Population data-based federated machine learning improves automated echocardiographic quantification of cardiac structure and function: the Automatisierte Vermessung der Echokardiographie project.

Aims: Machine-learning (ML)-based automated measurement of echocardiography images emerges as an option to reduce observer variability. The objective of the study is to improve the accuracy of a pre-existing automated reading tool ('original detector') by federated ML-based re-training. Methods and results: Automatisierte Vermessung der Echokardiographie was based on the echocardiography images of n = 4965 participants of the population-based Characteristics and Course of Heart Failure Stages A-B and Determinants of Progression Cohort Study. We implemented federated ML: echocardiography images were read by the Academic Core Lab Ultrasound-based Cardiovascular Imaging at the University Hospital Würzburg (UKW). A random algorithm selected 3226 participants for re-training of the original detector. According to data protection rules, the generation of ground truth and ML training cycles took place within the UKW network. Only non-personal training weights were exchanged with the external cooperation partner for the refinement of ML algorithms. Both the original detectors as the re-trained detector were then applied to the echocardiograms of n = 563 participants not used for training. With regard to the human referent, the re-trained detector revealed (i) superior accuracy when contrasted with the original detector's performance as it arrived at significantly smaller mean differences in all but one parameter, and a (ii) smaller absolute difference between measurements when compared with a group of different human observers. Conclusion: Population data-based ML in a federated ML set-up was feasible. The re-trained detector exhibited a much lower measurement variability than human readers. This gain in accuracy and precision strengthens the confidence in automated echocardiographic readings, which carries large potential for applications in various settings.

The MAS4AI framework for human-centered agile and smart manufacturing.

Volatility and uncertainty of today's value chains along with the market's demands for low-batch customized products mandate production systems to become smarter and more resilient, dynamically and even autonomously adapting to both external and internal disturbances. Such resilient behavior can be partially enabled by highly interconnected Cyber-Physical Production Systems (CPPS) incorporating advanced Artificial Intelligence (AI) technologies. Multi-agent solutions can provide better planning and control, improving flexibility and responsiveness in production systems. Small modular parts can autonomously take intelligent decisions and react to local events. The main goal of decentralization and interconnectivity is to enable autonomous and cooperative decision-making. Nevertheless, a more efficient orchestration of various AI components and deeper human integration are required. In addition, global behaviors of coalitions of autonomous agents are not easily comprehensible by workers. Furthermore, it is challenging to implement an Industry 4.0 paradigm where a human should be in charge of decision-making and execution. This paper discusses a Multi-Agent System (MAS) where several software agents cooperate with smart workers to enable a dynamic and reconfigurable production paradigm. Asset Administration Shell (AAS) submodels hold smart workers' descriptions in machine-readable format, serving as an integration layer between various system's components. The self-description capability of the AAS supports the system's adaptability and self-configuration. The proposed concept supports the plug-and-produce functionality of the production modules and improves human-machine integration in the shared assembly tasks.

A simple heuristic for Internet-based evidence search in primary care: a randomized controlled trial.

BACKGROUND: General practitioners (GPs) are confronted with a wide variety of clinical questions, many of which remain unanswered. METHODS: In order to assist GPs in finding quick, evidence-based answers, we developed a learning program (LP) with a short interactive workshop based on a simple three-step-heuristic to improve their search and appraisal competence (SAC). We evaluated the LP effectiveness with a randomized controlled trial (RCT). Participants (intervention group [IG] n=20; control group [CG] n=31) rated acceptance and satisfaction and also answered 39 knowledge questions to assess their SAC. We controlled for previous knowledge in content areas covered by the test. RESULTS: Main outcome - SAC: within both groups, the pre-post test shows significant (P=0.00) improvements in correctness (IG 15% vs CG 11%) and confidence (32% vs 26%) to find evidence-based answers. However, the SAC difference was not significant in the RCT. OTHER MEASURES: Most workshop participants rated "learning atmosphere" (90%), "skills acquired" (90%), and "relevancy to my practice" (86%) as good or very good. The LP-recommendations were implemented by 67% of the IG, whereas 15% of the CG already conformed to LP recommendations spontaneously (odds ratio 9.6, P=0.00). After literature search, the IG showed a (not significantly) higher satisfaction regarding "time spent" (IG 80% vs CG 65%), "quality of information" (65% vs 54%), and "amount of information" (53% vs 47%). CONCLUSION: Long-standing established GPs have a good SAC. Despite high acceptance, strong learning effects, positive search experience, and significant increase of SAC in the pre-post test, the RCT of our LP showed no significant difference in SAC between IG and CG. However, we suggest that our simple decision heuristic merits further investigation.

Distributed observers for pose estimation in the presence of inertial sensory soft faults.

Distributed Particle-Kalman Filter based observers are designed in this paper for inertial sensors (gyroscope and accelerometer) soft faults (biases and drifts) and rigid body pose estimation. The observers fuse inertial sensors with Photogrammetric camera. Linear and angular accelerations as unknown inputs of velocity and attitude rate dynamics, respectively, along with sensory biases and drifts are modeled and augmented to the moving body state parameters. To reduce the complexity of the high dimensional and nonlinear model, the graph theoretic tearing technique (structural decomposition) is employed to decompose the system to smaller observable subsystems. Separate interacting observers are designed for the subsystems which are interacted through well-defined interfaces. Kalman Filters are employed for linear ones and a Modified Particle Filter for a nonlinear non-Gaussian subsystem which includes imperfect attitude rate dynamics is proposed. The main idea behind the proposed Modified Particle Filtering approach is to engage both system and measurement models in the particle generation process. Experimental results based on data from a 3D MEMS IMU and a 3D camera system are used to demonstrate the efficiency of the method.

Cascaded Kalman and particle filters for photogrammetry based gyroscope drift and robot attitude estimation.

Based on a cascaded Kalman-Particle Filtering, gyroscope drift and robot attitude estimation method is proposed in this paper. Due to noisy and erroneous measurements of MEMS gyroscope, it is combined with Photogrammetry based vision navigation scenario. Quaternions kinematics and robot angular velocity dynamics with augmented drift dynamics of gyroscope are employed as system state space model. Nonlinear attitude kinematics, drift and robot angular movement dynamics each in 3 dimensions result in a nonlinear high dimensional system. To reduce the complexity, we propose a decomposition of system to cascaded subsystems and then design separate cascaded observers. This design leads to an easier tuning and more precise debugging from the perspective of programming and such a setting is well suited for a cooperative modular system with noticeably reduced computation time. Kalman Filtering (KF) is employed for the linear and Gaussian subsystem consisting of angular velocity and drift dynamics together with gyroscope measurement. The estimated angular velocity is utilized as input of the second Particle Filtering (PF) based observer in two scenarios of stochastic and deterministic inputs. Simulation results are provided to show the efficiency of the proposed method. Moreover, the experimental results based on data from a 3D MEMS IMU and a 3D camera system are used to demonstrate the efficiency of the method.