1
|
Valabregue R, Girka F, Pron A, Rousseau F, Auzias G. Comprehensive analysis of synthetic learning applied to neonatal brain MRI segmentation. Hum Brain Mapp 2024; 45:e26674. [PMID: 38651625 PMCID: PMC11036377 DOI: 10.1002/hbm.26674] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/02/2023] [Revised: 03/09/2024] [Accepted: 03/18/2024] [Indexed: 04/25/2024] Open
Abstract
Brain segmentation from neonatal MRI images is a very challenging task due to large changes in the shape of cerebral structures and variations in signal intensities reflecting the gestational process. In this context, there is a clear need for segmentation techniques that are robust to variations in image contrast and to the spatial configuration of anatomical structures. In this work, we evaluate the potential of synthetic learning, a contrast-independent model trained using synthetic images generated from the ground truth labels of very few subjects. We base our experiments on the dataset released by the developmental Human Connectome Project, for which high-quality images are available for more than 700 babies aged between 26 and 45 weeks postconception. First, we confirm the impressive performance of a standard UNet trained on a few volumes, but also confirm that such models learn intensity-related features specific to the training domain. We then confirm the robustness of the synthetic learning approach to variations in image contrast. However, we observe a clear influence of the age of the baby on the predictions. We improve the performance of this model by enriching the synthetic training set with realistic motion artifacts and over-segmentation of the white matter. Based on extensive visual assessment, we argue that the better performance of the model trained on real T2w data may be due to systematic errors in the ground truth. We propose an original experiment allowing us to show that learning from real data will reproduce any systematic bias affecting the training set, while synthetic models can avoid this limitation. Overall, our experiments confirm that synthetic learning is an effective solution for segmenting neonatal brain MRI. Our adapted synthetic learning approach combines key features that will be instrumental for large multisite studies and clinical applications.
Collapse
Affiliation(s)
- R. Valabregue
- CENIR, Institut du Cerveau (ICM)—Paris Brain Institute, Inserm U 1127, CNRS UMR 7225, Sorbonne UniversitéParisFrance
| | - F. Girka
- CENIR, Institut du Cerveau (ICM)—Paris Brain Institute, Inserm U 1127, CNRS UMR 7225, Sorbonne UniversitéParisFrance
| | - A. Pron
- Aix‐Marseille Université, CNRS, Institut de Neurosciences de la Timone, UMR 7289MarseilleFrance
| | - F. Rousseau
- IMT Atlantique, LaTIM INSERM U1101BrestFrance
| | - G. Auzias
- Aix‐Marseille Université, CNRS, Institut de Neurosciences de la Timone, UMR 7289MarseilleFrance
| |
Collapse
|
2
|
Horibe K, Taga G, Fujimoto K. Geodesic theory of long association fibers arrangement in the human fetal cortex. Cereb Cortex 2023; 33:9778-9786. [PMID: 37482884 PMCID: PMC10472492 DOI: 10.1093/cercor/bhad243] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/27/2023] [Revised: 06/12/2023] [Accepted: 06/13/2023] [Indexed: 07/25/2023] Open
Abstract
Association fibers connect different areas of the cerebral cortex over long distances and integrate information to achieve higher brain functions, particularly in humans. Prototyped association fibers are developed to the respective tangential direction throughout the cerebral hemispheres along the deepest border of the subplate during the fetal period. However, how guidance to remote areas is achieved is not known. Because the subplate is located below the cortical surface, the tangential direction of the fibers may be biased by the curved surface geometry due to Sylvian fissure and cortical poles. The fiber length can be minimized if the tracts follow the shortest paths (geodesics) of the curved surface. Here, we propose and examine a theory that geodesics guide the tangential direction of long association fibers by analyzing how geodesics are spatially distributed on the fetal human brains. We found that the geodesics were dense on the saddle-shaped surface of the perisylvian region and sparse on the dome-shaped cortical poles. The geodesics corresponded with the arrangement of five typical association fibers, supporting the theory. Thus, the geodesic theory provides directional guidance information for wiring remote areas and suggests that long association fibers emerge from minimizing their tangential length in fetal brains.
Collapse
Affiliation(s)
- Kazuya Horibe
- Department of Biological Sciences, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, 560-0043 Osaka, Japan
- Department of Systems Innovation, Graduate School of Engineering Science, Osaka University, 1-3 Machikaneyama-cho, Toyonaka, 560-8531 Osaka, Japan
| | - Gentaro Taga
- Department of Physical and Health Education, Graduate School of Education, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, 113-0033 Tokyo, Japan
| | - Koichi Fujimoto
- Department of Biological Sciences, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, 560-0043 Osaka, Japan
- Program of Mathematical and Life Sciences, Hiroshima University, 1-3-1 Kagamiyama, Higashi-Hiroshima, 739-8526 Hiroshima, Japan
| |
Collapse
|
3
|
Pogledic I, Schwartz E, Bobić-Rasonja M, Mitter C, Baltzer P, Gruber GM, Milković-Periša M, Haberler C, Bettelheim D, Kasprian G, Judaš M, Prayer D, Jovanov-Milošević N. 3T MRI signal intensity profiles and thicknesses of transient zones in human fetal brain at mid-gestation. Eur J Paediatr Neurol 2021; 35:67-73. [PMID: 34653829 DOI: 10.1016/j.ejpn.2021.09.014] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 05/03/2021] [Revised: 09/07/2021] [Accepted: 09/13/2021] [Indexed: 11/15/2022]
Abstract
In this study we compare temporal lobe (TL) signal intensity (SI) profiles, along with the average thicknesses of the transient zones obtained from postmortem MRI (pMRI) scans and corresponding histological slices, to the frontal lobe (FL) SI and zone thicknesses, in normal fetal brains. The purpose was to assess the synchronization of the corticogenetic processes in different brain lobes. Nine postmortem human fetal brains without cerebral pathologies, from 19 to 24 weeks of gestation (GW) were analyzed on T2-weighted 3T pMRI, at the coronal level of the thalamus and basal ganglia. The SI profiles of the transient zones in the TL correlate well spatially and temporally to the signal intensity profile of the FL. During the examined period, in the TL, the intermediate and subventricular zone are about the size of the subplate zone (SP), while the superficial SP demonstrates the highest signal intensity. The correlation of the SI profiles and the distributions of the transient zones in the two brain lobes, indicates a time-aligned histogenesis during this narrow time window. The 3TpMRI enables an assessment of the regularity of lamination patterns in the fetal telencephalic wall, upon comparative evaluation of sizes of the transient developmental zones and the SI profiles of different cortical regions. A knowledge of normal vs. abnormal transient lamination patterns and the SI profiles is a prerequisite for further advancement of the MR diagnostic tools needed for early detection of developmental brain pathologies prenatally, especially mild white matter injuries such as lesions of TL due to prenatal cytomegalovirus infections, or cortical malformations.
Collapse
Affiliation(s)
- Ivana Pogledic
- Division of Neuroradiology and Musculoskeletal Radiology, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, 1090, Vienna, Austria
| | - Ernst Schwartz
- Computational Imaging Research Lab, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, 1090, Vienna, Austria
| | - Mihaela Bobić-Rasonja
- University of Zagreb, School of Medicine, Croatian Institute for Brain Research, Section for Developmental Neuroscience, Scientific Centre of Excellence for Basic, Clinical and Translational Neuroscience, Šalata 12, 10000, Zagreb, Croatia; University of Zagreb, School of Medicine, Department of Biology, Šalata 3, 10000, Zagreb, Croatia
| | - Christian Mitter
- Division of Neuroradiology and Musculoskeletal Radiology, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, 1090, Vienna, Austria
| | - Pascal Baltzer
- Division of Molecular and Gender Imaging, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, 1090, Vienna, Austria
| | - Gerlinde Maria Gruber
- Department of Anatomy and Biomechanics, Karl Landsteiner University of Health Sciences, 3500, Krems, Austria
| | - Marija Milković-Periša
- University Hospital Centre Zagreb, Department of Pathology and Cytology, Petrova 13, 10000, Zagreb, Croatia; University of Zagreb, School of Medicine, Institute of Pathology, Šalata 10, 10000 Zagreb, Croatia
| | - Christine Haberler
- Division of Neuropathology and Neurochemistry, Department of Neurology, Medical University of Vienna, 1090, Vienna, Austria
| | - Dieter Bettelheim
- Division of Obstetrics and Feto-Maternal Medicine, Department of Obstetrics and Gynecology, Medical University of Vienna, 1090, Vienna, Austria
| | - Gregor Kasprian
- Division of Neuroradiology and Musculoskeletal Radiology, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, 1090, Vienna, Austria
| | - Miloš Judaš
- University of Zagreb, School of Medicine, Croatian Institute for Brain Research, Section for Developmental Neuroscience, Scientific Centre of Excellence for Basic, Clinical and Translational Neuroscience, Šalata 12, 10000, Zagreb, Croatia
| | - Daniela Prayer
- Division of Neuroradiology and Musculoskeletal Radiology, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, 1090, Vienna, Austria
| | - Nataša Jovanov-Milošević
- University of Zagreb, School of Medicine, Croatian Institute for Brain Research, Section for Developmental Neuroscience, Scientific Centre of Excellence for Basic, Clinical and Translational Neuroscience, Šalata 12, 10000, Zagreb, Croatia; University of Zagreb, School of Medicine, Department of Biology, Šalata 3, 10000, Zagreb, Croatia.
| |
Collapse
|
4
|
Transitory and Vestigial Structures of the Developing Human Nervous System. Pediatr Neurol 2021; 123:86-101. [PMID: 34416613 DOI: 10.1016/j.pediatrneurol.2021.07.003] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 02/20/2021] [Revised: 07/02/2021] [Accepted: 07/03/2021] [Indexed: 11/23/2022]
Abstract
As with many body organs, the human central nervous system contains many structures and cavities that may have had functions in embryonic and fetal life but are vestigial or atrophic at maturity. Examples are the septum pellucidum, remnants of the lamina terminalis, Cajal-Retzius neurons, induseum griseum, habenula, and accessory olfactory bulb. Other structures are transitory in fetal or early postnatal life, disappearing from the mature brain. Examples are the neural crest, subpial granular glial layer of Brun over cerebral cortex, radial glial cells, and subplate zone of cerebral cortex. At times persistent fetal structures that do not regress may cause neurological problems or indicate a pathologic condition, such as Blake pouch cyst. Transitory structures thus can become vestigial. Examples are an excessively wide cavum septi pellucidi, suprapineal recess of the third ventricle, trigeminal artery of the posterior fossa circulation, and hyaloid ocular artery. Arrested maturation might be considered another aspect of vestigial structure. An example is the persistent microcolumnar cortical architecture in focal cortical dysplasia type Ia, in cortical zones of chronic fetal ischemia, and in some metabolic/genetic congenital encephalopathies. Some transitory structures in human brain are normal adult structures in lower vertebrates. Recognition of transitory and vestigial structures by fetal or postnatal neuroimaging and neuropathologically enables better understanding of cerebral ontogenesis and avoids misinterpretations.
Collapse
|
6
|
Bobić-Rasonja M, Pogledić I, Mitter C, Štajduhar A, Milković-Periša M, Trnski S, Bettelheim D, Hainfellner JA, Judaš M, Prayer D, Jovanov-Milošević N. Developmental Differences Between the Limbic and Neocortical Telencephalic Wall: An Intrasubject Slice-Matched 3 T MRI-Histological Correlative Study in Humans. Cereb Cortex 2021; 31:3536-3550. [PMID: 33704445 DOI: 10.1093/cercor/bhab030] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/05/2020] [Revised: 01/19/2021] [Accepted: 01/26/2021] [Indexed: 12/20/2022] Open
Abstract
The purpose of the study was to investigate the interrelation of the signal intensities and thicknesses of the transient developmental zones in the cingulate and neocortical telencephalic wall, using T2-weighted 3 T-magnetic resonance imaging (MRI) and histological scans from the same brain hemisphere. The study encompassed 24 postmortem fetal brains (15-35 postconceptional weeks, PCW). The measurements were performed using Fiji and NDP.view2. We found that T2w MR signal-intensity curves show a specific regional and developmental stage profile already at 15 PCW. The MRI-histological correlation reveals that the subventricular-intermediate zone (SVZ-IZ) contributes the most to the regional differences in the MRI-profile and zone thicknesses, growing by a factor of 2.01 in the cingulate, and 1.78 in the neocortical wall. The interrelations of zone or wall thicknesses, obtained by both methods, disclose a different rate and extent of shrinkage per region (highest in neocortical subplate and SVZ-IZ) and stage (highest in the early second half of fetal development), distorting the zones' proportion in histological sections. This intrasubject, slice-matched, 3 T correlative MRI-histological study provides important information about regional development of the cortical wall, critical for the design of MRI criteria for prenatal brain monitoring and early detection of cortical or other brain pathologies in human fetuses.
Collapse
Affiliation(s)
- Mihaela Bobić-Rasonja
- Croatian Institute for Brain Research, School of Medicine University of Zagreb, 10000 Zagreb, Croatia.,Department of Biology, School of Medicine, University of Zagreb, 10000 Zagreb, Croatia
| | - Ivana Pogledić
- Division of Neuroradiology and Musculoskeletal Radiology, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, 1090 Vienna, Austria
| | - Christian Mitter
- Division of Neuroradiology and Musculoskeletal Radiology, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, 1090 Vienna, Austria
| | - Andrija Štajduhar
- Croatian Institute for Brain Research, School of Medicine University of Zagreb, 10000 Zagreb, Croatia.,Andrija Štampar School of Public Health, School of Medicine, University of Zagreb, 10000 Zagreb, Croatia
| | - Marija Milković-Periša
- University Hospital Centre Zagreb, Department of Pathology and Cytology, 10000 Zagreb, Croatia
| | - Sara Trnski
- Croatian Institute for Brain Research, School of Medicine University of Zagreb, 10000 Zagreb, Croatia
| | - Dieter Bettelheim
- Division of Obstetrics and Feto-Maternal Medicine, Department of Obstetrics and Gynecology, Medical University of Vienna, 1090 Vienna, Austria
| | - Johannes A Hainfellner
- Division of Neuropathology and Neurochemistry, Department of Neurology, Medical University of Vienna, 1090 Vienna, Austria
| | - Miloš Judaš
- Croatian Institute for Brain Research, School of Medicine University of Zagreb, 10000 Zagreb, Croatia
| | - Daniela Prayer
- Division of Neuroradiology and Musculoskeletal Radiology, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, 1090 Vienna, Austria
| | - Nataša Jovanov-Milošević
- Croatian Institute for Brain Research, School of Medicine University of Zagreb, 10000 Zagreb, Croatia.,Department of Biology, School of Medicine, University of Zagreb, 10000 Zagreb, Croatia
| |
Collapse
|