months Post-Doc position (possibly extendable) in language processing for the automatic detection of speech disorders during awake brain surgery

  • Brest, Finistère
  • CDD
  • Temps-plein
  • Il y a 15 jours
Offer DescriptionScientific contextLow-grade gliomas (LGG, WHO grade II) are infiltrative brain tumors that generally occur in young patients (average age 38 years). Due to their slow growth and the plasticity of the brain, most patients have a normal or almost normal clinical examination for several years. Surgery is considered the best treatment for LGG due to its impact on the time to anaplastic transformation and improved survival rate. When resecting LGGs located in eloquent areas, intraoperative mapping of electrical stimulation under awake conditions has been shown to be essential for localizing the patient's neurological functions, as it allows the surgeon to maximize the extent of resection while reducing the risk of permanent deficits [1, 2]. The localization of these functions is based on neuropsychological tests that are carried out just after cortical and subcortical stimulations during surgery. The awake patient is thus invited to perform different exercises depending on the areas stimulated. Electrical stimulation causes a reversible lesion in the targeted area and allows medical staff to analyze the patient's response by identifying possible neuropsychological disorders linked to these areas. Therefore, if a deficit occurs in the patient's response, the stimulated area is labeled as eloquent and is preserved. Since 2015, every Awake Craniotomies (AC) performed at the Tokyo Women's Medical University Hospital are recorded using the Intraoperative Examination Monitor for Awake Surgery (IEMAS) [3] which allows the acquisition of the patient's face and voice, the task currently administered and the video of the patient's brain, indicating when and where the surgeon is performing a stimulation. Since May 2023, every AC from the University Hospital of Brest are also recorded using the software presented in [4] to initiate the constitution of a French dedicated awake brain surgery database including vital signs, stimulation parameters and the patient's speech during exercises [4]. With the objective of optimizing tumor resection, our goal is to develop algorithms and digital solutions to further help the medical team by better detecting and identifying patient's intraoperative deficits following stimulation, starting with language for AC.MissionNumerous works on the detection of speech disorders have been published in the literature. They mainly concern stuttering [5], dysarthria [6] and Parkinson's disease [7]. Except for the work of Nishimura et al [8] carried out by the “Center for Advanced Biomedical Sciences” (TWIns), a joint research facility between Tokyo Women's Medical University and Waseda University, no research has been focused on speech disorders caused by intraoperative stimulation. In addition, all this work was carried out using English-language databases only. Based on the joint experiences between the LaTIM, LTSI and the Twins teams, our objective will be to analyze the correlations between speech alteration and the exact location of the stimulation. For that, the goal of this post-Doc position will be to: * Extend the intraoperative monitoring solutions in order to also record the position of the stimulation probe relative to the brain, and in particular the glioma, as described in [9]. Information about the glioma itself will be extracted from preoperative imaging data through existing machine learning-based tumor segmentation [10] and deformable registration to a standardized brain model.
  • Improve the accuracy of a first innovative approach based on deep learning for the detection of language deficits using a French and Japanese cross-language database [11]. A synthetic database will be created using existing approaches to potentially compensate for the lack of data [12] and Wave2Vec2-based Deep-Learning models will be investigated because of their efficiency and performances in other clinical contexts [13]. The improved method will be finally evaluated on the cross-language database and in a real awake neurosurgery context.
  • Highlight the correlations between speech alteration and the exact location of the stimulation. Specific registration approaches will be used to propagate the patient's brain as well as their tumor on a latent space allowing the homogeneous representation of the stimulations in relation to the tumor and thus being able to create a functional or pathological anatomical atlas.
EnvironmentThis post-doc position will be mainly hosted in the LaTIM (Brest, France). 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. But several mobilities will also be done either to the team of the LTSI lab (Rennes, France) or/and to the joint research facility (Tokyo, Japan).ProfilePhD in language processing, AI, applied mathematics.Excellent programming, especially in python and C++.Good English skills.High motivation for publications and for international mobilities.ApplicationCV with list of publications, cover letter and two letters of recommendation, must be sent to Guillaume Dardenne, LaTIM ( )The position is available as soon as possible for 18 months, contract possibly renewable.The salary will depend on the candidate's experience.References[1] Duffau, H. (2012). The challenge to remove diffuse low-grade gliomas while preserving brain functions. Acta neurochirurgica, 154, 569-574.[2] Soffietti, R., Baumert, B. G., Bello, L., Von Deimling, A., Duffau, H., Frénay, M., ... & Wick, W. (2010). Guidelines on management of low-grade gliomas: report of an EFNS-EANO* Task Force. European journal of neurology, 17(9), 1124-1133.[3] Yoshimitsu, K., Suzuki, T., Muragaki, Y., Chernov, M., & Iseki, H. (2010, August). Development of modified intraoperative examination monitor for awake surgery (IEMAS) system for awake craniotomy during brain tumor resection. In 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology (pp. 6050-6053). IEEE.[4] Maoudj, I., Garraud, C., Panheleux, C., Saliou, V., Seizeur, R., & Dardenne, G. (2023, July). A modular system for the synchronized multimodal data acquisition during Awake Surgery: towards the emergence of a dedicated clinical database. In 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) (pp. 1-4). IEEE.[5] Sheikh, S. A., Sahidullah, M., Hirsch, F., & Ouni, S. (2022). Machine learning for stuttering identification: Review, challenges and future directions. Neurocomputing.[6] Sekhar, S. M., Kashyap, G., Bhansali, A., & Singh, K. (2022). Dysarthric-speech detection using transfer learning with convolutional neural networks. ICT Express, 8(1), 61-64.[7] Laila, R., Salwa, L., & Mohammed, R. (2021, April). Detection of voice impairment for parkinson's disease using machine learning tools. In 2020 10th International Symposium on Signal, Image, Video and Communications (ISIVC) (pp. 1-6). IEEE.[8] Nishimura, T., Nagao, T., Iseki, H., Muragaki, Y., Tamura, M., & Minami, S. (2014, November). Classification of patient's reaction in language assessment during awake craniotomy. In 2014 IEEE 7th International Workshop on Computational Intelligence and Applications (IWCIA) (pp. 207-212). IEEE.[9] Tokuda, J., Fischer, G. S., Papademetris, X., Yaniv, Z., Ibanez, L., Cheng, P., ... & Hata, N. (2009). OpenIGTLink: an open network protocol for image-guided therapy environment. The International Journal of Medical Robotics and Computer Assisted Surgery, 5(4), 423-434.[10] Baid, U., Ghodasara, S., Mohan, S., Bilello, M., Calabrese, E., Colak, E., ... & Bakas, S. (2021). The rsna-asnr-miccai brats 2021 benchmark on brain tumor segmentation and radiogenomic classification. arXiv preprint arXiv:2107.02314.[11] Maoudj, I., Kuwano, A., Panheleux, C., Kubota, Y., Kawamata, T., Muragaki, Y., ... Dardenne, G., Tamura, M. (2024). Classification of Speech Arrests and Speech Impairments during Awake Craniotomy: a multi-databases analysis.[12] Kourkounakis, T., Hajavi, A., & Etemad, A. (2020). FluentNet: end-to-end detection of speech disfluency with deep learning. arXiv preprint arXiv:2009.11394.[13] Sheikh, S. A., Sahidullah, M., Hirsch, F., & Ouni, S. (2022). Machine learning for stuttering identification: Review, challenges and future directions. Neurocomputing, 514, 385-402.RequirementsResearch Field Computer science » Programming Education Level PhD or equivalentSkills/QualificationsPhD in language processing, AI, applied mathematics.Excellent programming, especially in python and C++.Good English skills.High motivation for publications and for international mobilities.Languages ENGLISH Level GoodResearch Field Computer science » Programming Years of Research Experience 1 - 4Additional InformationWork Location(s)Number of offers available 2 Company/Institute LaTIM Country France State/Province Brittany City Brest Postal Code 29200 GeofieldWhere to apply E-mailguillaume.dardenne@inserm.frContact State/ProvinceBrittany CityBrest WebsiteStreetLaTIM, Faculté de Médecine, 22 Avenue Camille Desmoulins, Bâtiment IBRBS, étage 1 Postal Code29238STATUS: EXPIRED

EURAXESS