Increasing Robustness of Brain-Computer Interfaces Through Automatic Detection and Removal of Corrupted Input Signals.

For brain-computer interfaces (BCIs) to be viable for long-term daily usage, they must be able to quickly identify and adapt to signal disruptions. Furthermore, the detection and mitigation steps need to occur automatically and without the need for user intervention while also being computationally tractable for the low-power hardware that will be used in a deployed BCI system. Here, we focus on disruptions that are likely to occur during chronic use that cause some recording channels to fail but leave the remaining channels unaffected. In these cases, the algorithm that translates recorded neural activity into actions, the neural decoder, should seamlessly identify and adjust to the altered neural signals with minimal inconvenience to the user. First, we introduce an adapted statistical process control (SPC) method that automatically identifies disrupted channels so that both decoding algorithms can be adjusted, and technicians can be alerted. Next, after identifying corrupted channels, we demonstrate the automated and rapid removal of channels from a neural network decoder using a masking approach that does not change the decoding architecture, making it amenable for transfer learning. Finally, using transfer and unsupervised learning techniques, we update the model weights to adjust for the corrupted channels without requiring the user to collect additional calibration data. We demonstrate with both real and simulated neural data that our approach can maintain high-performance while simultaneously minimizing computation time and data storage requirements. This framework is invisible to the user but can dramatically increase BCI robustness and usability.

Analysis of proportional data in reproductive and developmental toxicity studies: Comparison of sensitivities of logit transformation, arcsine square root transformation, and nonparametric analysis.

BACKGROUND: In developmental and reproductive toxicity studies, analysis of litter-based binary endpoints (e.g., incidence of malformed fetuses) is complex in that littermates often are not entirely independent of one another. It is well established that the litter, not the individual fetus, is the proper independent experimental unit in statistical analysis. Accordingly, analysis is often based on the proportion affected per litter and the litter proportions are analyzed as continuous data. Because these proportional data generally do not meet assumptions of symmetry or normality, data are typically analyzed by nonparametric methods, arcsine square root transformation, or logit transformation. METHODS: We conducted power calculations to compare different approaches (nonparametric, arcsine square root-transformed, logit-transformed, untransformed) for analyzing litter-based proportional data. A reproductive toxicity study with a control and one treated group provided data for two endpoints: prenatal loss, and fertility by in utero insemination (IUI). Type 1 error and power were estimated by 10,000 simulations based on two-sample one-tailed t tests with varying numbers of litters per group. To further compare the different approaches, we conducted additional analyses with shifted mean proportions to produce illustrative scenarios. RESULTS: Analyses based on logit-transformed proportions had greater power than those based on untransformed or arcsine square root-transformed proportions, or nonparametric procedures. CONCLUSION: The logit transformation is preferred to the other approaches considered when making inferences concerning litter-based proportional endpoints, particularly with skewed distributions. The improved performance of the logit transformation becomes increasingly pronounced as the response proportions are increasingly close to the boundaries of the parameter space.

Estimation of individual rodent water consumption from group consumption data for gestation, lactation, and postweaning life stages using linear regression models.

In rodent bioassays where chemicals are administered in the drinking water, water consumption data for individual animals are needed to estimate chemical exposures accurately. If multiple animals share a common water source, as occurs in some studies, only the total amount of drinking water consumed by all animals utilizing the common source is directly measurable, and water consumption rates for individual animals are not available. In the Four Lab Study of the US Environmental Protection Agency, which included a multigenerational rodent bioassay, a complex mixture of drinking water disinfection by-products was delivered to multiple Sprague-Dawley rats from a common drinking water container. To estimate disinfection by-product mixture exposure for each animal, authors developed four log-linear regression models to allocate water consumption among rats sharing a common water container. The four models represented three animal lifestages: Gestation, Lactation, and Postweaning, with separate Postweaning models for male and female. Authors used data from six Sprague-Dawley rat bioassays to develop these models from available individual cage data for the Postweaning models, and available individual animal data for the Gestation and Lactation models. The r(2) values for the model fits were good, ranging from 0.67 to 0.92. The Gestation and Lactation models were generally quite accurate in predicting average daily water consumption whereas the Postweaning models were less robust. These models can be generalized for use in other reproductive and developmental bioassays where common water sources are used and data on the explanatory variables are available.

Application of site-specific calibration data using the CALUX by XDS bioassay for dioxin-like chemicals in soil and sediment samples.

The Chemically Activated Luciferase Gene Expression (CALUX) by Xenobiotic Detection Systems (XDS) bioassay was evaluated for the determination of the presence of dioxin and dioxin-like compounds in soil and sediment in two studies conducted under the U.S. Environmental Protection Agency's Superfund Innovative Technology Evaluation Monitoring and Measurement Technologies Program. In the first study, the results were compared with those generated by established laboratory methods (EPA Method 1613B) using high-resolution mass spectrometry (HRMS). The study results demonstrated that the technology could be used to screen for dioxin concentrations above and below threshold values (e.g., less than or greater than 1 or 50 picograms of toxicity equivalents per gram [pg TEQ/g]); however, the results were not linearly correlated to the HRMS results. A second study was initiated to evaluate performance on a site-specific basis. During the second study, the data from the XDS technology were evaluated in four ways: (1) uncalibrated to HRMS, (2) calibrated using an overall statistical model, (3) calibrated using statistical models generated on a site-specific basis, and (4) calibrated using site-specific calibration factors. The results showed that TEQ data produced by the XDS technology were more precise than the data reported during the first study. The second study also demonstrated that site-specific statistical models were better tools for understanding the relationship between the XDS and HRMS data than a single overall model generated from data from multiple sites. Ultimately, site-specific calibration was shown to be the best approach because it was a simple and accurate way of correcting the XDS data and improving comparability with HRMS.