Estimating Hippocampal Shrinkage from MR Angiography
Poster No:
1643
Submission Type:
Abstract Submission
Authors:
benjamin thyreau1, Yasuko Tatewaki1, Tetsuya Maeda2, Kenji Nakashima3, Jun-ichi Iga4, Jun Hata5, Toshiharu Ninomiya5, Yasuyuki Taki6,1
Institutions:
1Institute of Development, Aging and Cancer, Tohoku University, Sendai, Japan, 2Iwate Medical University, Iwate, Japan, 3National Hospital Organization, Matsue Medical Center, Matsue, Japan, 4Ehime University Graduate School of Medicine, Matsuyama, Japan, 5Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan, 6Smart-Aging Research Center, Tohoku University, Sendai, Japan
First Author:
Co-Author(s):
Yasuyuki Taki
Smart-Aging Research Center, Tohoku University|Institute of Development, Aging and Cancer, Tohoku University
Sendai, Japan|Sendai, Japan
Smart-Aging Research Center, Tohoku University|Institute of Development, Aging and Cancer, Tohoku University
Sendai, Japan|Sendai, Japan
Introduction:
The hippocampus is a critical biomarker for neurodegenerative diseases, and thus, its segmentation from MR images enables to perform brain atrophy quantification. While hippocampal segmentation is commonly performed using 3D T1-weighted MRI, or a combination of T1 and higher-resolution T2 imaging in large cohort studies, 3D T1-weighted images are not easily available in some contexts, in particular in clinical setting.
We developed a hippocampal segmentation model specifically designed for MR Angiography (MRA), a modality primarily used for vascular assessment and available for routine screening exams in many Japanese medical check-ups. Although MRA images are not typically intended for structural analysis, advancements in image processing-such as Convolutional Neural Networks (CNNs)-have made the identification of the hippocampus feasible, even with the limited contrast or high noise obscuring anatomical landmarks.
This study provides a first evaluation of the ability of our CNN model to track hippocampal atrophy longitudinally from MR Angiography
We developed a hippocampal segmentation model specifically designed for MR Angiography (MRA), a modality primarily used for vascular assessment and available for routine screening exams in many Japanese medical check-ups. Although MRA images are not typically intended for structural analysis, advancements in image processing-such as Convolutional Neural Networks (CNNs)-have made the identification of the hippocampus feasible, even with the limited contrast or high noise obscuring anatomical landmarks.
This study provides a first evaluation of the ability of our CNN model to track hippocampal atrophy longitudinally from MR Angiography
Methods:
Segmentation Model. A CNN was trained using paired 3D T1-weighted and MRA images, carefully co-registered. The input to the model was the MRA, and its training target labels were obtained from Hippodeep (Thyreau et al, 2018) applied on the corresponding 3D T1-weighted images. This 'hippodeep-mra' model will be made available as opensource. In the present study, the resulting hippocampal volume was the primary metric of interest.
Dataset. A robust and generalizable model was trained using a large dataset of 3,300 images from three sources:
* JPSC-AD cohort: The Japan Prospective Studies Collaboration for Aging and Dementia (JPSC-AD) is a multisite, population-based prospective cohort study of dementia (Ninomiya et al, 2020). Data from four sites with MRA images available (3502 subjects, aged 50–90) were included, using acquisitions at 1.5T or 3T. Three sites were used for training the model (Kyushu University, Matsue Medical Center, and Ehime University), while the MRA from Iwate Medical University (a subset of 261 subjects) was reserved for testing and longitudinal evaluation
* Tohoku University Hospital. A sample of 438 subjects, aged 50–83. Acquisition at 3T.
* IXI dataset: 569 subjects, aged 20–86, including data from three hospitals in the UK (Hammersmith Hospital, Guy's Hospital, Institute Of Psychiatry), using 1.5 or 3T acquisitions
For longitudinal evaluation, the Iwate cohort included 261 subjects with two MRA scans acquired approximately three years apart. Scans were processed independently without co-registration to maintain simplicity. Furthermore, some additional variability was caused by some scanner inconsistency between the two timepoints.
Dataset. A robust and generalizable model was trained using a large dataset of 3,300 images from three sources:
* JPSC-AD cohort: The Japan Prospective Studies Collaboration for Aging and Dementia (JPSC-AD) is a multisite, population-based prospective cohort study of dementia (Ninomiya et al, 2020). Data from four sites with MRA images available (3502 subjects, aged 50–90) were included, using acquisitions at 1.5T or 3T. Three sites were used for training the model (Kyushu University, Matsue Medical Center, and Ehime University), while the MRA from Iwate Medical University (a subset of 261 subjects) was reserved for testing and longitudinal evaluation
* Tohoku University Hospital. A sample of 438 subjects, aged 50–83. Acquisition at 3T.
* IXI dataset: 569 subjects, aged 20–86, including data from three hospitals in the UK (Hammersmith Hospital, Guy's Hospital, Institute Of Psychiatry), using 1.5 or 3T acquisitions
For longitudinal evaluation, the Iwate cohort included 261 subjects with two MRA scans acquired approximately three years apart. Scans were processed independently without co-registration to maintain simplicity. Furthermore, some additional variability was caused by some scanner inconsistency between the two timepoints.
Results:
Fig 1. illustrate the result of MRA-based hippocampal segmentation on a test subject's MRA.
Cross-sectional Validation: The hippocampal volumes estimated by our model showed strong correlation (R = 0.95) with those obtained using Hippodeep on 3D T1-weighted images. (Fig 2.a)
Longitudinal Analysis: As expected, a global decrease of hippocampal volume was observed (Fig 2.b). Bilateral hippocampal volumes decreased from an average of 6611 ± 711 mm³ at the first timepoint to 6507 ± 789 mm³ at the second, corresponding to a mean atrophy of -104 ± 166 mm³ over three years. There was a small yet non-significant gender effect.
Cross-sectional Validation: The hippocampal volumes estimated by our model showed strong correlation (R = 0.95) with those obtained using Hippodeep on 3D T1-weighted images. (Fig 2.a)
Longitudinal Analysis: As expected, a global decrease of hippocampal volume was observed (Fig 2.b). Bilateral hippocampal volumes decreased from an average of 6611 ± 711 mm³ at the first timepoint to 6507 ± 789 mm³ at the second, corresponding to a mean atrophy of -104 ± 166 mm³ over three years. There was a small yet non-significant gender effect.
·MRA hippocampus segmentation outcome for a single Test Subject
·Test Sample volume statistics
Conclusions:
This study demonstrates that MRA, despite its primary focus on vasculature, can generally be used to track hippocampal atrophy longitudinally. This approach could be valuable in clinical settings where 3D T1-weighted MRI is unavailable.
Lifespan Development:
Aging
Modeling and Analysis Methods:
Segmentation and Parcellation 1
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Subcortical Structures 2
Novel Imaging Acquisition Methods:
Anatomical MRI
Imaging Methods Other
Keywords:
Aging
ANGIOGRAPHY
MR ANGIOGRAPHY
Segmentation
STRUCTURAL MRI
Sub-Cortical
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Thyreau, B., Sato, K., Fukuda, H., & Taki, Y. (2018). Segmentation of the hippocampus by transferring algorithmic knowledge for large cohort processing. Medical image analysis, 43, 214-228.
IXI Dataet, https://brain-development.org/ixi-dataset/