More Like this

1. Investigating the effect of indoor thermal environment on occupants’ mental workload and task performance using electroencephalogram

   Wang, Xi ; Li, Da ; Menassa, Carol. C ; and Kamat, Vineet R. ( January 2019 , Building and environment)

   Workers' performance in indoor offices can be greatly affected by the thermal condition of the environment. However, this effect can be difficult to quantify, especially when the thermal stress is a moderate increase or decrease in temperature and the work productivity cannot be directly measured. Subjects' high motivation to perform well under experimental conditions also causes difficulties in comparing their performance in different thermal environments. In order to overcome these limitations, this paper proposes a method to investigate the effect of the indoor thermal conditions on occupants' performance by studying occupants' mental workload measured by the electroencephalography (EEG) when they perform standardized cognitive tasks. An experiment integrating EEG mental workload measurement and cognitive tasks was implemented on 15 subjects. EEG data were collected while subjects were performing four cognitive tasks on computers. Based on previous studies, we propose a mental workload index calculated from the frontal theta and parietal alpha frequency band power. Within-subject comparisons were performed to investigate whether subjects' mental workload is statistically different under three different thermal environments, representing thermal sensations of slightly cool, neutral, and slightly warm. The results show that the effect of thermal environment varies across different individuals. By comparing the mental workload index among different thermal environments, we found that the slightly warm environment resulted in a relatively higher mental workload than the other two environments to achieve the same performance. The study provides promising insights into how the thermal environment influences occupants’ performance by affecting their mental workload from the neurophysiological perspective. 

   more » « less
2. Can Infrared Facial Thermography Disclose Mental Workload in Indoor Thermal Environments?

   https://doi.org/10.1145/3363459.3363528  

   Wang, Xi ; Li, Da ; Menassa, Carol ; Kamat, Vineet. ( November 2019 , UrbSys'19: Proceedings of the 1st ACM International Workshop on Urban Building Energy Sensing, Controls, Big Data Analysis, and Visualization)

   null (Ed.)

   Mental workload represents the mental resources an individual devotes to a task. In a building environment, understanding how ambient thermal conditions affect occupants' mental workload offers an opportunity to achieve optimal thermal settings for the heating, ventilation, and air conditioning (HVAC) systems. However, directly measuring mental workload on a large and continuous scale requires occupants to perform subjective tests or wear electroencephalogram (EEG) or similar devices, which is impractical. This paper assesses the feasibility of using infrared facial thermography captured by a low-cost thermal camera to disclose mental workload. An experiment was conducted to measure the facial skin temperature while subjects performed cognitive tasks in three different thermal environments, representing occupants' thermal sensation of slightly cool, neutral, and slightly warm. Mental workload was measured using an EEG headset to eliminate subjective bias. The correlations between facial temperature and mental workload vary with different individuals and thermal conditions. Relatively strong correlations are found in the neutral environment and in the regions of ears, mouth, and neck. The results also suggest that future work should collect data under extended experiment duration. This is because it was observed that the response of facial skin temperature to mental workload varies with task type; thus, increasing the repetitiveness for each type of task or using more challenging tasks in the experiment could potentially lead to more insights on this relationship. 

   more » « less
3. Identifying Patterns for Neurological Disabilities by Integrating Discrete Wavelet Transform and Visualization

   https://doi.org/10.3390/app14010273  

   Ji, Soo Yeon ; Jayarathna, Sampath ; Perrotti, Anne M. ; Kardiasmenos, Katrina ; Jeong, Dong Hyun ( January 2024 , Applied Sciences)

   Neurological disabilities cause diverse health and mental challenges, impacting quality of life and imposing financial burdens on both the individuals diagnosed with these conditions and their caregivers. Abnormal brain activity, stemming from malfunctions in the human nervous system, characterizes neurological disorders. Therefore, the early identification of these abnormalities is crucial for devising suitable treatments and interventions aimed at promoting and sustaining quality of life. Electroencephalogram (EEG), a non-invasive method for monitoring brain activity, is frequently employed to detect abnormal brain activity in neurological and mental disorders. This study introduces an approach that extends the understanding and identification of neurological disabilities by integrating feature extraction, machine learning, and visual analysis based on EEG signals collected from individuals with neurological and mental disorders. The classification performance of four feature approaches—EEG frequency band, raw data, power spectral density, and wavelet transform—is assessed using machine learning techniques to evaluate their capability to differentiate neurological disabilities in short EEG segmentations (one second and two seconds). In detail, the classification analysis is conducted under two conditions: single-channel-based classification and region-based classification. While a clear demarcation between normal (healthy) and abnormal (neurological disabilities) EEG metrics may not be evident, their similarities and distinctions are observed through visualization, employing wavelet features. Notably, the frontal brain region (frontal lobe) emerges as a crucial area for distinguishing abnormalities among different brain regions. Also, the integration of wavelet features and visual analysis proves effective in identifying and understanding neurological disabilities. 

   more » « less
4. Thermal and Exergy Analysis in UPS and Battery Rooms by Numerical Simulations

   Caceres, C ; Ortega, A ; Silva-Llanca, L ; Jones, G ; Sapia, N ( May 2018 , ITHERM)

   UPS (Uninterruptible Power Supply) units and batteries are essential subsystems in data centers or telecom industries to protect equipment from electrical power spikes, surges and power outages. UPS units handle electrical power and dissipate a large amount of heat, and possess a high efficiency. Therefore, cooling units (e.g., CRACs) are needed to manage the thermal reliability of this equipment. On the other hand, battery operating conditions and reliability are closely related to the ambient temperature according to battery manufacturers; reliability increases when the ambient room temperature is around 25ºC. This study analyzed different room configurations and scenarios using the commercial CFD software 6Sigma Room DCXTM. As a first approach, we evaluated the thermal behavior and cooling degradation using standard thermal performance metrics SHI (Supply Heat Index) and RHI (Return Heat Index). These are frequently implemented in data centers to measure the level of mixing between cold and hot air streams. The results from this evaluation showed that standard cooling practices are inefficient, as values for the two metrics differed considerably from industry recommendations. We also considered a metric from the second law of thermodynamics using exergy destruction. This technique allowed us to find the mechanisms that increase entropy generation the most, including viscous shear and air stream mixing. Reducing exergy destruction will result in lessening lost thermodynamic work and thus reduce energy required for cooling. Typically, UPS and batteries are located in different rooms due to the hydrogen generation by the batteries. The integration of both equipment in the same room is a new concept, and this study aims to analyze the thermal performance of the room. Adding controllability showed improvements by reducing the exergy destruction due to viscous dissipation while slightly increasing thermal mixing in the rooms. Ducting the return flows to avoid flow mixing increased pressure drop, but reduced heat transfer between the hot and cold air streams, which in turn, improved the thermal performance. In the study, we determined the optimal configuration and possible strategies to improve cooling while maintaining desirable battery temperatures. 

   more » « less
5. Recent Advances in the TUH EEG Corpus: Improving the Interrater Agreement for Artifacts and Epileptiform Events

   Buckwalter, Grace Chhin ( December 2021 , Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, Iyad Selesnick (Ed.)

   The Temple University Hospital EEG Corpus (TUEG) [1] is the largest publicly available EEG corpus of its type and currently has over 5,000 subscribers (we currently average 35 new subscribers a week). Several valuable subsets of this corpus have been developed including the Temple University Hospital EEG Seizure Corpus (TUSZ) [2] and the Temple University Hospital EEG Artifact Corpus (TUAR) [3]. TUSZ contains manually annotated seizure events and has been widely used to develop seizure detection and prediction technology [4]. TUAR contains manually annotated artifacts and has been used to improve machine learning performance on seizure detection tasks [5]. In this poster, we will discuss recent improvements made to both corpora that are creating opportunities to improve machine learning performance. Two major concerns that were raised when v1.5.2 of TUSZ was released for the Neureka 2020 Epilepsy Challenge were: (1) the subjects contained in the training, development (validation) and blind evaluation sets were not mutually exclusive, and (2) high frequency seizures were not accurately annotated in all files. Regarding (1), there were 50 subjects in dev, 50 subjects in eval, and 592 subjects in train. There was one subject common to dev and eval, five subjects common to dev and train, and 13 subjects common between eval and train. Though this does not substantially influence performance for the current generation of technology, it could be a problem down the line as technology improves. Therefore, we have rebuilt the partitions of the data so that this overlap was removed. This required augmenting the evaluation and development data sets with new subjects that had not been previously annotated so that the size of these subsets remained approximately the same. Since these annotations were done by a new group of annotators, special care was taken to make sure the new annotators followed the same practices as the previous generations of annotators. Part of our quality control process was to have the new annotators review all previous annotations. This rigorous training coupled with a strict quality control process where annotators review a significant amount of each other’s work ensured that there is high interrater agreement between the two groups (kappa statistic greater than 0.8) [6]. In the process of reviewing this data, we also decided to split long files into a series of smaller segments to facilitate processing of the data. Some subscribers found it difficult to process long files using Python code, which tends to be very memory intensive. We also found it inefficient to manipulate these long files in our annotation tool. In this release, the maximum duration of any single file is limited to 60 mins. This increased the number of edf files in the dev set from 1012 to 1832. Regarding (2), as part of discussions of several issues raised by a few subscribers, we discovered some files only had low frequency epileptiform events annotated (defined as events that ranged in frequency from 2.5 Hz to 3 Hz), while others had events annotated that contained significant frequency content above 3 Hz. Though there were not many files that had this type of activity, it was enough of a concern to necessitate reviewing the entire corpus. An example of an epileptiform seizure event with frequency content higher than 3 Hz is shown in Figure 1. Annotating these additional events slightly increased the number of seizure events. In v1.5.2, there were 673 seizures, while in v1.5.3 there are 1239 events. One of the fertile areas for technology improvements is artifact reduction. Artifacts and slowing constitute the two major error modalities in seizure detection [3]. This was a major reason we developed TUAR. It can be used to evaluate artifact detection and suppression technology as well as multimodal background models that explicitly model artifacts. An issue with TUAR was the practicality of the annotation tags used when there are multiple simultaneous events. An example of such an event is shown in Figure 2. In this section of the file, there is an overlap of eye movement, electrode artifact, and muscle artifact events. We previously annotated such events using a convention that included annotating background along with any artifact that is present. The artifacts present would either be annotated with a single tag (e.g., MUSC) or a coupled artifact tag (e.g., MUSC+ELEC). When multiple channels have background, the tags become crowded and difficult to identify. This is one reason we now support a hierarchical annotation format using XML – annotations can be arbitrarily complex and support overlaps in time. Our annotators also reviewed specific eye movement artifacts (e.g., eye flutter, eyeblinks). Eye movements are often mistaken as seizures due to their similar morphology [7][8]. We have improved our understanding of ocular events and it has allowed us to annotate artifacts in the corpus more carefully. In this poster, we will present statistics on the newest releases of these corpora and discuss the impact these improvements have had on machine learning research. We will compare TUSZ v1.5.3 and TUAR v2.0.0 with previous versions of these corpora. We will release v1.5.3 of TUSZ and v2.0.0 of TUAR in Fall 2021 prior to the symposium. ACKNOWLEDGMENTS Research reported in this publication was most recently supported by the National Science Foundation’s Industrial Innovation and Partnerships (IIP) Research Experience for Undergraduates award number 1827565. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the official views of any of these organizations. REFERENCES [1] I. Obeid and J. Picone, “The Temple University Hospital EEG Data Corpus,” in Augmentation of Brain Function: Facts, Fiction and Controversy. Volume I: Brain-Machine Interfaces, 1st ed., vol. 10, M. A. Lebedev, Ed. Lausanne, Switzerland: Frontiers Media S.A., 2016, pp. 394 398. https://doi.org/10.3389/fnins.2016.00196. [2] V. Shah et al., “The Temple University Hospital Seizure Detection Corpus,” Frontiers in Neuroinformatics, vol. 12, pp. 1–6, 2018. https://doi.org/10.3389/fninf.2018.00083. [3] A. Hamid et, al., “The Temple University Artifact Corpus: An Annotated Corpus of EEG Artifacts.” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2020, pp. 1-3. https://ieeexplore.ieee.org/document/9353647. [4] Y. Roy, R. Iskander, and J. Picone, “The NeurekaTM 2020 Epilepsy Challenge,” NeuroTechX, 2020. [Online]. Available: https://neureka-challenge.com/. [Accessed: 01-Dec-2021]. [5] S. Rahman, A. Hamid, D. Ochal, I. Obeid, and J. Picone, “Improving the Quality of the TUSZ Corpus,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2020, pp. 1–5. https://ieeexplore.ieee.org/document/9353635. [6] V. Shah, E. von Weltin, T. Ahsan, I. Obeid, and J. Picone, “On the Use of Non-Experts for Generation of High-Quality Annotations of Seizure Events,” Available: https://www.isip.picone press.com/publications/unpublished/journals/2019/elsevier_cn/ira. [Accessed: 01-Dec-2021]. [7] D. Ochal, S. Rahman, S. Ferrell, T. Elseify, I. Obeid, and J. Picone, “The Temple University Hospital EEG Corpus: Annotation Guidelines,” Philadelphia, Pennsylvania, USA, 2020. https://www.isip.piconepress.com/publications/reports/2020/tuh_eeg/annotations/. [8] D. Strayhorn, “The Atlas of Adult Electroencephalography,” EEG Atlas Online, 2014. [Online]. Availabl 

   more » « less