We developed a novel ensemble three-dimensional residual network (E3D-ResNet) for the reduction of false positives (FPs) in computer-aided detection (CADe) of polyps on CT colonography (CTC). To capture the volumetric multiscale information of CTC images, each polyp candidate was represented with three different sizes of volumes of interest (VOIs), which were enlarged to a common size and were individually subjected to three 3D-ResNets. These 3D-ResNets were trained to calculate three polyp-likelihood probabilities, p1, p2 and p3, corresponding to each input VOI. The final polyp likelihood, p, was obtained as the maximum of p1, p2 and p3. We compared the classification performance of the E3D-ResNet with that of a non-ensemble 3D-ResNet, ensemble 2D-ResNet, and ensemble of 2D- and 3D-convolutional neural network (CNN) models. All models were trained and evaluated with 21,021 VOIs of polyps and 19,557 VOIs of FPs that were sampled with data augmentation from the CADe detections on the CTC data of 20 patients. We evaluated the classification performance of the models with receiver operating characteristics (ROC) analysis using cross-validation, where the area under the ROC curve (AUC) was used as the figure of merit. Preliminary results showed that AUC value (0.98) of the E3D-ResNet was significantly higher than that of the reference models (P < 0.001), indicating that the E3D-ResNet has the potential of substantially reducing the FPs in CADe of polyps on CTC.
In this paper, we design a novel deep-sea near infrared light based imaging equipment for deep-sea mining observation systems. The spectral sensitivity peaks are in the red region of the invisible spectrum, ranging from 750nm to 900nm. In addition, we propose a novel underwater imaging model that compensates for the attenuation discrepancy along the propagation path. The proposed model fully considered the effects of absorption, scattering and refraction. We also develop a locally adaptive Laplacian filtering for enhancing underwater transmission map after underwater dark channel prior estimation. Furthermore, we propose a spectral characteristic-based color correction algorithm to recover the distorted color. In water tank experiments, we made a linear scale of eight turbidity steps ranging from clean to heavily scattered by adding deep sea soil to the seawater (from 500 to 2000 mg/L). We compared the results of different turbidity underwater scene, illuminated alternately with near infrared light vs. white light. Experiments demonstrate that the enhanced NIR images have a reasonable noise level after the illumination compensation in the dark regions and demonstrates an improved global contrast by which the finest details and edges are significantly enhanced. We also demonstrate that the effective distance of the designed imaging system is about 1.5 meters, which can meet the requirement of micro-terrain observation around the deep-sea mining systems. Remotely Operated Underwater Vehicle (ROV)-based experiments also certified the effectiveness of the proposed method.
Light curtain systems are used to detect intruders in various cases and places. However, it is necessary to adjust the position of the light detecting element accurately in order to receive the irradiating laser light. We propose a new type safety light curtain system that uses a hemispherical mirror and an LED in this research. A hemispherical mirror can reflect irradiating light rays surroundings of 180° in the vertical direction and 360° in the horizontal direction. When an LED is at a position that is higher than the hemispherical mirror, the LED irradiating light can be reflected by the hemispherical mirror, even if the LED is arbitrarily set up. In the case that, the light of LED is intercepted when an intruder passes between the LED and the hemispherical mirror, the output voltage of the light detecting element decreases. We can set a proper threshold voltage value of the detecting element to judge whether an intruder passes or not. Our system uses a PSOC microcomputer to judge the output voltage of the receiving element with the threshold voltage value. In addition, the LED output light is modulated by 10kHz in order to avoid the influence of the surrounding turbulence light. Our experiment succeeded to detect intruder using the proposed system without accurate light axis setting.
Proceedings Volume Editor (1)
This will count as one of your downloads.
You will have access to both the presentation and article (if available).