French
English
Lyon / San Franciso collaboration
The PhD will be carried out partly at the CREATIS laboratory in Lyon www.creatis.insa-lyon.fr and partly at Yan Li's laboratory at the University of California San Fancisco https://lilabimaging.ucsf.edu as part of an international thesis co-direction. This thesis project is co-funded by the graduate school medical device (https://graduate-plus.fr/en/medical-device-engineering-2/mde-call-for-project-phd-scholarshipp/) and UCSF.
Context:
Despite technical advancements, the evaluation of MRS spectra in clinical practice remains subjective, qualitative, and limited by a lack of robustness and automation. MRS has demonstrated its potential in elucidating key metabolic and neuronal pathways involved in brain diseases. However, current analytical methods rely on parametric model fitting, which is susceptible to modeling biases, with results varying depending on the quantification approach and parameterization. Given the growing emphasis on metabolic and genetic alterations in brain tumors, as highlighted by recent research and the WHO 2021 classification, a multimodal quantitative approach integrating MRS is increasingly relevant.
To address these limitations, we aim to leverage deep learning to enhance the objectivity, quantification, and automation of MRS analysis, thereby improving the characterization of primary brain lesions. Unlike traditional methods, deep learning can extract maximum information from real spectra databases without the constraints of simplified modeling. Our objective is to integrate routinely acquired clinical data—currently underutilized in quantitative analysis—into a comprehensive learning database. Additionally, we will expand a deep learning spectroscopy data library to facilitate pre-processing, quantification, and classification of primary brain tumor spectra. The analyzed and quantified data will then be correlated with clinical, histopathological—including immunohistochemistry—and radiological parameters, particularly diffusion/perfusion, to establish diagnostic probabilities based on MRS. However, the inherent challenge of lacking ground truth in real in vivo MRS acquisitions remains a constraint for supervised deep learning approaches. A challenge that needs to be adressed.
State of the art:
In the literature, machine learning has been explored and used for MRS data analysis, mainly for spectral classification (tumor and non-tumor spectra) and quality control. However, unlike medical image analysis, the analysis of in vivo MRS signals has only just begun to benefit from recent advances in deep learning [1], [2] [3], with problems very different from those encountered in high-resolution NMR used in chemistry or medical imaging.
In [1], a convolutional auto-encoder is trained on simulated data to reconstruct an artifact-free spectrogram from data containing artifacts. In [3], a CNN is trained to predict the evaluation of a spectrum (“good” or “poor” quality). For our part, we have been working on deep learning for data quantization and artifact removal. To our knowledge, we were the first to attempt to quantify MR spectra on synthetic data using a convolutional neural network [4]. As there is no ground truth for in vivo data, we used simulated data with known concentrations for training and testing. The simulations were based on the physical model that is generally used in standard quantification methods, which perform a parametric fit of a model to the acquired data. We also explored the UNET architecture for correcting signal artifacts [5] . However, this neural network architecture does not guarantee the preservation of metabolite content or physical properties. We believe that the constraints of generative networks to preserve certain physical properties as proposed in this project will be more suitable and powerful.
Project description
In vivo magnetic resonance spectroscopy (MRS) offers an unparalleled ability to provide non-invasive measurements of tumor metabolism, making it a valuable tool for diagnostic support, understanding pathophysiological processes, and optimizing patient treatment. As metabolism is fundamental to cellular and tissue function, characterizing chemical transformations through metabolite concentration measurements enhances neuro-oncological imaging by complementing anatomical, structural, and perfusion MRI. This multimodal approach improves lesion characterization in alignment with histo-molecular classification, optimizes biopsy targeting, and facilitates treatment monitoring. Despite significant technical advancements bringing MRS closer to clinical application, its integration into routine practice remains hindered by the subjective and qualitative nature of spectral interpretation.
This project aims to overcome these limitations by leveraging deep learning for the analysis of in vivo NMR spectra from tumor patients. We will explore state-of-the-art generative, regression, and classification models specifically adapted to MRS data. The project will build on a PyTorch-based library developed within a PNRIA initiative, which supports spectroscopy data, enables data augmentation, and facilitates the design of tailored architectures for classification, regression, and data generation. However, some key parameters—such as metabolite concentrations and macromolecule spectral behavior—lack a definitive "ground truth," making fully supervised learning impractical. Instead, we will explore unsupervised, weakly supervised, and semi-supervised approaches to address this challenge.
To mitigate overfitting risks associated with purely simulated data, we will develop unsupervised techniques based on generative score-based models, also known as diffusion models [6]. Additionally, we will implement hybrid AI methods that integrate various data sources: simulated spectra with known ground truth, real unannotated clinical data (from collaborator-led studies or publicly available datasets like “Big GABA”), and simulated data generated using classical NMR spectroscopy physical models. This work will build on advances in probabilistic scattering models, proposing novel ways to integrate physical properties into spectral inversion processes. Specifically, we aim to develop a probabilistic conditional diffusion model that, during its inverse process, incorporates targeted spectral properties such as tissue health status, water residue presence, or macromolecule expression.
Unlike standard computer vision problems, MRS analysis benefits from an existing, albeit incomplete, mathematical model based on NMR physics. It is therefore crucial to integrate this physical model into generative and analytical learning strategies, potentially leveraging domain adaptation techniques to bridge the gap between simulated and real data.
Expected results
The project is part of a drive to develop precision medicine, by improving information from in vivo MRS, the only technique to provide endogenous, non-ionizing, non-invasive information on tissue biochemistry. The techniques developed should 1) make it easier to exploit spectroscopy data on a routine basis, 2) make the quantitative data extracted more robust and reliable than existing methods, and 3) could help detect as yet uncharacterized metabolic or spectral profile modifications, which would provide a better understanding of the mechanisms underlying disease, and aid therapeutic choices.
Application
Background: applied mathematics, machine/deep learning or signal processing
Good software development skills (ideally with python/pytorch)
Taste for working in a highly multidisciplinary environment (deep learning, MRI physics, medical applications)
Applications (CV, transcript, recommendations,...) should be sent to michael.sdika[@]creatis.insa-lyon.fr and helene.ratiney[@]creatis.insa-lyon.fr
Bibliographie
[1] S. P. Kyathanahally, A. Döring, et R. Kreis, « Deep learning approaches for detection and removal of ghosting artifacts in MR spectroscopy », Magn Reson Med, vol. 80, no 3, p. 851 863, sept. 2018, doi: 10.1002/mrm.27096.
[2] S. S. Gurbani et al., « A convolutional neural network to filter artifacts in spectroscopic MRI », Magnetic Resonance in Medicine, vol. 80, no 5, p. 1765 1775, 2018, doi: 10.1002/mrm.27166.
[3] S. Vaziri et al., « Evaluation of deep learning models for quality control of MR spectra », Frontiers In Neuroscience, vol. 17, août 2023, doi: 10.3389/fnins.2023.1219343.
[4] N. Hatami, M. Sdika, et H. Ratiney, « Magnetic Resonance Spectroscopy Quantification Using Deep Learning », Medical Image Computing And Computer Assisted Intervention - MICCAI 2018, PT I. in Lecture Notes in Computer Science, vol. 11070. Springer, Switzerland, p. 467 475. doi: 10.1007/978-3-030-00928-1_53.
[5] N. Hatami, H. Ratiney, et M. Sdika, « MR spectroscopy artifact removal with U-Net convolutional neural network », in 27th Annual meeting of the ISMRM, Montreal, Canada, mai 2019. https://hal.archives-ouvertes.fr/hal-02129946
[6] J. Ho, A. Jain, et P. Abbeel, « Denoising Diffusion Probabilistic Models », in Advances in Neural Information Processing Systems, Curran Associates, Inc., 2020, p. 6840 6851. 6 mars 2023. https://proceedings.neurips.cc/paper/2020/hash/4c5bcfec8584af0d967f1ab10179ca4b-Abstract.html