In recent years, with the increasing incidence of cancer, regular physical examination is an important way to find cancer. Nuclear screening is an important method for the diagnosis of gastrointestinal diseases, but it is challenging in the face of small and fuzzy gastrointestinal images. Different from traditional medical objects, pathological slice images are mostly blurry and tiny, which is somewhat difficult to detect and segment. The traditional diagnostic method lacks rapid quantitative analysis and has a certain delay in medical diagnosis, and traditional image processing uses morphological features and pixel distribution to extract features; it is often difficult to achieve the desired effect on small blurry images. This paper proposes a small, microfuzzy pathology detection algorithm based on the attention mechanism; the YOLOv5 is improved under small and micro fuzzy scenarios of the detection of cancer cells in the full field of digital pathology and tests it in the gastric cancer slice dataset. The network structure is improved, and the ability to learn features on small and micro targets is enhanced according to the law of feature distribution. Spatial and channel changes in network attention and attention weight distribution. In the deep blur scenario, the attention mechanism is added to optimize its recognition ability, and the test result shows F1_score is 0.616, and the mAP is 0.611, which can provide the decision support for clinical judgment.
3 April 2020 in Proceedings of the AAAI Conference on Artificial Intelligence
Reinforcement learning (RL) has been widely used to aid training in language generation. This is achieved by enhancing standard maximum likelihood objectives with user-specified reward functions that encourage global semantic consistency. We propose a principled approach to address the difficulties associated with RL-based solutions, namely, high-variance gradients, uninformative rewards and brittle training. By leveraging the optimal transport distance, we introduce a regularizer that significantly alleviates the above issues. Our formulation emphasizes the preservation of semantic features, enabling end-to-end training instead of ad-hoc fine-tuning, and when combined with RL, it controls the exploration space for more efficient model updates. To validate the effectiveness of the proposed solution, we perform a comprehensive evaluation covering a wide variety of NLP tasks: machine translation, abstractive text summarization and image caption, with consistent improvements over competing solutions.
The recognition of medical images, especially endoscopic ultrasound images, has the characteristics of changing images and insignificant gray-scale changes, which requires repeated observation and comparison by medical staff. In view of the above-mentioned characteristics of ultrasound imaging, a system scheme suitable for image processing is proposed, which can analyze the biliary tract, gallbladder, abdominal lymph nodes, liver, descending duodenum, duodenal bulb, stomach, pancreas, pancreatic lymph nodes, there are a total of 10 ultrasonic organs, including 21 kinds of sub-categories and 3510 images. The images are preprocessed using binarization, histogram equalization, median filtering and edge enhancement algorithms. The improved YoloV4 convolutional neural network algorithm is used to train the data set and perform high accuracy is detected in real time. Finally, the average accuracy of this algorithm has reached 91.59%. The algorithm proposed in this paper can make up for the shortcomings of manual detection in the original image detection system, improve the efficiency of detection, and at the same time as an auxiliary system can reduce detection misjudgments, and promote the development of automated and intelligent detection in the medical field.
Placental volume is an important indicator for evaluating fetal health, and automatic identification of placenta and automatic calculation of its volume is of great significance in clinical practice. Due to the complexity and diversity of placental images, domestic and foreign scholars have proposed a variety of image recognition and segmentation methods. Among them, the convolutional neural network method is widely used, but this method has black box theory, which is difficult to explain and requires a large number of training samples. In this context, this paper proposes an improved Canny algorithm based on meshing and OTSU for edge detection, recognition and segmentation of placental images. The placenta volume was calculated by using Gurukin's theorem, and the calculation results were compared with the results of manual calculation by doctors. The results show that the algorithm model proposed in this paper is convenient to calculate, and the error of manual calculation is within 10%. The feasibility of the method has been verified. Especially when the sample size is not enough to use machine learning algorithms, it provides strong support for clinical decision-making.