Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

U-net based prediction of cerebrospinal fluid distribution and ventricular reflux grading

Melanie Rieff, Fabian Holzberger, Oksana Lapina, Geir Ringstad, Lars Magnus Valnes, Bogna Warsza, Kent-Andre Mardal, Per Kristian Eide, Barbara Wohlmuth

TL;DR

This study addresses predicting CSF tracer distribution from MRI to reduce the need for multiple scans. It applies a 2D U-net to baseline and early post-injection sagittal and axial MRI slices to forecast the 24-hour gadobutrol distribution and to assess ventricle reflux against neuroradiologist grading. The results show that training with data from the first 1–2 hours post-injection achieves comparable accuracy to models that incorporate later-stage scans, with low pixel-wise errors and favorable grading alignment, suggesting potential clinical efficiency gains. The work highlights the practical impact of data-driven CSF flow predictions in reducing imaging burden while preserving diagnostic utility, while noting limitations and outlining paths toward 3D volumetric modeling and integration of clinical data.

Abstract

Previous work indicates evidence that cerebrospinal fluid (CSF) plays a crucial role in brain waste clearance processes, and that altered flow patterns are associated with various diseases of the central nervous system. In this study, we investigate the potential of deep learning to predict the distribution in human brain of a gadolinium-based CSF contrast agent (tracer) administered intrathecal. For this, T1-weighted magnetic resonance imaging (MRI) scans taken at multiple time points before and after injection were utilized. We propose a U-net-based supervised learning model to predict pixel-wise signal increase at its peak after 24 hours. Performance is evaluated based on different tracer distribution stages provided during training, including predictions from baseline scans taken before injection. Our findings show that training with imaging data from only the first two hours post-injection yields tracer flow predictions comparable to models trained with additional later-stage scans. Validation against ventricular reflux gradings from neuroradiologists confirmed alignment with expert evaluations. These results demonstrate that deep learning-based methods for CSF flow prediction deserve more attention, as minimizing MR imaging without compromising clinical analysis could enhance efficiency, improve patient well-being, and lower healthcare costs.

U-net based prediction of cerebrospinal fluid distribution and ventricular reflux grading

TL;DR

This study addresses predicting CSF tracer distribution from MRI to reduce the need for multiple scans. It applies a 2D U-net to baseline and early post-injection sagittal and axial MRI slices to forecast the 24-hour gadobutrol distribution and to assess ventricle reflux against neuroradiologist grading. The results show that training with data from the first 1–2 hours post-injection achieves comparable accuracy to models that incorporate later-stage scans, with low pixel-wise errors and favorable grading alignment, suggesting potential clinical efficiency gains. The work highlights the practical impact of data-driven CSF flow predictions in reducing imaging burden while preserving diagnostic utility, while noting limitations and outlining paths toward 3D volumetric modeling and integration of clinical data.

Abstract

Previous work indicates evidence that cerebrospinal fluid (CSF) plays a crucial role in brain waste clearance processes, and that altered flow patterns are associated with various diseases of the central nervous system. In this study, we investigate the potential of deep learning to predict the distribution in human brain of a gadolinium-based CSF contrast agent (tracer) administered intrathecal. For this, T1-weighted magnetic resonance imaging (MRI) scans taken at multiple time points before and after injection were utilized. We propose a U-net-based supervised learning model to predict pixel-wise signal increase at its peak after 24 hours. Performance is evaluated based on different tracer distribution stages provided during training, including predictions from baseline scans taken before injection. Our findings show that training with imaging data from only the first two hours post-injection yields tracer flow predictions comparable to models trained with additional later-stage scans. Validation against ventricular reflux gradings from neuroradiologists confirmed alignment with expert evaluations. These results demonstrate that deep learning-based methods for CSF flow prediction deserve more attention, as minimizing MR imaging without compromising clinical analysis could enhance efficiency, improve patient well-being, and lower healthcare costs.
Paper Structure (13 sections, 1 equation, 7 figures, 2 tables)

This paper contains 13 sections, 1 equation, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Materials and Methods
  4. Study population
  5. Injection protocol, MRI hardware, and data preprocessing
  6. Distribution pattern
  7. Model architecture
  8. Ventricular reflux grading
  9. Experimental Results
  10. Image prediction from baseline and tracer-enriched scans
  11. Ventricular reflux grading performance compared with neuroradiologists' assessments
  12. Discussion
  13. Loss curves for U-net training and testing over 250 epochs ($\ell_2$ loss)

Figures (7)

  • Figure 1: Defaced sagittal slices from CSF tracer–enhanced MRI of a sample patient. After 24 hours, the tracer has enriched the CSF spaces around the entire brain and is completely cleared after four weeks.
  • Figure 2: Exemplary illustration of the training and testing steps. Here, the network was trained by minimizing the $\ell_2$-loss of the reconstructed sagittal images to the real images and final model evaluation made use of the mean of all squared errors among the testing data.
  • Figure 3: \ref{['fig:UNET']}: Modified U-net architecture used in this work. Here, axial and sagittal MRI planes (one each) scanned before tracer injection are used to predict its distribution 24 hours after tracer injection. \ref{['fig:processing_workflow']}: Processing workflow from original MRI scans to U-Net model predictions, comparison with real MRI images, and final clinical diagnosis.
  • Figure 4: A sample test case (sagittal and axial plane) of gadobutrol distribution prediction based from baseline MRI scans (pre-injection). \ref{['inpt']}: real MR imaging taken before injection. \ref{['tgt']}: real MR imaging taken approximately 24 hours after intrathecal tracer injection. \ref{['mse']}: predicted tracer distribution 24 hours post-injection using an $\ell_2$ loss function. \ref{['mse_diff']}: absolute difference between \ref{['tgt']} and \ref{['mse']}. \ref{['mae']}: predicted tracer distribution using an $\ell_1$loss function. \ref{['mae_diff']}: absolute difference between \ref{['tgt']} and \ref{['mae']}.
  • Figure 5: Two sample test cases (sagittal and axial planes). Rows 1 and 2 display to the first test case, while rows 3 and 4 display the second. Each column corresponds to a different set of MR images (artificial or real). \ref{['inpt_2']}: real MR imaging taken within 1-2 hours post-injection. \ref{['tgt_2']}: real MR imaging taken approximately 24 hours after intrathecal tracer injection. \ref{['mse_2']}: predicted tracer distribution 24 hours post-injection ($\ell_2$ loss). \ref{['mse_2_diff']}: absolute difference between \ref{['tgt_2']} and \ref{['mse_2']}. \ref{['mae_2']}: predicted tracer distribution 24 hours post-injection ($\ell_1$loss). \ref{['mae_2_diff']}: absolute difference between \ref{['tgt_2']} and \ref{['mae_2']}.
  • ...and 2 more figures