Scanning Brain Networks
(Hamamatsu Photonics, Japan). T e data acquisition conditions are summarized in Table 1 .
X-ray images were subjected to a convolution back-projection calculation using the RecView program (available from
http://www.el.u-tokai.ac.jp/
Figure 3: Skeletonized model of the left hemisphere of the fruit fl y brain. The entire model (a) can be divided into CB and OL regions. The upper panel of (a) shows a side view with dorsal side toward the top. The lower panel of (a) shows a front view of the left hemisphere with the dorsal side toward the top The OL (b) can be further divided into the second optic chiasma (c), medulla (d), lobula plate (e), and lobula (f). Structural groups are color-coded. Scale bar = 20 µm.
X-ray tomographic microscopy with Fresnel zone plate (FZP) optics was performed at the BL37XU and BL47XU beamlines. In the BL37XU experiment, a FZP with an outermost zone width of 100 nm and diameter of 310 µm was used as an X-ray objective lens to visualize structures at 160–200 nm resolution. In the BL47XU experiment, a FZP with an outermost zone width of 100 nm and diameter of 774 µm was used as an X-ray objective lens to take images with a viewing field wider than 500 µm. Transmission images were recorded with a CMOS-based imaging detector
14
3D structural groups. Results
Fly brain network . Figure 2a shows the 3D structure of the
fruit fl y brain visualized at the BL37XU beamline. T e spatial resolution of this image was estimated to be 160–200 nm. Neuronal processes were clearly visualized as network structures. In order to analyze the overall network, a skeletonized model was built using a 3D image obtained at BL47XU ( Figure 2b ). T e obtained model consists of neuronal processes with a total length of 378 mm in a volume of 0.220 × 0.328 × 0.314 mm 3 .
www.microscopy-today.com • 2015 September
ryuta/ ). T e obtained tomographic cross sections were stacked in order to reconstruct a 3D image of the sample. Spatial resolution was estimated with 3D square-wave patterns [ 3 ]. Network analysis . In order to analyze the brain network, the structure of each neuron should be described by building a skeletonized model in the 3D image. The model was built by using a method like those used in crystallographic studies of macromolecular structures [ 1 ]. Because it takes considerable person-hours to manually build models of neuronal networks, an initial model was automatically built by machine interpretation of the 3D image [ 1 , 2 ]. First, large structures, such as cell body somas and blood vessels, were scanned to mask those regions from the subsequent tracing procedure. Cartesian coordinates of the large structures were also used for locating neuronal somas. Next, fibriform structures corresponding to dendrites and axons were searched by evaluating the gradient vector flow. The resultant coordi- nates with high fibriform scores were used as starting points to trace the neuronal processes using a Sobel filter [ 1 ]. This procedure was implemented with dedicated software to build models automati- cally [ 1 , 2 ]. The models obtained were manually examined and edited so as to assemble neuronal processes as neurons or to classify them into
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68