Spacecraft exploring Earth and other bodies in the solar system are producing increasingly large and complex volumes of scientific data. Recently, the field of science autonomy has begun to develop data analysis algorithms to select, prioritize, and characterize science instrument data directly onboard the spacecraft. These data-derived insights can then inform downstream autonomous decisions like data prioritization to mitigate data bandwidth constraints. However, science autonomy development has largely focused on bespoke instrument-specific algorithms; there are no generalized platforms for sharing reusable utilities or centralizing the operation of multiple science autonomy algorithms. To accelerate future development of flight-ready science autonomy, we developed a software package called Science Yield Improvement via Onboard Prioritization and Summary of Information System (SYNOPSIS). SYNOPSIS manages the execution of one or more science autonomy algorithms, stores their outputs in a centralized database, and prioritizes data even across multiple instruments. We simulated the use of SYNOPSIS with science autonomy algorithms to analyze and prioritize data collected by the Curiosity rover on Mars and integrated it into two popular flight software packages-core Flight System and F-Prime-to ease infusion of future science autonomy algorithms. As opportunities for onboard data analysis grow with the volume and complexity of data, we expect that platform technologies like SYNOPSIS will facilitate future exploration of our solar system.

Martian dust storms are valuable to study both for atmospheric science and for characterizing risks for future exploration missions. While dust plays an outsized role in the atmospheric of Mars, we lack a firm characterization of how dust storms initiate, grow, dissipate, and perturb the atmosphere over their lifecycles. Small (local) dust storms on Mars are particularly difficult to study; hundreds to thousands occur each Mars year, but their limited spatial extent and lifetime means we have relatively little to no atmospheric observations for each event. To investigate these local storms, we identified atmospheric measurements from the Mars Climate Sounder that coincide with recently mapped small storm boundaries from Mars Color Imager observations spanning Mars years 28-34 during the northern spring and summer. Using data science techniques to identify and aggregate these coincident observations, we extracted dust and temperature changes aligned with dust storm initiation across ~1350 events. Generally, we found that atmospheric perturbations are strong but short lived in the south, while northern local dust storms are weak. Our findings provide new insights into the atmospheric changes within small dust storms, which can inform future modeling studies and mission planning for exploring Mars.

A dominant driver of Martian atmospheric, landscape, and landform activity is the annual cycle of CO _2 and water, seasonally moving between atmospheric gases and surface frost. Orbital data set-based maps of the distribution of seasonal frost provide a broad look at how these volatiles move through time and space, as documented in past analyses. These past works typically take the form of a global study using a single, coarse-resolution, global data set or a spatially targeted site study combining data from multiple instruments. However, there does not yet exist a global map incorporating data sets from both the kilometer-resolution maps of the contiguous seasonal cap with landform-scale (about 10 m to hundreds of meters) patches of frost. Therefore, we produce a comprehensive, global map of the seasonal distribution of CO _2 frost through a single Mars year, integrating the information contained in visible (HiRISE/Context Camera), thermal (Mars Climate Sounder/THEMIS), and spectral (CRISM) data sets so as to generate 2.4 billion frost probability estimates. Combining global instrument data across diverse spectral, spatial, and temporal regimes required the rigorous application of multiple data science methods and a multidisciplinary collaboration between data scientists and planetary scientists. We produce two papers: this first focuses on the analysis of the individual data sets, highlighting key data science techniques and metrics used to capture the underpinning physical science intuition within robust and efficient autonomous “frost detection” methods. We generate, for each data set, 36 frost probability maps showing the distribution of surface CO _2 frost through a Mars year, with a temporal resolution of 10° L _s and a spatial resolution of 64 points-per-degree.

The quest to find extraterrestrial life is a critical scientific endeavor with civilization-level implications. Icy moons in our solar system are promising targets for exploration because their liquid oceans make them potential habitats for microscopic life. However, the lack of a precise definition of life poses a fundamental challenge to formulating detection strategies. To increase the chances of unambiguous detection, a suite of complementary instruments must sample multiple independent biosignatures (e.g., composition, motility/behavior, and visible structure). Such an instrument suite could generate 10,000x more raw data than is possible to transmit from distant ocean worlds like Enceladus or Europa. To address this bandwidth limitation, Onboard Science Instrument Autonomy (OSIA) is an emerging discipline of flight systems capable of evaluating, summarizing, and prioritizing observational instrument data to maximize science return. We describe two OSIA implementations developed as part of the Ocean Worlds Life Surveyor (OWLS) prototype instrument suite at the Jet Propulsion Laboratory. The first identifies life-like motion in digital holographic microscopy videos, and the second identifies cellular structure and composition via innate and dye-induced fluorescence. Flight-like requirements and computational constraints were used to lower barriers to infusion, similar to those available on the Mars helicopter, "Ingenuity." We evaluated the OSIA's performance using simulated and laboratory data and conducted a live field test at the hypersaline Mono Lake planetary analog site. Our study demonstrates the potential of OSIA for enabling biosignature detection and provides insights and lessons learned for future mission concepts aimed at exploring the outer solar system.

As tracers of the major volatile cycles of MarsCO2, H2O, and dustclouds are important for understanding the circulation of the martian atmosphere and hence martian climate. We present the spatial and seasonal distribution of laterally-confined clouds in the middle atmosphere of Mars during one Mars Year as identified in limb radiance measurements by the Mars Climate Sounder. Cloud identifications were made by citizen scientists through the “Cloudspotting on Mars” citizen science project, hosted on the citizen science platform Zooniverse. A method to aggregate the crowdsourced data using a novel clustering algorithm is developed. The derived cloud catalog is presented and the seasonal and spatial distribution of clouds is discussed in terms of key populations.

Seasonal frosting and defrosting on the surface of Mars is hypothesized to drive both climate processes and the formation and evolution of geomorphological features such as gullies. Past studies have focused on manually analyzing the behavior of the frost cycle in the northern mid-latitude region of Mars using high-resolution visible observations from orbit. Extending these studies globally requires automating the detection of frost using data science techniques such as convolutional neural networks. However, visible indications of frost presence can vary significantly depending on the geologic context on which the frost is superimposed. In this study, we (1) present a novel approach for spatially partitioning data to reduce biases in model performance estimation, (2) illustrate how geologic context affects automated frost detection, and (3) propose mitigations to observed biases in automated frost detection.

Ocean worlds such as Europa and Enceladus are high priority targets in the search for past or extant life beyond Earth. Evidence of life may be preserved in samples of surface ice by processes such as deposition from active plumes, hydrofracturing, or thermal convection. Terrestrial life produces unique distributions of organic molecules that translate into recognizable biosignatures. Identification and quantification of these organic compounds can be achieved by separation science such as capillary electrophoresis coupled to mass spectrometry (CE-MS). However, the data generated by such an instrument can be multiple orders of magnitude larger than what can be transmitted back to Earth during an ocean world's mission. This requires onboard science data analysis capabilities that summarize and prioritize CE-MS observations with limited computational resources. In response, the autonomous capillary electrophoresis mass-spectra examination (ACME) onboard science autonomy system was created for application to the ocean world's life surveyor (OWLS) instrument suite. ACME is able to compress raw mass spectra by two to three orders of magnitude while preserving most of its scientifically relevant information content. This summarization is achieved by the extraction of raw data surrounding autonomously identified ion peaks and the detection and parameterization of unique background regions. Prioritization of the summarized observations is then enabled by providing estimates of scientific utility, including presence of key target compounds, and the uniqueness of an observation relative to previous observations.

We describe a system for high-temperature investigations of bacterial motility using a digital holographic microscope completely submerged in heated water. Temperatures above 90°C could be achieved, with a constant 5°C offset between the sample temperature and the surrounding water bath. Using this system, we observed active motility in Bacillus subtilis up to 66°C. As temperatures rose, most cells became immobilized on the surface, but a fraction of cells remained highly motile at distances of >100 μm above the surface. Suspended non-motile cells showed Brownian motion that scaled consistently with temperature and viscosity. A novel open-source automated tracking package was used to obtain 2D tracks of motile cells and quantify motility parameters, showing that swimming speed increased with temperature until ∼40°C, then plateaued. These findings are consistent with the observed heterogeneity of B. subtilis populations, and represent the highest reported temperature for swimming in this species. This technique is a simple, low-cost method for quantifying motility at high temperatures and could be useful for investigation of many different cell types, including thermophilic archaea.

The Ocean Worlds Life Surveyor (OWLS) is a field prototype instrument suite designed to autonomously search for evidence of water-based life, developed in preparation for potential future missions to ocean worlds such as Enceladus and Europa. One instrument included in this suite is a Capillary Electrophoresis-Electrospray Ionization Mass Spectrometer (CE-ESI MS), which can detect the presence of organic molecules and other potential biosignature compounds. Due to the extreme energy costs involved in communication from these distant worlds, a mission’s downlink bandwidth is insufficient to return raw data from even a single recorded dataset. We developed two onboard capabilities to address this constraint: compression via knowledge summarization, and prioritization for the most scientifically useful observations. To summarize and prioritize the data generated by the CE-ESI MS, we developed the Autonomous CE-ESI Mass-Spectra Examination (ACME) system. ACME performs content summarization while ensuring that scientifically valuable signals are retained. First, ACME identifies and characterizes potential peaks in the mass spectra, each of which may indicate the presence of a specific compound. Then, ACME uses a decision tree model trained on expert-labeled data and peak properties such as width and signal-to-noise ratio to filter only for peaks of likely scientific interest. Finally, ACME produces a series of Autonomous Science Data Products (ASDPs): crops of small regions of the raw mass spectra data around each peak, a summary of the background noise to provide context and justification for its decisions, estimates of the scientific utility of the observation, and a brief description of its contents to enable downlink prioritization based on known science targets of interest as well as diversity sampling. Typical data sizes of the peak locations, crops, and background noise summary satisfy the mission downlink bandwidth constraints with an average compression ratio of 900:1. ACME was validated on lab- and field-collected data to confirm that scientists are able to successfully analyze and make valid scientific conclusions using only ACME’s ASDPs, compared to analyzing the raw data directly.

There are still many obstacles that must be solved before brain-computer interfaces (BCIs) can advance out of controlled research settings and into real-world scenarios. Some of these obstacles arise because of a growing separation between neuroscience and neuroengineering. Much of modern neuroscience research operates at the cortical level, so it is difficult to leverage this basic science research in BCIs that use noninvasive recordings (e.g., electroencephalography or magnetoencephalography) because those BCI data are recorded from outside the head. A neuroimaging technique called "source imaging" may offer a way to alleviate this disconnect as it allows the estimation cortical activity (on the surface of the brain) from noninvasive data (recorded on or above the surface of the scalp). We start by explaining the fundamentals of source imaging and then explore research using this technique to address two issues in BCI research. First, targeting brain activity from the most relevant cortical region can improve classification accuracy compared to the traditional approach. Second, source imaging was shown to improve transfer learning-an area of research aimed at reducing the 20- to 30-min calibration period required for most noninvasive BCIs. Overall, this chapter illustrates how tools and research findings from neuroscience can serve as a principled way to advance BCI methodology.

Introduction: Attention-mediated neural signals such as the P300 response allow some paralyzed patients to communicate via a speller device. New auditory speller devices that aid a listener's ability to selectively listen may increase the bit rate of communication. However, basic psychophysical studies probing stimulus design and ergonomic issues are lacking. Furthermore, there are untested usability concerns related to learning and memory load of new fixed-order and alphabetic-based auditory P300 BCI design, such as the new charStreamer paradigm [1]. Methods: Trials were divided into three conditions: alphabetic, fixed-order (non-alphabetic), and random (changing order). The alphabetic condition closely matches the charStreamer paradigm proposed in [1]: tokens (letters plus several additional commands) were parsed into three spatial locations in alphabetic order (left to right). In the fixed-order condition, tokens with similar pronunciations (like letters ‘b’, ‘c’, ‘e’) were separated. The random condition also began with this same separation of similar letters; however, the ordering was pseudo-randomly shuffled, such that subjects could not predict when the target token would occur. In order to investigate learning effects, data was analyzed at the initial (first 9 trials) and final (last 9 trials) stage of the experiment (27 trials in total). Behavioral: To test each subject’s ability to detect target tokens in each condition, subjects were asked whether the target occurred once or twice. Physiological: Pupillometry is a corollary of the attention-based effort and brain activation in a task [2]. Pupillometry was measured using EyeLink1000 eye tracker. Subjective: To assess the subject's experience of cognitive load, the NASA Task Load Index (TLX; [3]) survey was completed after the experiment. Results: Behavioral: There were no significant differences between the accuracies of any condition in the initial trials or the final trials (p>0.12, all) or between any one condition’s accuracy from initial to the final trials (p>0.3, all). Physiological: Mean pupil size in the fixed-order condition was significantly greater than alphabetic and random in early trials (p=0.04, p=0.02 respectively; uncorrected, Fig. 1a). Within the fixed-order trials, mean pupil size decreased from the initial trials to the final trials (p=0.03). Subjective: Subjects rated the random condition significantly harder than both the fixed-order and alphabetic condition (p=0.04, p=0.0005 respectively; uncorrected, Fig. 1b). The greater difficulty of fixed-order vs. alphabetic was not significant (p=0.07). Discussion: Behavioral: Accuracy in discriminating the target was the only measure used that relates each condition to a projected bit rate, thus these findings mainly provide an opportunity to compare usability issues. Physiological: The fixed-order condition elicits the highest relative pupil size (p=0.03), which is likely due to the relatively high cognitive load involved in memorizing the fixedordering—an activity that would not have been necessary in alphabetic-order or feasible in random-order trials. The significant decrease in pupil size for only the fixed-order condition from the initial to final trials (Fig. 1b) suggests that subjects learned and used the fixed-ordering over the course of the experiment. Subjective: Subjects rated the fixed-order condition as easier than the randomorder condition, which also suggests that they were able to learn the fixed ordering as an informative cue to reduce task difficulty. From these findings we conclude that, with exposure, a paradigm with an arbitrary, but fixed-order presentation, may approach the same usability as a known (i.e., alphabetic) order, whereas an unpredictable pattern may always impact usability. Therefore, alphabetic-ordering or alternative fixed-orderings may increase the usability of speller systems. Significance: These findings suggest that leveraging both subjective and objective measures of user effort can lead to further optimizations of BCI speller paradigms. References: [1] Höhne J and Tangermann M (2014). “Towards User-Friendly Spelling with an Auditory Brain-Computer Interface: The charStreamer Paradigm.” PLoS ONE, 9(7):e102630. [2] Zekveld AA et al., (2014). “The eye as a window to the listening brain: Neural correlates of pupil size as a measure of cognitive load.” Neuroimage, 101: 76-86. [3] Zickler, C, Halder, S, Kleih, SC, Herbert, C and Kübler, A (2014). “Brain Painting: usability testing according to the user-centered design in end users with severe motor paralysis.” Artificial Intelligence in medicine, 59(2):99-110.

Objective. Brain-computer interface (BCI) technology allows users to generate actions based solely on their brain signals. However, current non-invasive BCIs generally classify brain activity recorded from surface electroencephalography (EEG) electrodes, which can hinder the application of findings from modern neuroscience research. Approach. In this study, we use source imaging-a neuroimaging technique that projects EEG signals onto the surface of the brain-in a BCI classification framework. This allowed us to incorporate prior research from functional neuroimaging to target activity from a cortical region involved in auditory attention. Main results. Classifiers trained to detect attention switches performed better with source imaging projections than with EEG sensor signals. Within source imaging, including subject-specific anatomical MRI information (instead of using a generic head model) further improved classification performance. This source-based strategy also reduced accuracy variability across three dimensionality reduction techniques-a major design choice in most BCIs. Significance. Our work shows that source imaging provides clear quantitative and qualitative advantages to BCIs and highlights the value of incorporating modern neuroscience knowledge and methods into BCI systems.

Recent technological advances now allow for the collection of vast data sets detailing the intricate neural connectivity patterns of various organisms. Oh et al. (2014) recently published the most complete description of the mouse mesoscale connectome acquired to date. Here we give an in-depth characterization of this connectome and propose a generative network model which utilizes two elemental organizational principles: proximal attachment ‒ outgoing connections are more likely to attach to nearby nodes than to distant ones, and source growth ‒ nodes with many outgoing connections are likely to form new outgoing connections. We show that this model captures essential principles governing network organization at the mesoscale level in the mouse brain and is consistent with biologically plausible developmental processes.

Introduction: There exist a number of open challenges facing brain-computer interfaces (BCIs) for both clinical and commercial applications. In electroencephalography (EEG) based BCIs, several of these issues stem from analyzing recorded data in the sensor space (i.e., the surface of the scalp) where the electrodes are placed. Here, we focus on two of these obstacles: First, inclusion of priors from neuroscience (e.g., localization of brain regions associated with particular tasks) is difficult when analyzing data in the sensor space as the vast majority of neuroscience operates in the cortical domain. Second, the bulk of BCI research has focused on three canonical paradigms: P300, motor-imagery, and visually evoked potential BCIs all of which are well developed, but presumably, there remain many more brain regions useful for control that could advance the field as a whole [1]. Materials and Methods: Source imaging is a method for estimating cortical sources of activity from noninvasive recordings [2]. We modeled the entire cortex using ~7000 distributed current dipoles each of which represents a small patch of cortex. Using the MNE-Python library, we solved the inverse problem allowing estimation of cortical activity at each dipole from (non-invasive) EEG data. Source imaging also provides a principled path to include neuroscience priors from other research ‒ particularly, from neuroimaging. To this point, the right temporoparietal junction (RTPJ) was recently discovered to be significantly more active when switching auditory attention compared to maintaining it to a single sound source [3]. We hypothesized that a source-based approach incorporating this knowledge (by targeting activity from only this region) would provide significantly better single-trial classification accuracy compared to a naive sensor space approach. We tested our hypothesis by comparing a sensor-based and a source-based BCI approach both attempting to classify (offline) if a subject switched or maintained attention in an auditory task (previously conducted in the laboratory) [3]. Briefly, subjects listened to one of two talkers and were instructed to either maintain attention to one throughout a trial or switch halfway through. For both approaches, we employed different dimensionality reduction techniques: principal component analysis (PCA), independent component analysis (ICA), and common spatial patterns (CSP) using an identical range of parameters for each. Support vector machines were used to classify the resulting signal and we employed 10-fold cross validation to obtain a stable accuracy estimate. Results: We found that the source-based approach significantly outperformed its sensor-based counterpart (p=0.003; corrected 2-way repeated-measures ANOVA) conferring a 5.2% absolute accuracy increase between the best sensorand source-based strategies. Interestingly, the relative difference between the dimensionality reduction techniques appeared to diminish once in the source space. Discussion: These results suggest that the source space provides an avenue to target more informative signals for classification by incorporating neuroscience priors. The absolute accuracy attained here may not be robust enough to be useful for immediate use; however, the significant gains relative to a sensor approach gives credence to the notion that neuroscience priors can provide a significant performance gain across multiple signal processing strategies. Importantly, this work also demonstrates that activity from a region not associated with the canonical BCI paradigms can be objectively targeted to provide a useful control signal. Note that source imaging is not amenable to all BCI studies; the additional time and cost required to obtain a structural MRI scan will be prohibitive in some cases. That said, there are simplified source imaging techniques that require only 3D localization of electrodes and a generic head model. Future research will evaluate the necessity of MRI information by testing these simplified techniques. Significance: In this study, we found that leveraging neuroscience priors via the source space provides a significant increase in classification accuracy and seems to reduce the dependence of this accuracy on the dimensionality technique chosen. Acknowledgements: Funded by NSF GRFP to M.W. and AFOSR to A.L. References: [1] Vansteensel MJ, Hermes D, Aarnoutse EJ, Bleichner MG, Schalk G, van Rijen PC, Leijten FSS, Ramsey NF. Brain-Computer Interfacing Based on Cognitive Control. Annals of Neurology, 67(6): 809-16, 2010. [2] Baillet S, Mosher JC, Leahy RM. Electromagnetic Brain Mapping. IEEE Signal Processing Magazine, 18(6): 14-30, 2001. [3] Larson E, Lee AKC. Switching auditory attention using spatial and non-spatial features recruits difference cortical networks. NeuroImage, 84(2014): 681-687, 2014. Figure 1. Accuracy of sensorand source-space approaches (using 3 dimensionality reduction techniques) when predicting a switch in attention. Source space yields significant improvement. Bars indicate mean (± SEM) and the red line represents chance. DOI: 10.3217/978-3-85125-467-9-30 Proceedings of the 6th International Brain-Computer Interface Meeting, organized by the BCI Society

Objective. Brain-computer interfaces (BCIs) represent a technology with the potential to rehabilitate a range of traumatic and degenerative nervous system conditions but require a time-consuming training process to calibrate. An area of BCI research known as transfer learning is aimed at accelerating training by recycling previously recorded training data across sessions or subjects. Training data, however, is typically transferred from one electrode configuration to another without taking individual head anatomy or electrode positioning into account, which may underutilize the recycled data. Approach. We explore transfer learning with the use of source imaging, which estimates neural activity in the cortex. Transferring estimates of cortical activity, in contrast to scalp recordings, provides a way to compensate for variability in electrode positioning and head morphologies across subjects and sessions. Main results. Based on simulated and measured electroencephalography activity, we trained a classifier using data transferred exclusively from other subjects and achieved accuracies that were comparable to or surpassed a benchmark classifier (representative of a real-world BCI). Our results indicate that classification improvements depend on the number of trials transferred and the cortical region of interest. Significance. These findings suggest that cortical source-based transfer learning is a principled method to transfer data that improves BCI classification performance and provides a path to reduce BCI calibration time.

: Harnessing the capability to read and classify brainwaves into the myriad of possible human cognitive states (referred to as brain-states) has been a long-standing engineering challenge. Brain signals are generally captured non-invasively by electroencephalography (EEG) a cheapand portable brain imaging tool with time resolution fine enough to track the dynamic changes of different brain-states. While the scientific quest to map human brain function has exploded in the last two decades, the ability to link patterns in EEG signals to specific cognitive states remains elusive, owing perhaps to limited crosstalk between the fields of neuroscience and engineering. Here, we report a framework we developed that leverages the latest neuroscience knowledge to transform the current engineering approach to brain-state classification. We used inverse imaging techniques and surface-based spatial normalization algorithms to interpret brain signals across a large pool of subjects and cross-validated our findings with simulated and actual brain data. We concluded that decoding brain signals in the brain (a.k.a. source-space approach) confers two major benefits compared to classifying brain signals directly on the EEG sensor readings (a.k.a. sensor-space approach): i) it provides a principled method to transfer data from one subject to another, thereby reducing BCI calibration time; and ii) it increases classification accuracy regardless of which dimensionality-reduction techniques were used to preprocess the data. Overall, this innovative approach establishes a formal integrated neuroengineering framework that allows us to capitalize on the similarity in brain function across subjects (a traditional neuroscience approach) and optimally incorporate a priori information to maximize classification algorithm performance at an individual level (a traditional engineering goal) that ultimately improves our ability to classify human brain-states.

Traditional brain-state classifications are primarily based on two well-known neural biomarkers: P300 and motor imagery / event-related frequency modulation. Currently, many brain-computer interface (BCI) systems have successfully helped patients with severe neuromuscular disabilities to regain independence. In order to translate this neural engineering success to hearing aid applications, we must be able to capture brain waves across the population reliably in cortical regions that have not previously been incorporated in these systems before, for example, dorsolateral prefrontal cortex (DLPFC) and right temporoparietal junction. Here, we present a brain-state classification framework that incorporates individual anatomical information and accounts for potential anatomical and functional differences across subjects by applying appropriate cortical weighting functions prior to the classification stage. Using an inverse imaging approach, use simulated EEG data to show that our method can outperform the traditional brain-state classification approach that trains only on individual subject's data without considering data available at a population level.

Brain–computer interface (BCI) systems have emerged as a method to restore function and enhance communication in motor impaired patients. To date, this has been applied primarily to patients who have a compromised motor outflow due to spinal cord dysfunction, but an intact and functioning cerebral cortex. The cortical physiology associated with movement of the contralateral limb has typically been the signal substrate that has been used as a control signal. While this is an ideal control platform in patients with an intact motor cortex, these signals are lost after a hemispheric stroke. Thus, a different control signal is needed that could provide control capability for a patient with a hemiparetic limb. Previous studies have shown that there is a distinct cortical physiology associated with ipsilateral, or same-sided, limb movements. Thus far, it was unknown whether stroke survivors could intentionally and effectively modulate this ipsilateral motor activity from their unaffected hemisphere. Therefore, this study seeks to evaluate whether stroke survivors could effectively utilize ipsilateral motor activity from their unaffected hemisphere to achieve this BCI control. To investigate this possibility, electroencephalographic (EEG) signals were recorded from four chronic hemispheric stroke patients as they performed (or attempted to perform) real and imagined hand tasks using either their affected or unaffected hand. Following performance of the screening task, the ability of patients to utilize a BCI system was investigated during on-line control of a one-dimensional control task. Significant ipsilateral motor signals (associated with movement intentions of the affected hand) in the unaffected hemisphere, which were found to be distinct from rest and contralateral signals, were identified and subsequently used for a simple online BCI control task. We demonstrate here for the first time that EEG signals from the unaffected hemisphere, associated with overt and imagined movements of the affected hand, can enable stroke survivors to control a one-dimensional computer cursor rapidly and accurately. This ipsilateral motor activity enabled users to achieve final target accuracies between 68% and 91% within 15 min. These findings suggest that ipsilateral motor activity from the unaffected hemisphere in stroke survivors could provide a physiological substrate for BCI operation that can be further developed as a long-term assistive device or potentially provide a novel tool for rehabilitation.