Abstract Human activity recognition (HAR) research has recently focused on multiple individuals within videos. However, conventional models are trained using supervised or semi-supervised learning, which makes their direct application to real-world videos challenging. The purpose of this study is to achieve HAR from real-world videos through completely unsupervised learning. As real-world videos, we target operating room surveillance videos with surgery ongoing. We extract visual features using two autoencoders based on Inception 3D (I3D) and spatial features measured by the L2 norm between the operating table and individuals. These individual features were clustered using a centroid-based model. We evaluated our method on 29 pieces of different operating room videos from 6 different operating rooms in 145 seconds in total, and achieved 0.83 and 0.71 of accuracy in the training and test datasets, respectively, for individual clustering. Our method allows automatic analysis of operating room videos, which contributes to improving the efficiency and effectiveness of postoperative analysis and further medical education.

Most real-world applications in EHRs involve temporal data with skewed distributions. The imbalanced classification problem becomes more difficult in sporadic temporal data that variables exist on correlation and have some missing values. A common solution to classification tasks with imbalanced data is the oversampling methods, which generate new samples to re-balancing the classes. However, traditional oversampling methods usually change the distribution, thereby leading to bias. This paper proposed a self-adaptive integrated oversampling method for imbalanced sporadic temporal data in EHRs. The masking vectors and density vectors have been introduced to measure missing value distribution of samples, and the minority samples are divided into high density samples and sparse density samples. We extend the resampling strategies combining a subsample alignment method and structure preserving oversampling method. The weight of sample difference is used to improve classification performance. Furthermore, the filter mechanism is proposed to remove the noise samples with good efficiency. The experimental results show that the proposed method increases performance compared to traditional resampling methods in terms of AUC, F1, and G-mean evaluation metrics.

In this paper, we present an agent-based completion that cooperates with a physician to optimize parameter-text pairs collection from a medical note. Currently, in hospitals, there is a medical coding process, which collects certain parameter-text pairs from the medical note's narrative text. The process starts with categorizing the text into certain parameter-text pairs. Then, the collected pairs are studied by the coders to produce correct medical codes. However, since the physicians may not aware of the coders' requirements, the medical coding process is quite problematic, such as lack of parameter-text pairs. To address this problem, we propose an agent-based completion, which represents the coder's view to categorize the text into certain parameter-text pairs and recommend the parameters to be filled. This paper shows a basic design of the agent and the background technologies to support the completion system.

OBJECTIVE: For chronic diseases, medical history reconstruction is essential for retrospective database analyses. One important aspect is determining which prescriptions belong to the same episode. However, a standard framework for this task is still lacking, particularly for multitherapy datasets. This paper presents a medication episode construction framework for the medical history of patients with chronic diseases. METHODS: Allen's relaxed temporal relations (i.e., temporal relations with time constraints relaxed by ) is used to define the consecutive prescription relations considering the patients' behavior. For example, patients occasionally arrive earlier or later than their appointment. RESULTS: influences the generation of stable periods (i.e., periods of time, at least three months, in which a medication is continuously taken by a patient). When using the lowest selected value (7 days), considerably fewer shorter stable periods (for durations less than 300 days) are produced and more longer stable periods are produced compared to cases without using . Furthermore, the results show that by using , regarding the number of events, where a stable period continues the previous stable period, decreases and the number of medication transition events available to be observed increases. CONCLUSION: Using in medication episode construction from multitherapy prescription datasets enables the longer expression of short-duration fragmented prescriptions and pruning repetitive prescriptions. SIGNIFICANCE: Our proposed framework is designed for multitherapy datasets, which has not been addressed by previous studies. The concept of relaxes the prescription relation against noise caused by the patient behavior and consequently provides a compact, but informative search space for observing medication transition events in a longitudinal analysis.

Walking 8,000 steps in a day is one of the important criteria to maintain our health. However, we often miss a chance to walk due to the difficulty to keep our motivation toward our health in a daily life. We propose an algorithm to search an appropriate daily walking pattern from the user's past walking record. The searched walking patterns are used for making a health promotion agent recommend an effective timing to walk. With this recommendation, users will not miss the timing when they can walk. In this study, we focused on designing the algorithm for searching a daily walking pattern, which satisfied both conditions, achieving 8,000 steps a day and being similar to the current user walking record. In a pilot performance study, it was revealed that the proposed algorithm can narrow down the walking pattern candidates and properly search the walking pattern similar to the current user walking record.

This paper proposes The Rainbow Marker, a planar marker for estimating the direction of a light source using a structural color. A structural color is a color produced by microscopically structured surfaces that vary in appearance according to the viewpoint, the direction and the spectrum of the light source. The proposed marker contains a planar material which causes structural coloration. The direction of the light source is estimated by structural color pattern matching between an input pattern and referential color patterns. In this paper, two types of the marker were implemented, with a grating sheet and with a holographic sheet, to demonstrate that the proposed method is applicable in the field of augmented reality.

A Hospital Information Systems (HIS) have turned a hospital into a gigantic computer with huge computational power, huge storage and wired/wireless local area network. On the other hand, a modern medical device, such as echograph, is a computer system with several functional units connected by an internal network named a bus. Therefore, we can embed such a medical device into the HIS by simply replacing the bus with the local area network. This paper designed and developed two embedded systems, a ubiquitous echograph system and a networked digital camera. Evaluations of the developed systems clearly show that the proposed approach, embedding existing clinical systems into HIS, drastically changes productivity in the clinical field. Once a clinical system becomes a pluggable unit for a gigantic computer system, HIS, the combination of multiple embedded systems with application software designed under deep consideration about clinical processes may lead to the emergence of disruptive innovation in the clinical field.

The fine concavo-convex shape of the object surface is a critical component that determines its texture. A previous electrotactile display could not tactually present fine concavo-convex objects but only a line pattern on the plane. In this study, the authors proposed a physiology-based tactile mapping method by formulating nerve activities that stem from a mechanical stimulus. The proposed electrical stimulus allows a human to perceive a concavo-convex shape. A stimulus frequency coded the skin displacement given by a concavo-convex shape and represented the adaptation of receptors with exponential approximation. As an example of tactually presenting a concavo-convex shape, roughness was introduced by using a Brownian surface, which enables a user to touch a surface of varying degrees of roughness with a developed tactile display mounted on a computer mouse. The results of subjective experiments showed a statistical difference among the perceived roughness of displayed rough surfaces. This study concluded that the proposed tactile mapping method was effective to present a touch sensation for concavo-convex shapes corresponding to a stimulus frequency below 100 Hz.