REMODEL: Rethinking Deep CNN Models to Detect and Count on a NeuroSynaptic System.

In this work, we perform analysis of detection and counting of cars using a low-power IBM TrueNorth Neurosynaptic System. For our evaluation we looked at a publicly-available dataset that has overhead imagery of cars with context present in the image. The trained neural network for image analysis was deployed on the NS16e system using IBM's EEDN training framework. Through multiple experiments we identify the architectural bottlenecks present in TrueNorth system that does not let us deploy large neural network structures. Following these experiments we propose changes to CNN model to circumvent these architectural bottlenecks. The results of these evaluations have been compared with caffe-based implementations of standard neural networks that were deployed on a Titan-X GPU. Results showed that TrueNorth can detect cars from the dataset with 97.60% accuracy and can be used to accurately count the number of cars in the image with 69.04% accuracy. The car detection accuracy and car count (-/+ 2 error margin) accuracy are comparable to high-precision neural networks like AlexNet, GoogLeNet, and ResCeption, but show a manifold improvement in power consumption.

Activation of disulfide bond cleavage triggered by hydrophobization and lipophilization of functionalized dihydroasparagusic acid.

Concisely synthesized and functionalized dihydroasparagusic acid (DHAA) derivatives were used to show that the introduction of a hydrophobic functional group dramatically reduced air oxidation activity at the dithiol moieties and dominantly activated the cleavage of S-S bonds in proteins, presumably due to the hydrophobization and lipophilization. Notably, the reaction sites of water-reactive dithiol moieties behaved similarly to hydrophobic and lipophilic functional groups, which suggests impersonation of the reaction site.

[Study of CT Automatic Exposure Control System (CT-AEC) Optimization in CT Angiography of Lower Extremity Artery by Considering Contrast-to-Noise Ratio].

PURPOSE: To evaluate the image quality and effect of radiation dose reduction by setting for computed tomography automatic exposure control system (CT-AEC) in computed tomographic angiography (CTA) of lower extremity artery. METHODS: Two methods of setting were compared for CT-AEC [conventional and contrast-to-noise ratio (CNR) methods]. Conventional method was set noise index (NI): 14and tube current threshold: 10-750 mA. CNR method was set NI: 18, minimum tube current: (X+Y)/2 mA (X, Y: maximum X (Y)-axis tube current value of leg in NI: 14), and maximum tube current: 750 mA. The image quality was evaluated by CNR, and radiation dose reduction was evaluated by dose-length-product (DLP). RESULTS: In conventional method, mean CNRs for pelvis, femur, and leg were 19.9±4.8, 20.4±5.4, and 16.2±4.3, respectively. There was a significant difference between the CNRs of pelvis and leg (P<0.001), and between femur and leg (P<0.001). In CNR method, mean CNRs for pelvis, femur, and leg were 15.2±3.3, 15.3±3.2, and 15.3±3.1, respectively; no significant difference between pelvis, femur, and leg (P=0.973) in CNR method was observed. Mean DLPs were 1457±434 mGy⋅cm in conventional method, and 1049±434 mGy·cm in CNR method. There was a significant difference in the DLPs of conventional method and CNR method (P<0.001). CONCLUSION: CNR method gave equal CNRs for pelvis, femur, and leg, and was beneficial for radiation dose reduction in CTA of lower extremity artery.

Air-stable cationic gold(I) catalyst featuring a Z-type ligand: promoting enyne cyclizations.

An air-stable cationic Au(I) complex featuring a Z-type ligand (boron atom) as a σ-acceptor was developed for elucidating the effect of B on catalytic reactions. An enyne cyclization in the presence of either [Au→B](+) or [Au](+) showed that [Au→B](+) promotes the reactivity, which enabled the effective construction of not only five- and six-membered rings, but also seven-membered rings.

[Usefulness of subtraction computed tomography angiography employing orbital synchronized helical scanning for diagnosis of lower extremity arteries with vessel wall calcification].

PURPOSE: Massive calcification complicates the diagnosis of the blood vessel lumen in computed tomography angiography (CTA) of the arteries of the lower extremities. The purpose of this study was to evaluate subtraction CTA with the use of orbital synchronized helical scanning (OS-SCTA). METHOD: Phantom study: We performed OS-SCTA and non-OSCTA of a calcified vessel phantom (ψ2.5 mm), and compared them with a non-calcified vessel phantom as the reference by full width at half maximum (FWHM) and full width at tenth maximum (FWTM) of maximum intensity projection (MIP) images. Clinical study: 58 patients with peripheral artery disease who were referred for angiography also underwent OS-SCTA. OS-SCTA was produced using MIP images. Findings were graded according to three categories: (1) stenosis greater than 50% or occluded; (2) stenosis less than 50%; (3) not detected due to insufficient image quality. OS-SCTA findings were compared with the angiographic findings for each arterial segment. RESULTS: In the phantom study, FWHM showed no significant difference between OS-SCTA and the reference (P=0.135), whereas FWTM showed a significant difference (P<0.001). FWHM and FWTM showed a significant difference between non-OS-SCTA and the reference (P<0.001), due to misregistration with helical artifacts. In a clinical study comparing OS-SCTA with angiography, the sensitivity and specificity were 93.3% and 95.1% in calcified segments, 91.8% and 93.9% in non-calcified segments, and 92.2% and 94.6% in all segments. There was no significant difference between calcified segments and non-calcified segments (sensitivity: P=0.568, specificity: P=0.549). CONCLUSION: OS-SCTA is beneficial for the diagnosis of lower extremity arteries with vessel wall calcification, since it shows detection accuracy comparable to that of angiography.

[Optimization of the chest computed tomography scan by varying the position of the arms].

Computed tomography automatic exposure control (CT-AEC) technique is calculated from a localizer radiograph. When we perform neck and chest CT examination, at first, we acquire localizer radiograph and neck images by placing the arm in a lowered position. Next, the arm is raised for the chest scan. Therefore, the localizer radiograph and subject information are different in the chest scan. In this situation, the chest scan with the use of the CT-AEC causes radiation over-dose. The purpose of this study is to optimize the CT-AEC by controlling noise index (NI), and make a chest CT scan condition considering the position of the arms. We measured the image noise (SD) in the phantom by using CT-AEC. In addition, dose length product (DLP) was recorded. Moreover, we examined the correlation with the clinical images. The results of our experiments show that radiation dose can be reduced with the image quality kept by controlling NI.