-
Murine AI excels at cats and cheese: Structural differences between human and mouse neurons and their implementation in generative AIs
Authors:
Rino Saiga,
Kaede Shiga,
Yo Maruta,
Chie Inomoto,
Hiroshi Kajiwara,
Naoya Nakamura,
Yu Kakimoto,
Yoshiro Yamamoto,
Masahiro Yasutake,
Masayuki Uesugi,
Akihisa Takeuchi,
Kentaro Uesugi,
Yasuko Terada,
Yoshio Suzuki,
Viktor Nikitin,
Vincent De Andrade,
Francesco De Carlo,
Yuichi Yamashita,
Masanari Itokawa,
Soichiro Ide,
Kazutaka Ikeda,
Ryuta Mizutani
Abstract:
Mouse and human brains have different functions that depend on their neuronal networks. In this study, we analyzed nanometer-scale three-dimensional structures of brain tissues of the mouse medial prefrontal cortex and compared them with structures of the human anterior cingulate cortex. The obtained results indicated that mouse neuronal somata are smaller and neurites are thinner than those of hu…
▽ More
Mouse and human brains have different functions that depend on their neuronal networks. In this study, we analyzed nanometer-scale three-dimensional structures of brain tissues of the mouse medial prefrontal cortex and compared them with structures of the human anterior cingulate cortex. The obtained results indicated that mouse neuronal somata are smaller and neurites are thinner than those of human neurons. These structural features allow mouse neurons to be integrated in the limited space of the brain, though thin neurites should suppress distal connections according to cable theory. We implemented this mouse-mimetic constraint in convolutional layers of a generative adversarial network (GAN) and a denoising diffusion implicit model (DDIM), which were then subjected to image generation tasks using photo datasets of cat faces, cheese, human faces, and birds. The mouse-mimetic GAN outperformed a standard GAN in the image generation task using the cat faces and cheese photo datasets, but underperformed for human faces and birds. The mouse-mimetic DDIM gave similar results, suggesting that the nature of the datasets affected the results. Analyses of the four datasets indicated differences in their image entropy, which should influence the number of parameters required for image generation. The preferences of the mouse-mimetic AIs coincided with the impressions commonly associated with mice. The relationship between the neuronal network and brain function should be investigated by implementing other biological findings in artificial neural networks.
△ Less
Submitted 28 October, 2024;
originally announced October 2024.
-
Structural aging of human neurons is the opposite of the changes in schizophrenia
Authors:
Ryuta Mizutani,
Rino Saiga,
Yoshiro Yamamoto,
Masayuki Uesugi,
Akihisa Takeuchi,
Kentaro Uesugi,
Yasuko Terada,
Yoshio Suzuki,
Vincent De Andrade,
Francesco De Carlo,
Susumu Takekoshi,
Chie Inomoto,
Naoya Nakamura,
Youta Torii,
Itaru Kushima,
Shuji Iritani,
Norio Ozaki,
Kenichi Oshima,
Masanari Itokawa,
Makoto Arai
Abstract:
Human mentality develops with age and is altered in psychiatric disorders, though their underlying mechanism is unknown. In this study, we analyzed nanometer-scale three-dimensional structures of brain tissues of the anterior cingulate cortex from eight schizophrenia and eight control cases. The distribution profiles of neurite curvature of the control cases showed a trend depending on their age,…
▽ More
Human mentality develops with age and is altered in psychiatric disorders, though their underlying mechanism is unknown. In this study, we analyzed nanometer-scale three-dimensional structures of brain tissues of the anterior cingulate cortex from eight schizophrenia and eight control cases. The distribution profiles of neurite curvature of the control cases showed a trend depending on their age, resulting in an age-correlated decrease in the standard deviation of neurite curvature (Pearson's r = -0.80, p = 0.018). In contrast to the control cases, the schizophrenia cases deviate upward from this correlation, exhibiting a 60% higher neurite curvature compared with the controls (p = 7.8 x 10^(-4)). The neurite curvature also showed a correlation with a hallucination score (Pearson's r = 0.80, p = 1.8 x 10^(-4)), indicating that neurite structure is relevant to brain function. We suggest that neurite curvature plays a pivotal role in brain aging and can be used as a hallmark to exploit a novel treatment of schizophrenia. This nano-CT paper is the result of our decade-long analysis and is unprecedented in terms of number of cases.
△ Less
Submitted 13 November, 2022;
originally announced November 2022.
-
Schizophrenia-mimicking layers outperform conventional neural network layers
Authors:
Ryuta Mizutani,
Senta Noguchi,
Rino Saiga,
Yuichi Yamashita,
Mitsuhiro Miyashita,
Makoto Arai,
Masanari Itokawa
Abstract:
We have reported nanometer-scale three-dimensional studies of brain networks of schizophrenia cases and found that their neurites are thin and tortuous compared to healthy controls. This suggests that connections between distal neurons are suppressed in microcircuits of schizophrenia cases. In this study, we applied these biological findings to the design of schizophrenia-mimicking artificial neur…
▽ More
We have reported nanometer-scale three-dimensional studies of brain networks of schizophrenia cases and found that their neurites are thin and tortuous compared to healthy controls. This suggests that connections between distal neurons are suppressed in microcircuits of schizophrenia cases. In this study, we applied these biological findings to the design of schizophrenia-mimicking artificial neural network to simulate the observed connection alteration in the disorder. Neural networks having a "schizophrenia connection layer" in place of a fully connected layer were subjected to image classification tasks using the MNIST and CIFAR-10 datasets. The results revealed that the schizophrenia connection layer is tolerant to overfitting and outperforms a fully connected layer. The outperformance was observed only for networks using band matrices as weight windows, indicating that the shape of the weight matrix is relevant to the network performance. A schizophrenia convolution layer was also tested using the VGG configuration, showing that 60% of the kernel weights of the last three convolution layers can be eliminated without loss of accuracy. The schizophrenia layers can be used instead of conventional layers without any change in the network configuration and training procedures; hence, neural networks can easily take advantage of these layers. The results of this study suggest that the connection alteration found in schizophrenia is not a burden to the brain, but has functional roles in brain performance.
△ Less
Submitted 1 April, 2022; v1 submitted 22 September, 2020;
originally announced September 2020.
-
Structural diverseness of neurons between brain areas and between cases
Authors:
Ryuta Mizutani,
Rino Saiga,
Yoshiro Yamamoto,
Masayuki Uesugi,
Akihisa Takeuchi,
Kentaro Uesugi,
Yasuko Terada,
Yoshio Suzuki,
Vincent De Andrade,
Francesco De Carlo,
Susumu Takekoshi,
Chie Inomoto,
Naoya Nakamura,
Youta Torii,
Itaru Kushima,
Shuji Iritani,
Norio Ozaki,
Kenichi Oshima,
Masanari Itokawa,
Makoto Arai
Abstract:
The cerebral cortex is composed of multiple cortical areas that exert a wide variety of brain functions. Although human brain neurons are genetically and areally mosaic, the three-dimensional structural differences between neurons in different brain areas or between the neurons of different individuals have not been delineated. Here, we report a nanometer-scale geometric analysis of brain tissues…
▽ More
The cerebral cortex is composed of multiple cortical areas that exert a wide variety of brain functions. Although human brain neurons are genetically and areally mosaic, the three-dimensional structural differences between neurons in different brain areas or between the neurons of different individuals have not been delineated. Here, we report a nanometer-scale geometric analysis of brain tissues of the superior temporal gyrus of 4 schizophrenia and 4 control cases by using synchrotron radiation nanotomography. The results of the analysis and a comparison with results for the anterior cingulate cortex indicated that 1) neuron structures are dissimilar between brain areas and that 2) the dissimilarity varies from case to case. The structural diverseness was mainly observed in terms of the neurite curvature that inversely correlates with the diameters of the neurites and spines. The analysis also revealed the geometric differences between the neurons of the schizophrenia and control cases, suggesting that neuron structure is associated with brain function. The area dependency of the neuron structure and its diverseness between individuals should represent the individuality of brain functions.
△ Less
Submitted 30 June, 2020;
originally announced July 2020.
-
Synchrotron radiation microtomography of brain hemisphere and spinal cord of a mouse model of multiple sclerosis revealed a correlation between capillary dilation and clinical score
Authors:
Rino Saiga,
Masato Hoshino,
Akihisa Takeuchi,
Kentaro Uesugi,
Katsuko Naitou,
Akemi Kamijo,
Noboru Kawabe,
Masato Ohtsuka,
Shunya Takizawa,
Ryuta Mizutani
Abstract:
Multiple sclerosis is a neurological disorder in which the myelin sheaths of axons are damaged by the immune response. We report here a three-dimensional structural analysis of brain and spinal cord tissues of a mouse model of multiple sclerosis, known as experimental autoimmune encephalomyelitis (EAE). EAE-induced mice were raised with or without administration of fingolimod, which is used in the…
▽ More
Multiple sclerosis is a neurological disorder in which the myelin sheaths of axons are damaged by the immune response. We report here a three-dimensional structural analysis of brain and spinal cord tissues of a mouse model of multiple sclerosis, known as experimental autoimmune encephalomyelitis (EAE). EAE-induced mice were raised with or without administration of fingolimod, which is used in the treatment of multiple sclerosis. Brains and spinal cords dissected from the EAE mice were lyophilized so as to reconstitute the intrinsic contrast of tissue elements, such as axons, in X-ray images. Three-dimensional structures of the brain hemispheres and spinal cords of the EAE mice were visualized with synchrotron radiation microtomography. Microtomographic cross sections reconstructed from the X-ray images revealed dilation of capillary vessels and vacuolation in the spinal cord of the EAE mice. Vacuolation was also observed in the cerebellum, suggesting that the neuroinflammatory response progressed in the brain. The vessel networks and vacuolation lesions in the spinal cords were modelled by automatically tracing the three-dimensional image in order to analyze the tissue structures quantitatively. The results of the analysis indicated that the distribution of vacuolations was not uniform but three-dimensionally localized. The mean vessel diameter showed a linear correlation with the clinical score, indicating that vasodilation is relevant to paralysis severity in the disease model. We suggest that vasodilation and vacuolation are related with neurological symptoms of multiple sclerosis.
△ Less
Submitted 1 October, 2018;
originally announced October 2018.
-
Three-dimensional alteration of neurites in schizophrenia
Authors:
Ryuta Mizutani,
Rino Saiga,
Akihisa Takeuchi,
Kentaro Uesugi,
Yasuko Terada,
Yoshio Suzuki,
Vincent De Andrade,
Francesco De Carlo,
Susumu Takekoshi,
Chie Inomoto,
Naoya Nakamura,
Itaru Kushima,
Shuji Iritani,
Norio Ozaki,
Soichiro Ide,
Kazutaka Ikeda,
Kenichi Oshima,
Masanari Itokawa,
Makoto Arai
Abstract:
This paper reports nano-CT analysis of brain tissues of schizophrenia and control cases. The analysis revealed that: (1) neuronal structures vary between individuals, (2) the mean curvature of distal neurites of the schizophrenia cases was 1.5 times higher than that of the controls, and (3) dendritic spines were categorized into two geometrically distinctive groups, though no structural difference…
▽ More
This paper reports nano-CT analysis of brain tissues of schizophrenia and control cases. The analysis revealed that: (1) neuronal structures vary between individuals, (2) the mean curvature of distal neurites of the schizophrenia cases was 1.5 times higher than that of the controls, and (3) dendritic spines were categorized into two geometrically distinctive groups, though no structural differences were observed between the disease and control cases. The differences in the neurite curvature result in differences in the spatial trajectory and hence alter neuronal circuits. We suggest that the structural alteration of neurons in the schizophrenia cases should reflect psychiatric symptoms of schizophrenia.
△ Less
Submitted 16 February, 2019; v1 submitted 2 April, 2018;
originally announced April 2018.
-
Method for estimating modulation transfer function from sample images
Authors:
Rino Saiga,
Akihisa Takeuchi,
Kentaro Uesugi,
Yasuko Terada,
Yoshio Suzuki,
Ryuta Mizutani
Abstract:
The modulation transfer function (MTF) represents the frequency domain response of imaging modalities. Here, we report a method for estimating the MTF from sample images. Test images were generated from a number of images, including those taken with an electron microscope and with an observation satellite. These original images were convolved with point spread functions (PSFs) including those of c…
▽ More
The modulation transfer function (MTF) represents the frequency domain response of imaging modalities. Here, we report a method for estimating the MTF from sample images. Test images were generated from a number of images, including those taken with an electron microscope and with an observation satellite. These original images were convolved with point spread functions (PSFs) including those of circular apertures. The resultant test images were subjected to a Fourier transformation. The logarithm of the squared norm of the Fourier transform was plotted against the squared distance from the origin. Linear correlations were observed in the logarithmic plots, indicating that the PSF of the test images can be approximated with a Gaussian. The MTF was then calculated from the Gaussian-approximated PSF. The obtained MTF closely coincided with the MTF predicted from the original PSF. The MTF of an x-ray microtomographic section of a fly brain was also estimated with this method. The obtained MTF showed good agreement with the MTF determined from an edge profile of an aluminum test object. We suggest that this approach is an alternative way of estimating the MTF, independently of the image type.
△ Less
Submitted 24 November, 2017;
originally announced November 2017.
-
Estimating the resolution of real images
Authors:
Ryuta Mizutani,
Rino Saiga,
Susumu Takekoshi,
Chie Inomoto,
Naoya Nakamura,
Makoto Arai,
Kenichi Oshima,
Masanari Itokawa,
Akihisa Takeuchi,
Kentaro Uesugi,
Yasuko Terada,
Yoshio Suzuki
Abstract:
Image resolvability is the primary concern in imaging. This paper reports an estimation of the full width at half maximum of the point spread function from a Fourier domain plot of real sample images by neither using test objects, nor defining a threshold criterion. We suggest that this method can be applied to any type of image, independently of the imaging modality.
Image resolvability is the primary concern in imaging. This paper reports an estimation of the full width at half maximum of the point spread function from a Fourier domain plot of real sample images by neither using test objects, nor defining a threshold criterion. We suggest that this method can be applied to any type of image, independently of the imaging modality.
△ Less
Submitted 2 March, 2017;
originally announced March 2017.
-
Three-dimensional X-ray visualization of axonal tracts in mouse brain hemisphere
Authors:
Ryuta Mizutani,
Rino Saiga,
Masato Ohtsuka,
Hiromi Miura,
Masato Hoshino,
Akihisa Takeuchi,
Kentaro Uesugi
Abstract:
Neurons transmit active potentials through axons, which are essential for the brain to function. In this study, the axonal networks of the murine brain were visualized with X-ray tomographic microscopy, also known as X-ray microtomography or micro-CT. Murine brain samples were freeze-dried to reconstitute the intrinsic contrast of tissue constituents and subjected to X-ray visualization. A whole b…
▽ More
Neurons transmit active potentials through axons, which are essential for the brain to function. In this study, the axonal networks of the murine brain were visualized with X-ray tomographic microscopy, also known as X-ray microtomography or micro-CT. Murine brain samples were freeze-dried to reconstitute the intrinsic contrast of tissue constituents and subjected to X-ray visualization. A whole brain hemisphere visualized by absorption contrast illustrated three-dimensional structures including those of the striatum, corpus callosum, and anterior commissure. Axonal tracts observed in the striatum start from the basal surface of the cerebral cortex and end at various positions in the basal ganglia. The distribution of X-ray attenuation coefficients indicated that differences in water and phospholipid content between the myelin sheath and surrounding tissue constituents account for the observed contrast. A rod-shaped cutout of brain tissue was also analyzed with a phase retrieval method, wherein tissue microstructures could be resolved with up to 2.7 μm resolution. Structures of axonal networks of the striatum were reconstructed by tracing axonal tracts. Such an analysis should be able to delineate the functional relationships of the brain regions involved in the observed network.
△ Less
Submitted 2 October, 2016;
originally announced October 2016.
-
A method for estimating spatial resolution of real image in the Fourier domain
Authors:
Ryuta Mizutani,
Rino Saiga,
Susumu Takekoshi,
Chie Inomoto,
Naoya Nakamura,
Masanari Itokawa,
Makoto Arai,
Kenichi Oshima,
Akihisa Takeuchi,
Kentaro Uesugi,
Yasuko Terada,
Yoshio Suzuki
Abstract:
Spatial resolution is a fundamental parameter in structural sciences. In crystallography, the resolution is determined from the detection limit of high-angle diffraction in reciprocal space. In electron microscopy, correlation in the Fourier domain is used for estimating the resolution. In this paper, we report a method for estimating the spatial resolution of real images from a logarithmic intens…
▽ More
Spatial resolution is a fundamental parameter in structural sciences. In crystallography, the resolution is determined from the detection limit of high-angle diffraction in reciprocal space. In electron microscopy, correlation in the Fourier domain is used for estimating the resolution. In this paper, we report a method for estimating the spatial resolution of real images from a logarithmic intensity plot in the Fourier domain. The logarithmic intensity plots of test images indicated that the full width at half maximum of a Gaussian point-spread function can be estimated from the images. The spatial resolution of imaging X-ray microtomography using Fresnel zone-plate optics was also estimated with this method. A cross section of a test object visualized with the imaging microtomography indicated that square-wave patterns up to 120-nm pitch were resolved. The logarithmic intensity plot was calculated from a tomographic cross section of brain tissue. The full width at half maximum of the point spread function estimated from the plot coincided with the resolution determined from the test object. These results indicated that the logarithmic intensity plot in the Fourier domain provides an alternative measure of the spatial resolution without explicitly defining a noise criterion.
△ Less
Submitted 8 September, 2016;
originally announced September 2016.
-
Three-dimensional network of Drosophila brain hemisphere
Authors:
Ryuta Mizutani,
Rino Saiga,
Akihisa Takeuchi,
Kentaro Uesugi,
Yoshio Suzuki
Abstract:
The first step to understanding brain function is to determine the brain's network structure. We report a three-dimensional analysis of the brain network of the fruit fly Drosophila melanogaster by synchrotron-radiation tomographic microscopy. A skeletonized wire model of the left half of the brain network was built by tracing the three-dimensional distribution of X-ray absorption coefficients. Th…
▽ More
The first step to understanding brain function is to determine the brain's network structure. We report a three-dimensional analysis of the brain network of the fruit fly Drosophila melanogaster by synchrotron-radiation tomographic microscopy. A skeletonized wire model of the left half of the brain network was built by tracing the three-dimensional distribution of X-ray absorption coefficients. The obtained models of neuronal processes were classified into groups on the basis of their three-dimensional structures. These classified groups correspond to neuronal tracts that send long-range projections or repeated structures of the optic lobe. The skeletonized model is also composed of neuronal processes that could not be classified into the groups. The distribution of these unclassified structures correlates with the distribution of contacts between neuronal processes. This suggests that neurons that cannot be classified into typical structures should play important roles in brain functions. The quantitative description of the brain network provides a basis for structural and statistical analyses of the Drosophila brain. The challenge is to establish a methodology for reconstructing the brain network in a higher-resolution image, leading to a comprehensive understanding of the brain structure.
△ Less
Submitted 8 September, 2016;
originally announced September 2016.
