Pathological is crucial to cancer diagnosis. Usually, Pathologists draw their conclusion based on observed cell and tissue structure on histology slides. Rapid development in machine learning, especially deep learning...
详细信息
ISBN:
(纸本)9781728118680
Pathological is crucial to cancer diagnosis. Usually, Pathologists draw their conclusion based on observed cell and tissue structure on histology slides. Rapid development in machine learning, especially deep learning have established robust and accurate classifiers. They are being used to analyze histopathological slides and assist pathologists in diagnosis. Most machine learning systems rely heavily on annotated data sets to gain experiences and knowledge to correctly and accurately perform various tasks such as classification and segmentation. Generally, annotations made in pathology-related datasets have inherited annotation methods from natural scene images. This work investigates different granularity of annotations in histopathological data set including image-wise, bounding box, ellipse-wise, and pixel-wise to verify the influence of annotation in pathological slide on deep learning models. We design corresponding experiments to test classification and segmentation performance of deep learning models based on annotations with different annotation granularity. In classification, state-of-the-art deep learning-based classifiers perform better when trained by pixel-wise annotation dataset. On average, precision, recall and F1-score improves by 7.87%, 8.83% and 7.85% respectively. Thus, it is suggested that finer granularity annotations are better utilized by deep learning algorithms in classification tasks. Similarly, semantic segmentation algorithms can achieve 8.33% better segmentation accuracy when trained by pixel-wise annotations. Our study shows not only that finer-grained annotation can improve the performance of deep learning models, but also help they extract more accurate phenotypic information from histopathological slides. The accurate and spatially precise acquisitions of phenotypic information can improve the reliability of the model prediction. Intelligence systems trained on granular annotations may help pathologists inspecting certain regions a
In the process of infrared spectrum analysis, wavebands selections is very important to deduce the dimension of spectrum data and improve the accuracy of analysis model. There are some methods are used in wavebands se...
详细信息
In the process of infrared spectrum analysis, wavebands selections is very important to deduce the dimension of spectrum data and improve the accuracy of analysis model. There are some methods are used in wavebands selections, such asGenetic Algorithm (GA) andParticle Swarm Optimization (PSO), variance analysis method, correlation coefficient method, Uninformative Variables Elimination (UVE) method, interval Partial Least Squares (iPLS) method, stepwise regression method, etc.. But most of these methods are used in nearly infrared spectrum analysis, and are not good enough in waveband selection. In this paper, Cloud-Based Adaptive Particle Swarm Optimization (CAPSO) algorithm is introduced to select the waveband. Its optimization process is based on cloud theory in the fuzzy control field. According to the prior computation result, it can adjust the current evolution strategy using three ways . Due to this strategy, CAPSO algorithm can reduce the number of particle easily and the convergence speed is faster than the other compared algorithms. Experiments showed that CAPSO algorithm has clear advantage in optimization time, the dimension of the selected spectrum data and the accuracy of the analysis model compared with PSO and GA.
The statistic information of connected components are fundamental for imageprocessing, which could be acquired through connected components labeling. This paper proposes a hardware-efficient method for extracting sta...
详细信息
The statistic information of connected components are fundamental for imageprocessing, which could be acquired through connected components labeling. This paper proposes a hardware-efficient method for extracting statistic information of connected components in a binary image to accelerate imageprocessing in embedded application. The proposed method scans two adjacent rows with 2 x 2 template simultaneously, meanwhile, statistic information of runs are recorded. After scanning two rows, the equivalent runs are merged, and then statistic information of completed connected region is exported directly. This method scans an image only once, which could reduce off-chip memory access massively. For a determined image resolution, the requirement of on-chip memory resource is also confirmed and not affected by the number of connected components. This algorithm is modeled with Verilog, and the simulation result shows that average processing speed could be real-time for various images with different resolution. Furthermore, the memory cost is little compared to other hardware based algorithms for labeling connected components, and the proposed method is appropriated for hardware implementation.
An optimization algorithm for image recovery is a core issue in the field of compressive sensing (CS). This paper deeply studied the CS reconstruction algorithm based on split Bregman iteration with ℓ 1 norm, which e...
详细信息
ISBN:
(纸本)9781728102481;9781728102474
An optimization algorithm for image recovery is a core issue in the field of compressive sensing (CS). This paper deeply studied the CS reconstruction algorithm based on split Bregman iteration with ℓ 1 norm, which enables the ℓ 1 norm to approximate the original ℓ 0 norm during the optimization process. Consequently, we proposed another novel algorithm improving the precision and the convergence speed based on split quadratic Bregman iteration (SQBI) with ℓ 0 norm. Besides, we analyzed its convergence by proving two monotonically decreasing theorems. Inspired by previous researches, we applied smoothed ℓ 0 norm for the optimization problem to replace the traditional ℓ 0 norm in CS. The improvement is made by using a Gaussian function to approximate the ℓ 0 norm, transforming it into a convex optimization problem, and eventually achieved a convergent solution by the steepest descent method. The experimental results show that under the same conditions, compared with other state-of-the-art algorithms, the reconstruction accuracy of the CS reconstruction algorithm based on the SQBI with smoothed ℓ 0 norm is improved significantly, and its convergence rate is also accelerated as well.
We propose a new ballistic imaging method that is capable of imaging an object through an intense scattering medium. In this method, a femtosecond supercontinuum and a roundabout spatial gate were used to suppress spe...
详细信息
We propose a new ballistic imaging method that is capable of imaging an object through an intense scattering medium. In this method, a femtosecond supercontinuum and a roundabout spatial gate were used to suppress speckles and filter background noise, respectively. The roundabout spatial gate extracts ballistic light and avoids low- pass spatial filtering to ensure the high resolution of images. The experimental results showed that even when the optical depth of the scattering medium reached 17, the images extracted by the method had improved identifiability and contrast. (C) 2017 Optical Society of America
Haze removal is useful in computational photography and computer vision applications. Although many haze removal algorithms have been proposed, their computational efficiency requires improvement. A real-time haze rem...
详细信息
Haze removal is useful in computational photography and computer vision applications. Although many haze removal algorithms have been proposed, their computational efficiency requires improvement. A real-time haze removal method is presented in this paper. The method is based on the concept of a dark channel prior. To enhance the haze removal performance, an approximate method to estimate the atmospheric light and transmission is employed. For embedded system applications, a hardware architecture to perform real-time haze removal is proposed. The hardware can achieve 116 MHz on Stratix FPGA. The simulation results indicate that the hardware is highly efficient and performs well. It obtains good image recovery results and satisfies the real-time requirement even for large images.
This paper focuses on the detection of crowd abnormal behaviors in surveillance systems. By an improved Harris feature point extraction method in multi-scale space, feature points needed for crowd anomaly behavior det...
详细信息
This paper focuses on the detection of crowd abnormal behaviors in surveillance systems. By an improved Harris feature point extraction method in multi-scale space, feature points needed for crowd anomaly behavior detection are extracted. A feature point optimization method based on the motion foreground extraction algorithm is designed. The optimized feature points are classified according to the motion attributes. In the crowd abnormality behavior detection stage, a comprehensive abnormality decision method combining speed and direction is given. Experiment results shows this method has good adaptability in different scenarios.
暂无评论