ポスター
H. 方法論
H. Methodology |
|---|
| 2022年6月30日 13:00～14:00 沖縄コンベンションセンター 展示棟 ポスター会場1
1P-318
安静時脳活動からの個体情報の解読とそのげっ歯類への応用
Decoding of subject information from resting-state brain activity and its application to rodents
*稗田 健美(1)、地村 弘二(2)、石田 綾(3)、松井 鉄平(1,4)
1. 岡山大学、2. 慶應義塾大学、3. 理化学研究所、4. JSTさきがけ
*Takemi Hieda(1), Koji Jimura(2), Aya Ito-Ishida(3), Teppei Matsui(1,4)
1. Okayama University, 2. Keio University, 3. RIKEN Center for Brain Science, 4. JST-PRESTO
Keyword: RESTING STATE, fMRI, DECODING, CALCIUM IMAGIG
Recently, calcium imaging using the genetically encoded calcium indicator (GCaMP) and optical imaging of intrinsic signals have enabled the analysis of cortex-wide resting brain activity in mice (Matsui et al., PNAS, 2016; Matsui et al., Comm. Int. Bio., 2018). Recent studies using these methods have reported significant changes in resting brain activity in pathological mouse models and in the cerebral cortex of mice treated with psychotropic drugs (Vesuna et al., Nature, 2020; Busche et al., Nat. Neurosci., 2015). On the other hand, the analysis of resting brain activity is more advanced in human studies than in mouse studies. It is now known that various information about subjects, such as age, gender, and general fluid intelligence, can be decoded from resting brain activity. In particular, it has recently been reported that the accuracy of decoding individual information can be improved by incorporating the dynamic nature of resting brain activity (Liegeois et al., Nat. Comm., 2019). These methods of decoding individual information developed for the analysis of resting brain activity in humans are also expected to be useful for the analysis of resting brain activity in mice. In this study, we developed analysis methods for decoding subjects’ information from resting brain activity in humans, and then applied the analysis methods to extract phenotypic information from resting brain activity in mice. For the initial development of analysis methods, we used human resting brain activity data from the Human Connectome Project. We used data for 200 subjects each containing 1200 TRs of fMRI data with TR = 0.72 sec per scan (one subject was excluded from the analysis due to abnormal data). Time series were extracted from 264 Regions of Interest (ROIs) in the cortex based on a previous study by Power et al. (Power et al., Neuroimage, 2013), and a functional connectivity matrix was obtained by using temporal correlations between all ROIs. The functional connectivity matrix was used for classification by the Support Vector Machine. Initial investigation found that the gender of the subject could be correctly identified from the functional connectivity matrix in 75% of test data. We are now testing the autoregressive model to incorporate dynamic aspects of resting brain activity, and see if a reliable decoder could be obtained with less data. Concurrently we are acquiring widefield calcium imaging data on resting brain activity in mice to test these methods. | | 2022年6月30日 13:00～14:00 沖縄コンベンションセンター 展示棟 ポスター会場1
1P-319
マーモセットの個体識別・追跡を目的とした機械学習による標識タグの開発
Development of marker tags for individual identification and tracking marmosets by machine learning
*曾 彩華(1)、鴻池 菜保(2)、中村 克樹(2)、花沢 明俊(1)
1. 九州工業大学工学府、2. 京都大学霊長類研究所
*Ayaka So(1), Naho Konoike(2), Katsuki Nakamura(2), Akitoshi Hanazawa(1)
1. Grad Sch Eng, Kyusyu Institute of Technology, Fukuoka, Japan, 2. Primate Res Inst, Kyoto Univ, Aichi, Japan
Keyword: MARMOSET MONKEY, TRACKING, MACHINE LEARNING, DEEP LEARNING
We propose marker tags as a method for detecting and tracking the position of each experimental animal under collective rearing from video images. There are two main features in our method. Firstly, we can design the graphic patterns of the tags freely. Secondly, no specially developed library is required to build the system. These features are derived from that our method uses one of the general deep learning methods. Thus, we can conform the design of the marker tags to the purpose of use and the required number of the tags.
Our method used one of general object detection algorithms, Single Shot Multibox Detector (SSD), to detect and identify marker tags with black and white geometric graphic patterns. We prepared 1440 images (45 images for 32 types) as training data and 160 images (5 images for 32 types) as test data. For evaluation of the accuracy, we employed 10-fold cross validation.
After the learning, our system could detect marker tags in images. The result of each detection was shown by drawing a bounding box for indicating the detected position of the tag accompanied with texts showing the class and score indicating marker type and classification confidence. The accuracy was about 86%. Misclassification occurred between vertical or horizontal symmetry patterns. We would like to find better patterns for the marker tags and improve the detection system for better performance.
We will apply this method to marmoset monkeys (Callithrix jacchus) . The marmoset is one of the smallest primates that has around 25cm length and 350g weight. Due to the small body size and three-dimensional movement, there are difficulties in individual identification of the animals in video images. We think we can use the marker tags not only for the observation of experimental animals but also for various situations. For example, our system can be used as a tool for managing number of various items in factories and for checking safety of children in preschools or patients in hospitals. | | 2022年6月30日 13:00～14:00 沖縄コンベンションセンター 展示棟 ポスター会場1
1P-320
マウスナー細胞の軸索起始部を取り囲む特殊なグリア構造のμCTによる可視化
Visualization of a specialized glial structure surrounding the initial segment of the Mauthner cell with the synchrotron radiation X-ray micro computed tomography (SR−μCT)
*谷口 夏輝(1)、岩谷 将太(2)、上杉 健太郎(3)、安武 正展(3)、八田 公平(1,2)
1. 兵庫県立大学理学部、2. 兵庫県立大学大学院理学研究科、3. 高輝度光科学研究センター
*natsuki taniguchi(1), shota iwatani(2), kentaro uesgi(3), masahiro yasutake(3), kohei hatta(1,2)
1. Sch Sci, Univ of Hyogo, Hyogo, Japan, 2. Grad Sch Sci, Univ of Hyogo, Hyogo, Japan, 3. JASRI
Keyword: ZEBRAFISH, AXON CAP, SPRING-8, DENDRITE
The Mauthner (M-) cells are a pair of giant excitatory interneurons located in the rhombomere 4 of the hindbrain in teleost. They control a rapid C-start escape response, in which the body moves away from the stimulus input to initiate escape behavior. The initial segment of the Mauthner cell is surrounded by axons of excitatory spiral neurons and axons of inhibitory feed-back neurons. This specialized structure is called the Axon Cap. In addition, in many teleost species, a specialized group of glial cells, called Axon Cap Glia, surrounds the Axon Cap, forming an electrical insulator. It enhances the extracellular field derived from the action potentials of the feed-back neurons and causes an unusually rapid ‘electrical inhibition' (Furukawa and Furshpan, 1963; Hatta and Korn, 1998).
In this work, we used a high-resolution synchrotron radiation X-ray micro computed tomography (SR−μCT) at SPring-8, to observe the M-cells and the surrounding structures in the larval and adult zebrafish brains embedded in paraffin with or without phosphotangstic acid (PTA) staining. In the adult brain, the Axon Cap, the initial segment, and Axon Cap Glia including their cellular nuclei are clearly visualized. In some samples, we could also trace the main branches of lateral and ventral dendrites, visualizing the over-all 3D structure of the M-cell. Recently we have identified a transgenic line which expresses GFP in the Axon Cap Glia. Combination of the µCT and fluorescent imaging should facilitate the study of the morphology, development and function of the glial structure in the vertebrate.
The experiments at SPring-8 were performed with the approval of JASRI (2018B1039, 2019B1421, 2021A1535, 2021B1697).
Reference)
T Furukawa, EJ Furshpan Two inhibitory mechanisms in the Mauthner neurons of goldfish J Neurophysiol. 1963 26:140-76.
K Hatta, H Korn Physiological properties of the Mauthner system in the adult zebrafish. J Comp Neurol. 1998 15395(4):493-509. | | | |
|
|
