Pennsylvania Commonwealth Universal Research Enhancement Program (PA CURE):
Enabling the Application of Deep Learning to Automated Seizure Detection

Database Evolution & Ideology

The Temple University Seizure Corpus was started with the development of an evaluations set, which was created from the EEG records of 50 TUH_EEG patients. More information regarding the TUH EEG can be found here.

This set was created with the goal of maintaining as much diversity as possible in order to show resemblance to real-world EEG morphologies. Variety is preserved in terms of seizure types and clinical conditions. This set includes both convulsive seizures, such as tonic-clonic and myoclonic, as well as non-convulsive seizures, such as absence and complex-partial. Patients with conditions that cause extremely complicated electrographic morphologies, such as Lennox-Gestaut syndrome, have been included as well. Diversity in both age and gender have also been preserved. This aims to reflect the overall distribution of TUH_EEG patients.

Manual annotation is done through a closed loop process where at least three annotators are required to look at a particular EEG before release. Channel specific annotations are made with clear electrographic spread through the hemisphere. All non-conclusive/gray-zone EEGs are put on Q&A queries for neurologists' feedback and are discussed amongst annotators in weekly group meetings.

Database Progression

Release Version Description
v1.5.2 This version include new annotations for the entire training database.
v1.5.1 The annotations for the dev and eval sets have been manually reviewed in preparation for the Neureka™ 2020 Epilepsy Challenge.
v1.5.0 This release includes the expansion of the training dataset from 1,984 files to 4,597. Calibration sequences of the new data have been manually annotated and added to the seizure spreadsheet. Annotation corrections were made to the files already existing in the training set.
v1.4.0 This release includes improvements to the quality of annotations. Annotation corrections were made in the development test and training sets.
v1.3.0 This release contains quality improvements of the annotations, as manually labeled calibration sequences. The main reason for this release is that we have created a blind evaluation set, often referred to as a held-out set.
v1.2.1 This release contains enhanced documentation and corrected data structuring.
v1.2.0 This version directly uses the official patient numbers used in v1.1.0 of the TUH_EEG database.
v1.1.0 This version includes an expanded training set. Seizure time marks were also adjusted to a finer resolution.
v1.0.4 This release contains bi-class annotations files (seizure/no-seizure). A detailed spreadsheet, which classifies each session according to normal/abnormal EEG types and subtypes, is also included. Annotation endpoints were quantized to a 1 second resolution.
v1.0.3 This release contains the same data as v1.0.0 but includes more detailed documentation and adjusted naming conventions.
v1.0.0 This is the first official release of an annotated data set.
v0.6.0 This version's data set was used for a large portion of the initial image classification system's experiments.
v0.0.0 This is the first release for TUH_EEG Seizure. Seizure annotations were created using Encevis and Persyst. Annotations follow the formatting guidelines of the Auto_EEG Demo 0.2.

Data Set Statistics (V1.5.2)

Development Test Set

Type Total
Files and Sessions
Files 1013
Files Containing Seizures 280
Sessions 238
Sessions Containing Seizures 104
Patients 50
Patients Containing Seizures 40
Signal Data
Seizures 58,445 secs (9.53%)
Background 554,787 secs    (90.47%)
Total 613,232 secs
Files Containing Seizures 230,031 secs (37.51% of the total data)

Training Set

Type Total
Files and Sessions
Files 4599
Files Containing Seizures 869
Sessions 1185
Sessions Containing Seizures 343
Patients 592
Patients Containing Seizures 202
Signal Data
Seizures 169,794 secs (6.26%)
Background 2,540,689 secs (93.74%)
Total 2,710,483 secs
Files Containing Seizures 637,689 secs (23.52% of the total data)


Publications

  • Shah, V., Golmohammadi, M., Obeid, I., & Picone, J. (2021). Objective Evaluation Metrics for Automatic Classification of EEG Events. In I. Obeid, I. Selesnick, & J. Picone (Eds.), Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications (1st ed., pp. 1–26). (Download).

  • Ferrell, S., Mathew, V., Ahsan, T., & Picone, J. (2020). The Temple University Hospital EEG Corpus: Electrode Location and Channel Labels. Philadelphia, Pennsylvania, USA. (Download).

  • Golmohammadi, M., Shah, V., Obeid, I., & Picone, J. (2020). Deep Learning Approaches for Automatic Seizure Detection from Scalp Electroencephalograms. In I. Obeid, I. Selesnick, and J. Picone (Eds.), Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications (1st ed., pp. 233–274). (Download).

  • Lin, R., Marquez, D., Jacobson, M., Castaldi, H., Buckman, S., Shah, V., & Picone, J. (2020). Accuracy of Automated Machine Learning Software in Identifying EEGs with Prolonged Seizures. Annual Meeting of the American Academy of Neurology (AAN), (p. P6.002). Philadelphia, Pennsylvania, USA. (Download).

  • Ochal, D., Rahman, S., Ferrell, S., Elseify, T., Obeid, I., & Picone, J. (2020). The Temple University Hospital EEG Corpus: Annotation Guidelines. Philadelphia, Pennsylvania, USA. (Download).

  • Golmohammadi, M., Harati, A., Lopez, S., Obeid, I., & Picone, J. (2019). Automatic Analysis of EEGs Using Big Data and Hybrid Deep Learning Architectures. Frontiers in Human Neuroscience, 13, 76. (Download).

  • Jean-Paul, S., Elseify, T., Obeid, I., & Picone, J. (2019). Issues in the Reproducibility of Deep Learning Results. I. Obeid & J. Picone (Eds.), Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB) (pp. 1-4). (Download).

  • Kiral, I., Roy, S., Mummert, T., Braz, A., Tsay, J., Tang, J., Asif, U., Schaffter, T., Mehmet, E., Picone, J., Obeid, I., Marques, B., Maetschke, S., Khalaf, R., Rosen-Zvi, M., Stolovitzky, G., Mirmomeni, M., Harrer, S., Yanagisawa, H., Iwamori, T., Madan, P., Qin, Y., Ma, L., Ti, W., Liu, W., Mei, J., Hensley, S., Chandra, R., Hake, P., Henessy, R., Babaali, P., Yuenreed, G., Kather, R., Arcos-Diaz, D., Cherner, M. (2019). The Deep Learning Epilepsy Detection Challenge: Design, Implementation, and Test of a New Crowd-Sourced AI Challenge Ecosystem, in Challenges in Machine Learning Competitions for All (CiML) (pp. 1-3). Vancouver, Canada (Download).

  • Obeid, I., & Picone, J. (2019). Applying Speech Processing Approaches to EEG. EEG: Analytical Approaches and Applications Virtual Symposium. Philadelphia, Pennsylvania, USA. (Download).

  • Obeid, I., & Picone, J. (2019). Applying Speech Processing Approaches to EEG. EEG: Analytical Approaches and Applications Virtual Symposium. Philadelphia, Pennsylvania, USA. (Download).

  • Picone, J., & Obeid, I. (2019). The Temple University Hospital (TUH) Electroencephalogram (EEG) Corpus. EEG: Analytical Approaches and Applications Virtual Symposium. Philadelphia, Pennsylvania, USA. (Download).

  • Picone, J., & Obeid, I. (2019). The Temple University Hospital (TUH) Electroencephalogram (EEG) Corpus. EEG: Analytical Approaches and Applications Virtual Symposium. Philadelphia, Pennsylvania, USA. (Download).

  • Rahman, S., Miranda, M., Obeid, I., & Picone, J. (2019). Software and Data Resources to Advance Machine Learning Research in Electroencephalography. I. Obeid & J. Picone (Eds.), Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB) (pp. 1-4). (Download).

  • Shah, V., von Weltin, E., Ahsan, T., Ziyabari, S., Golmohammadi, M., Obeid, I. and Picone, J. (2019). On the Use of Non-Experts for Generation of High-Quality Annotations of Seizure Events. Journal of Clinical Neurophysiology. (Download).

  • Capp, N., Campbell, C., Elseify, T., Obeid, I., & Picone, J. (2018). Optimizing EEG Visualization Through Remote Data Retrieval and Asynchronous Processing. Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB) (pp. 1–2). Philadelphia, Pennsylvania, USA. (Download).

  • Ferrell, S., von Weltin, E., Obeid, I., & Picone, J. (2018). Open Source Resources to Advance EEG Research. Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB) (pp. 1–3). Philadelphia, Pennsylvania, USA. (Download).

  • Golmohammadi, M., Ziyabari, S., Shah, V., Obeid, I., & Picone, J. (2018). Deep Architectures for Spatio-Temporal Modeling: Automated Seizure Detection in Scalp EEGs. Proceedings of the International Conference on Machine Learning and Applications (ICMLA) (pp. 1–6). Orlando, Florida, USA. (Download).

  • Golmohammadi, M., Obeid, I. and Picone, J. (2018). Deep Residual Learning for Automatic Seizure Detection. 26th Conference on Intelligent Systems for Molecular Biology (p. 1). Chicago, Illinois, USA. (Download).

  • Lopez, S., Obeid, I. and Picone, J. (2018). Automated Interpretation of Abnormal Adult Electroencephalograms. 26th Conference on Intelligent Systems for Molecular Biology (p. 1). Chicago, Illinois, USA. (Download).

  • Obeid, I., & Picone, J. (2018). The Temple University Hospital EEG Data Corpus. In Augmentation of Brain Function: Facts, Fiction and Controversy. Volume I: Brain-Machine Interfaces (1st ed., pp. 394–398). Lausanne, Switzerland: Frontiers Media S.A. (Download).

  • Obeid, I., & Picone, J. (2018). Machine Learning Approaches to Automatic Interpretation of EEGs. In E. Sejdik & T. Falk (Eds.), Signal Processing and Machine Learning for Biomedical Big Data (1st ed., p. 70). Boca Raton, Florida, USA: CRC Press. (Download).

  • Picone, J. and Obeid, I. (2018). Enabling Deep Learning Approaches for Automatic Interpretation of EEGs. Neural Interfaces Conference (p. 23). Minneapolis, Minnesota, USA. (Download).

  • Shah, V., von Weltin, E., Lopez, S., McHugh, J. R., Veloso, L., Golmohammadi, M., … Picone, J. (2018). The Temple University Hospital Seizure Detection Corpus. Frontiers in Neuroinformatics, 12, 1-6. (Download).

  • Shah, V., Anstotz, R., Obeid, I., & Picone, J. (2018). Adapting an Automatic Speech Recognition System to Event Classification of Electroencephalograms. Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB) (pp. 1–4). Philadelphia, Pennsylvania, USA. (Download).

  • Veloso, L., McHugh, J. R., von Weltin, E., Obeid, I. and Picone, J. (2017). Big Data Resources for EEGs: Enabling Deep Learning Research. Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB) (p. 1). Philadelphia, Pennsylvania, USA: IEEE. (Download).

  • von Weltin, E., Ahsan, T., Shah, V., Jamshed, D., Golmohammadi, M., Obeid, I. and Picone, J. (2017). Electroencephalographic Slowing: A Primary Source of Error in Automatic Seizure Detection. Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (pp. 1-5). Philadelphia, Pennsylvania, USA: IEEE. (Download).

If you have any additional comments or questions regarding the listed data, please direct them to help.nedcdata.org