How to contour the different heart subregions for future deep-learning modeling of the heart: A practical pictorial proposal for radiation oncologists.

There are currently no accurate rules for manually delineating the subregions of the heart (cavities, vessels, aortic/mitral valves, Planning organ at Risk Volumes for coronary arteries) with the perspective of deep-learning based modeling. Our objective was to present a practical pictorial view for radiation oncologists, based on the RTOG atlas and anatomical complementary considerations for the cases where the RTOG guidelines are missing.

Auto-MatRegressor: liberating machine learning alchemists.

Machine learning (ML) is widely used to uncover structure-property relationships of materials due to its ability to quickly find potential data patterns and make accurate predictions. However, like alchemists, materials scientists are plagued by time-consuming and labor-intensive experiments to build high-accuracy ML models. Here, we propose an automatic modeling method based on meta-learning for materials property prediction named Auto-MatRegressor, which automates algorithm selection and hyperparameter optimization by learning from previous modeling experience, i.e., meta-data on historical datasets. The meta-data used in this work consists of 27 meta-features that characterize the datasets and the prediction performances of 18 algorithms commonly used in materials science. To recommend optimal algorithms, a collaborative meta-learning method embedded with domain knowledge quantified by a materials categories tree is designed. Experiments on 60 datasets show that compared with the traditional modeling method from scratch, Auto-MatRegressor automatically selects appropriate algorithms at lower computational cost, which accelerates constructing ML models with good prediction accuracy. Auto-MatRegressor supports dynamic expansion of meta-data with the increase of the number of materials datasets and other required algorithms and can be applied to any ML materials discovery and design task.

Development of an automatic modeling method for patient-specific aortic graft reconstruction guide in thoracoabdominal aortic repair.

BACKGROUND AND OBJECTIVES: Because repairing visceral and segmental arteries in open surgical repair for thoracoabdominal aortic aneurysms is essential, two types of patient-specific graft reconstruction guides for reconstruction in the operating room have been developed that are applied clinically. However, designing the patient-specific guides is a time-consuming, laborious task. The aim of this study was to develop an automatic modeling method and to evaluate its accuracy. METHODS: In 10 patients with thoracoabdominal aortic aneurysms, computer-aided designing was performed with conventional and automatic modeling methods for aortic reconstruction guides as follows: 1) a visualizing guide that presented the accurate shape of the aortic graft, visualizing the main aortic body and major blood vessels; and 2) a marking guide wherein the vessels in the visualizing guide were replaced by the protruding marking regions detectable by tactile sense. The script-based automatic guide modeling program was developed using an application programming interface presented in the 3-matic software with Python. For accuracy, the absolute mean differences of both modeling methods were assessed using Hausdorff distance. The modeling between conventional and automatic modeling methods was compared and evaluated using the Wilcoxon signed-rank test. RESULTS: The absolute mean difference between the conventional and automatic modeling methods were 6.05 ± 4.86 µm for the visualizing guide and 5.51 ± 4.85 µm for the marking guide. For the visualizing guide, the modeling time of the automatic modeling method was reduced by approximately more than thirtyfold than the conventional modeling method (p<0.001). The marking guide was reduced about fortyfold (p<0.001). CONCLUSIONS: Compared to the conventional method, the automatic modeling method was demonstrated to reduce the modeling time with reasonable accuracy, which could lead to a more efficient modeling and clinical application.

Multi-Agent-Based Urban Vegetation Design.

Urban vegetation is an essential element of the urban city pedestrian walkway. Despite city forest regulations and urban planning best practices, vegetation planning lacks clear comprehension and compatibility with other urban elements surrounding it. Urban planners and academic researchers currently devote vital attention to include most of the urban elements and their impact on the occupants and the environment in the planning stage of urban development. With the advancement in computational design, they have developed various algorithms to generate design alternatives and measure their impact on the environment that meets occupants' needs and perceptions of their city. In particular, multi-agent-based simulations show great promise in developing rule compliance with urban vegetation design tools. This paper proposed an automatic urban vegetation city rule compliance approach for pedestrian pathway vegetation, leveraging multi-agent system and algorithmic modeling tools. This approach comprises three modules: rule compliance (T-Rule), street vegetation design tool (T-Design), and multi-agent alternative generation (T-Agent). Notably, the scope of the paper is limited to trees, shrubbery, and seating area configurations in the urban pathway context. To validate the developed design tool, a case study was tested, and the vegetation design tool generated the expected results successfully. A questionnaire was conducted to give feedback on the use of the developed tool for enhancing positive experience of the developed tool. It is anticipated that the proposed tool has the potential to aid urban planners in decision-making and develop more practical vegetation planting plans compared with the conventional Two-Dimensional (2D) plans, and give the city occupants the chance to take part in shaping their city by merely selecting from predefined parameters in a user interface to generate their neighborhood pathway vegetation plans. Moreover, this approach can be extended to be embedded in an interactive map where city occupants can shape their neighborhood greenery and give feedback to urban planners for decision-making.

Influencing Factors of Femoro-Tibial Contact Area for Knee Prostheses / 医用生物力学

Objective By developing an automatic procedure for optimization of femoro-tibial contact area for knee prosthesis, to summarize the influence pattern of design parameters on contact area, and discover the relationship between the maximum contact stress and contact area. Methods A parametric finite element (FE) model was developed in the Isight software, which included three components: automatic parameter changes for the geometric model, automatic modeling in the FE software, and automatic FE calculation. The automatic workflow was realized, and then contact areas were statistically analyzed. Results The FE model was validated by using Tekscan pressure distribution system. When the femoral sagittal radius was gradually close to the tibial sagittal radius, the contact area gradually reached to the maximum 295 mm2. The femoral sagittal radius had a positive effect on contact area, while the tibial sagittal radius had a negative effect. The maximum contact stress had a linear relationship with contact area approximately. Conclusions This study analyzed the influence of femoro-tibial sagittal radius on contact stress and contact area, and the research findings would provide references for the design on reducing wear of tibial insert in clinic.

An Evolutionary Search Algorithm for Covariate Models in Population Pharmacokinetic Analysis.

Building a covariate model is a crucial task in population pharmacokinetics. This study develops a novel method for automated covariate modeling based on gene expression programming (GEP), which not only enables covariate selection, but also the construction of nonpolynomial relationships between pharmacokinetic parameters and covariates. To apply GEP to the extended nonlinear least squares analysis, the parameter consolidation and initial parameter value estimation algorithms were further developed and implemented. The entire program was coded in Java. The performance of the developed covariate model was evaluated for the population pharmacokinetic data of tobramycin. In comparison with the established covariate model, goodness-of-fit of the measured data was greatly improved by using only 2 additional adjustable parameters. Ten test runs yielded the same solution. In conclusion, the systematic exploration method is a potentially powerful tool for prescreening covariate models in population pharmacokinetic analysis.

Application of workflow technology to construct automatic rapid drug testing system / 中国药学杂志

OBJECTIVE: To introduce the implementation method and design ideas to achieve automatic NIR (near-infrared spectroscopy) modeling and comparison function during the construction of "National Rapid Drug Testing Database Network Platform" and investigate and verify the feasibility of constructing the automatic rapid drug testing module using workflow technology. METHODS: Both the scientific data workflow and analysis tool Pipeline Pilot and the basic principle of two quick comparison algorithms of NIR were employed to construct a spectral processing and modeling workflow of drugs. Moreover, validation and comparison were carried out between our method and the instrument workstation software OPUS. RESULTS: The established workflow method was not only consistent with the results of OPUS, but also had the advantages of automation and easinessto use. CONCLUSION: The scientific workflow technology can be used to achieve automatic modeling and comparison function of NIR and easily introduce other chemometrics method in the future.