Fractures of the distal radius are a public health issue as they are the most common fractures. They can occur at any age, both in young active people (high-velocity trauma) and elderly people (height falls with osteoporotic bones, especially women over 50-years-old). Although the use of 3D imaging technologies improves the understanding of many joint fractures, they are still rarely used for the forearm1,2. Surgeons mainly use diagnosis and measurement methods only based on 2D X-rays with several limitations: the bony deformations are only visible in 3D and, X-rays do not reflect entirely the interrelationships between the ulna and the radius due to the complex movement of pronosupination in both distal and proximal parts. Moreover, the accuracy of morphometric measurements of interest on the radius (such as volar tilt, radial inclination or ulnar variance) highly depends on the position of the forearm during image acquisition. Any deviation from the standard protocol during image acquisition, e.g. due to the inability of the patient to hold the forearm in the predefined standard positions, affects the quality of these measures. In addition, diagnoses and presurgical planning are currently manually performed which is time-consuming and affects the reproducibility. To this end, a funded research project “PERForms” (ANR PRCE) gathering industrial, academic, and medical partners, have been recently launched. PERForms will develop a robust approach for building a digital twin of the patient's forearm able to simulate the premorbid anatomy and kinematics and help the surgical planning solutions. The Laboratory of Medical Information Processing LaTIM (INSERM U1101) will be in charge of the 3D shape-function modeling, allowing (1) the automatic segmentation of the addressed bones (ulna and radius), starting from CT-scans, which is often the first step in 3D modeling, (2) the prediction of the missing anatomy of the forearm due to a potential truncation by the scanner field of view (FOV) which is generally focused on the wrist and not the whole forearm in case of distal fractures and, (3) the prediction of the pronosupination movement, which is two-folds, to standardize the position of the forearm in anatomical position and, to understand the potential bony conflicts that could exist in the antebrachial framework. By exploiting learning approaches, the recruited person (postdoctoral position) will oversee the prediction of the missing anatomy and the pronosupination movement estimation.
Statistical shape models (SSM) provide a distribution of the morphology for a specific organ through a population. SSM have also the capability to predict missing parts thanks to the correlations embedded in the model. To date, they have already been used to successfully predict the 3D anatomy of the scapula3, proximal humerus4 and, the femur5,6 from a partial scanner field of view (FOV). Nevertheless, SSM have not yet been used to predict the missing part of the forearm from a distal part of ulna and radius. They have neither been used to estimate the pronosupination motion which is crucial to obtain robust morphological parameters from a standardized position. New statistical models incorporating motion have been recently proposed to describe a joint motion, particularly for the shoulder7. In this configuration, the joint model was able to predict the motion of the humerus and the scapula during the abduction motion7. The use of SSM for joint dynamic analysis has shown that the prediction of both the shape and position of bones could allow a joint standardization. In this postdoctoral position, Dynamic Multi-Object Gaussian Process Models7 that combines a SSM for shape and pose, will be investigated to predict the patient’s complete radius and ulna from their visible distal part alone in a standardized protocol. This SSM will be trained and validated using 3D clinical data that are already available for the project. The proposed approach will be finally integrated, in collaboration with engineers, inside our planning software to be used by surgeons.
The person will be hired by IMT Atlantique and hosted in the LaTIM (https://latim.univ-brest.fr/). Born from the complementarity between health and data science, the LaTIM laboratory develops multi-disciplinary research driven by members from IMT Atlantique, CHRU Brest, University of Western Brittany and Inserm. The recruited postdoc will work in collaboration with academic, industrial and hospital partners within the context of the ANR PERForms project. Access will be given to data from our clinical partners as well as to the PLaTIMed platform (https://platimed.fr/) to make realistic evaluation of the proposed approaches.
PhD in computer vision, AI, applied mathematics.
Good programming skills is an important requisite, especially in python and C++.
Autonomy, open-mindedness, and motivation.
Good English skills are also expected.
Cover letter and CV with list of publications, thesis reports (both reviewers and defense) and, a recommendation letter from the PhD supervisor, have to be sent to Guillaume Dardenne, INSERM researcher (guillaume.dardenne@inserm.fr) and Valérie Burdin, Professor at IMT Atlantique (valerie.burdin@imt-atlantique.fr).
The position is available as soon as possible for 20 months, possibly extendable.
The salary will depend on the candidate’s experience.
[1] Winter, R., Citarel, A., Chabrand, P,. Chenel, A., Bronsard, N., Poujade, T., Gauci, M. O. (2024). An Evaluation of the Reliability of Manual Landmark Identification on 3D Segmented Wrists. The Journal of Bone and Joint Surgery. 106(4), 315-322. doi: 10.2106/JBJS.23.00173
[2] Casari, F. A., Roner, S., Fürnstahl, P., Nagy, L., & Schweizer, A. (2021). Computer-assisted open reduction internal fixation of intraarticular radius fractures navigated with patient-specific instrumentation, a prospective case series. Archives of Orthopaedic and Trauma Surgery, 141, 1425-1432.
[3] Salhi A, Burdin V, Boutillon A, Brochard S, Mutsvangwa T, Borotikar B. (2020). Statistical Shape Modeling Approach to Predict Missing Scapular Bone. Ann Biomed Eng. 2020;48(1):367-379. doi:10.1007/s10439-019-02354-6
[4] Poltaretskyi S, Chaoui J, Mayya M, et al. (2017). Prediction of the pre-morbid 3D anatomy of the proximal humerus based on statistical shape modelling. The Bone & Joint Journal. 2017;99-B(7):927-933.
doi:10.1302/0301-620X.99B7.BJJ-2017-0014
[5] Asvadi A, Dardenne G, Troccaz J, Burdin V. (2021). Bone surface reconstruction and clinical features estimation from sparse landmarks and Statistical Shape Models: a feasibility study on the femur. Medical Engineering & Physics. 2021;95:30-38. doi:10.1016/j.medengphy.2021.07.005
[6] L. C. Ebert, D. Rahbani, M. Lüthi, M. J. Thali, A. M. Christensen, B. Fliss. (2022). Reconstruction of full femora from partial bone fragments for anthropological analyses using statistical shape modeling. Forensic Science International, doi:332.10.1016/j.forsciint.2022.111196
[7] Fouefack J.-R., Borotikar B., Lüthi M., Douglas T. S., Burdin V., Mutsvangwa T. E.M. (2023). Dynamic multi feature-class Gaussian process models. Medical Image Analysis, 85. doi.org/10.1016/j.media.2022.102730
(c) GdR IASIS - CNRS - 2024.