent specific to image use).

🔒 Storage, extraction, anonymization, and declaration of data

• Storage (AP-HP NAS). Identifying clinical data is stored on the AP-HP's NAS, accessible only to physicians. It is never kept on a personal drive, a USB key, an unencrypted laptop, or a consumer cloud service (Google Drive, Dropbox, etc.).
• Extraction and download. This clinical data may only be extracted from the NAS as part of MR-004-type retrospective studies, in anonymized form, for use on the lab's research computers. Downloading is done from the hospital's PACS, with a systematic check that the patient and the exam match before saving. Inter-center exchanges go through secure repositories and anonymized data, governed by a data-sharing agreement.
• Anonymization. Each exam is anonymized right at export (via MRIcron for MRIs, DICOM Anonymizer for CT scans) and named using the lab's anonymization key (IPP-FCC), which makes the patient unidentifiable outside the lab's members. The conversion method itself is never disclosed.
• Storage and archiving. Downloaded and anonymized data is stored on the NAS, organized by disease and then by imaging type. In shared files, the original identifier (IPP) and the anonymized identifier (IPP-FCC) always appear in two separate columns; for any file leaving the lab, only the IPP-FCC is kept.
• MR-004. The CNIL's MR-004 reference methodology governs research that does not directly involve human subjects (retrospective studies, reuse of care data). It requires: a study entered in the processing register, informing patients and families with the option to object, a clearly defined purpose, and compliance with retention periods. Any study outside this framework (prospective research, a trial, collection of new data) falls under a different regime (CPP ethics committee, dedicated consent).
• Mandatory declaration. Every retrospective study using data acquired through patient care must be declared to the AP-HP's Data Protection Officer (DPO), Philippe Tourenne, before it begins.

Every lab member handling medical data individually commits to never disclosing patient-record access codes or the anonymization method, to never storing non-anonymized data on personal equipment, and to never retaining medical data beyond the end of their engagement with the lab.

3. Reporting back and sharing results

Reporting results back is part of research, not an afterthought.

• Results are shared internally first (lab meetings) before any external release.
• Any data shared externally (a collaborator, a conference, a funder) is pseudonymized and validated by the study lead.
• Results are reported back to the clinical teams that made the study possible, and, where relevant, to patient associations and funders, in accessible language.
• Preliminary results are never presented as if established; exploratory findings are clearly distinguished from confirmed ones.

4. Use of funding

The lab's funding comes from foundations, public bodies, and partners (Fondation Les Gueules Cassées, Fondation Princesse Grace, INSERM, City of Paris, PEPR, university partnerships, etc.). It is committed to research and training.

• Collegial decision-making: any use of funding is discussed at a lab meeting, as is the recruitment of any new student.
• Acknowledgment: every funding source is credited in the corresponding publications and presentations (see §6).
• Travel expenses. The lab covers: second-class (train) or economy-class (plane) transport costs; hotel costs at establishments up to three stars. It does not cover daily expenses (meals and other day-to-day costs on site).

5. The Marie-Paule-Vazquez library

The lab has its own library, the Marie-Paule-Vazquez library, which is part of it and is presented on the lab's website (§7).

Every useful reference is added to the shared library, correctly documented (DOI, complete metadata).

6. Scientific publications

• Authorship: author order and status follow ICMJE criteria (substantial contribution to design or analysis, drafting or critical revision, final approval, accountability). Each person's contribution is discussed early, not at the end.
• Affiliations: every publication carries the lab's complete, harmonized affiliation: Laboratoire « Forme et croissance du crâne », Institut Imagine, Université Paris Cité, Paris, France; Department of Maxillofacial and Plastic Surgery, Necker – Enfants Malades Hospital, AP-HP, Paris, France.
• Funding and conflicts of interest: systematically stated in the dedicated section.

7. The lab's website

The lab has a website with two parts: a section open to everyone (presenting the lab, the team, its work, publications, the Marie-Paule-Vazquez library) and a password-protected private section, reserved for the lab's internal life.

The following decisions are discussed at a lab meeting, then implemented on the lab's website (in its private section):

• computer workstation assignments;
• IT equipment purchases and requests for equipment or resources;
• internship and master's position postings.

The private section thus brings together the elements that matter for the lab's day-to-day life — including meeting minutes (§8) and internal resources.

8. Communication and social media

The lab communicates about its work (website, LinkedIn, Instagram, press). This communication involves the whole team.

• Consistent affiliations: any public communication names the lab and its parent institutions consistently (the same wording as in publications), and points back to the lab's website as the reference source. For social media posts (LinkedIn, Instagram, etc.) related to the lab, the following entities are named — and tagged/mentioned where such accounts exist on the platform used: the lab's name, the lab director's name, Institut Imagine, Université Paris Cité, Assistance Publique – Hôpitaux de Paris (AP-HP), Necker – Enfants Malades Hospital, Filière TêteCou, CRMR MAFACE, CRMR CRANIOST, and Fondation Princesse Grace.
• No patient data: never a photograph, image, or identifying information about a patient on social media, even blurred, without explicit written consent specifically covering that use. When in doubt, refrain.
• Validation: any post made in the lab's name (LinkedIn, Instagram, website) is validated by the director or a designated person; personal accounts that mention the lab respect its image and rules.
• Tone: accurate popularization, without overselling or promising therapeutic outcomes; research findings are clearly distinguished from clinical application.
• Credit: people, collaborators, and funders involved are credited.

9. Lab life and meeting minutes

Respect, kindness, and mutual support are the rule. Everyone contributes to meetings, to training younger members, and to keeping shared spaces and tools in good order.

Systematic minutes. Every scientific event held as part of the lab and/or funded by the lab is written up, and regularly uploaded to the lab's website (private section, §7). This includes in particular: thesis lunches, PhD follow-up meetings, scientific lunches, scientific travel, and seminars held as part of specific lab projects, etc.

Minutes should be brief, listing the names of those present and the topics discussed. This record is the lab's memory: it applies to any gathering that involves its name or its resources.

"By joining the lab, I acknowledge having read this charter and commit to respecting its principles."