Identification of Chinese Herb Pieces Based on YOLOv4 / 中国实验方剂学杂志

Chinese herbal piece is an important component of the traditional Chinese medicine （TCM） system， and identifying their quality and grading can promote the development and utilization of Chinese herbal pieces. Utilizing deep learning for intelligent identification of Chinese herbal pieces can save time， effort， and cost， while also reasonably avoiding the constraints of human subjectivity， providing a guarantee for efficient identification of Chinese herbal pieces. In this study， a dataset containing 108 kinds of Chinese herbal pieces （14 058 images） was constructed，the basic YOLOv4 algorithm was employed to identify the 108 kinds of Chinese herbal pieces of our database The mean average precision （mAP） of the developed basic YOLOv4 model reached 85.3%. In addition， the receptive field block was introduced into the neck network of YOLOv4 algorithm， and the improved YOLOv4 algorithm was used to identify Chinese herbal pieces. The mAPof the improved YOLOv4 model achieved 88.7%， the average precision of 80 kinds of decoction pieces exceeded 80%， the average precision of 48 kinds of decoction pieces exceeded 90%. These results indicate that adding the receptive field module can help to some extent in the identification of Chinese herbal medicine pieces with different sizes and small volumes. Finally， the average precision of each kind of Chinese herbal medicine piece by the improved YOLOv4 model was further analyzed. Through in-depth analysis of the original images of Chinese herbal medicine pieces with low prediction average precision， it was clarified that the quantity and quality of original images of Chinese herbal medicine pieces are key to performing intelligent object detection. The improved YOLOv4 model constructed in this study can be used for the rapid identification of Chinese herbal pieces， and also provide reference guidance for the manual authentication of Chinese herbal medicine decoction pieces.

Research on the influence of light with different wavelength on the motion behavior of carp robots / 生物医学工程学杂志

In order to study the effect of light with different wavelengths on the motion behavior of carp robots, phototaxis experiment, anatomical experiment, light control experiment and speed measurement experiment were carried out in this study. Blue, green, yellow and red light with different wavelength were used to conduct phototaxis experiments on carp to observe their movement behavior. By dissecting the skull bones of the carp to determine the appropriate location to carry the light control device, we independently developed a light control carrying device which was suitable for any illumination intensity environment. The experiment of the light-controlled carp robots was carried out. The motion behavior of the carp robot was checked by using computer binocular stereo vision technology. The motion trajectory of the carp robot was tracked and obtained by applying kernel correlation filter (KCF) algorithm. The motion velocity of the carp robot at different wavelengths was calculated according to their motion trajectory. The results showed that carps' sensitivity to different light changed from strong to weak in the order of blue, red, yellow and green, so that using light with different wavelengths to control the speed of the carp robot has certain laws to follow. A new method to avoid brain damage in carp robots control can be provided in this study.

Research on computer aided diagnosis of otitis media based on faster region convolutional neural network / 生物医学工程学杂志

Otitis media is one of the common ear diseases, and its accurate diagnosis can prevent the deterioration of conductive hearing loss and avoid the overuse of antibiotics. At present, the diagnosis of otitis media mainly relies on the doctor's visual inspection based on the images fed back by the otoscope equipment. Due to the quality of otoscope equipment pictures and the doctor's diagnosis experience, this subjective examination has a relatively high rate of misdiagnosis. In response to this problem, this paper proposes the use of faster region convolutional neural networks to analyze clinically collected digital otoscope pictures. First, through image data enhancement and preprocessing, the number of samples in the clinical otoscope dataset was expanded. Then, according to the characteristics of the otoscope picture, the convolutional neural network was selected for feature extraction, and the feature pyramid network was added for multi-scale feature extraction to enhance the detection ability. Finally, a faster region convolutional neural network with anchor size optimization and hyperparameter adjustment was used for identification, and the effectiveness of the method was tested through a randomly selected test set. The results showed that the overall recognition accuracy of otoscope pictures in the test samples reached 91.43%. The above studies show that the proposed method effectively improves the accuracy of otoscope picture classification, and is expected to assist clinical diagnosis.

FILM: finding the location of microaneurysms on the retina

Diabetes retinopathy (DR) is one of the leading cause of blindness among people suffering from diabetes. It is a lesion based disease which starts off as small red spots on the retina. These small red lesions are known as microaneurysms (MA). These microaneurysms gradually increase in size as the DR progresses, which eventually leads to blindness. Thus, DR can be prevented at a very early stage by eliminating the retinal microaneurysms. However, elimination of MA is a two step process. The first step requires detecting the presence of MA on the retina. The second step involves pinpointing the location of MA on the retina. Even though, these two steps are interdependent, there is no model available that can perform both steps simultaneously. Most of the models perform the first step successfully, while the second step is performed by opthamologists manually. Hence we have proposed an object detection model that integrates the two steps by detecting (first step) and pinpointing (second step) the MA on the retina simultaneously. This would help the opthamologists in directly finding the exact location of MA on the retina, thereby simplifying the process and eliminating any manual intervention.

Medical computer-aided detection method based on deep learning / 生物医学工程学杂志

This paper performs a comprehensive study on the computer-aided detection for the medical diagnosis with deep learning. Based on the region convolution neural network and the prior knowledge of target, this algorithm uses the region proposal network, the region of interest pooling strategy, introduces the multi-task loss function: classification loss, bounding box localization loss and object rotation loss, and optimizes it by end-to-end. For medical image it locates the target automatically, and provides the localization result for the next stage task of segmentation. For the detection of left ventricular in echocardiography, proposed additional landmarks such as mitral annulus, endocardial pad and apical position, were used to estimate the left ventricular posture effectively. In order to verify the robustness and effectiveness of the algorithm, the experimental data of ultrasonic and nuclear magnetic resonance images are selected. Experimental results show that the algorithm is fast, accurate and effective.

Vision-based human detection in wounded search task / 医疗卫生装备

Objective To design a vision-based detection method for rotated human bodies to fulfill unmanned wounded search in the rescue operation.Methods HOG (histogram of oriented gradient) which was the most successful visual feature in pedestrian detection was involved in,and the human detection in the wounded search task was realized by multi-directional detection.Furthermore,two human bodies datasets were established by imitating the views of unmanned ground vehicle (UGV)and unmanned aerial vehicle (UAV).Results The application to the two datasets proved the method's feasibility in UGV and UAV.Conclusion The method is robust to the in-plane rotations and out-plane rotations of human bodies,which is of vital significance to promote the efficiency of the wounded searching and rescuing.

Detecção de artefatos estranhos em radiografias de tórax / Foreign artifacts detection in chest radiographs

A radiografia de tórax é um dos exames preconizados para detecção de pneumonia na infância pela Organização Mundial de Saúde. Durante a execução do exame é comum encontrar artefatos nessas imagens radiográficas como:rótulos de identificação, marcas de dedos, botões de camisa, dentre outros, principalmente quando digitalizadas por meio de câmeras fotográficas. Esses artefatos além de tirarem o anonimato da radiografia, afetam significativamente sua análise por sistemas informatizados de detecção e suporte a identificação de doenças. Este trabalho apresenta um método eficiente para identificação dos artefatos, composto de 3 etapas principais: filtragem de pixels baseada em histograma,detecção de bordas com algoritmo de Roberts e filtragem espacial por filtro de desvio padrão. Este método foi experimentado em uma base de 200 imagens e inspecionado visualmente para identificação de erros. Resultados experimentais como, eficiência (tempo processamento/radiografia) ≈ 7ms e precisão de 0,98 demonstram que o método é bastante promissor.

Chest radiography is one of recommended imaging test by World Health Organization for childhood pneumonia diagnosis. However, during patient examination is very usual finding artifacts in these images, such as identification labels, fingerprints, shirt buttons, and so forth. Moreover, when these images are digitally scanned, other problems raise suchas noise, brightness control and so on. Artifacts can reveal private data and expose patient identification. Furthermore, these artifacts can significantly damage automatic analysis by computer diagnosis aided systems. This works presents anefficient method for artifact identification composed by 3 main stages: histogram based pixel filtering, edge detection withRoberts algorithm and standard deviation spacial filtering. This method has been experimented upon 200 images databaseand presented about 7ms of time processing per image. Visually inspection was used to error measuring and we achieve 0,98 of precision. As a result of this, the method demonstrate a very promising preprocessing tool.

Rapid method of object detection based on color feature / 医疗卫生装备

Object detection systems are widely used in many fields. To speed up object detection, a rapid method based on color feature is presented in this paper. Artificial neural network is used for color classification. A series of original objects are gained through searching the most outstanding feature of the marker based on multi-resolution. A set of features obtained from these original objects in the original image, and artificial neural network are used for object classification. Experimental results prove the effectiveness of this method.