2020
Muhammad Saif-ur-Rehman; Omair Ali; Susanne Dyck; Robin Lienkämper; Marita Metzler; Yaroslav Parpaley; Jörg Wellmer; Charles Liu; Brian Lee; Spencer Kellis; Richard A Andersen; Ioannis Iossifidis; Tobias Glasmachers; Christian Klaes
SpikeDeep-Classifier: A deep-learning based fully automatic offline spike sorting algorithm Journal Article
In: Journal of Neural Engineering, 2020.
Abstract | Links | BibTeX | Tags: BCI, CNN, Machine Learning, Spike Sorting
@article{10.1088/1741-2552/abc8d4,
title = {SpikeDeep-Classifier: A deep-learning based fully automatic offline spike sorting algorithm},
author = {Muhammad Saif-ur-Rehman and Omair Ali and Susanne Dyck and Robin Lienkämper and Marita Metzler and Yaroslav Parpaley and Jörg Wellmer and Charles Liu and Brian Lee and Spencer Kellis and Richard A Andersen and Ioannis Iossifidis and Tobias Glasmachers and Christian Klaes},
url = {http://iopscience.iop.org/article/10.1088/1741-2552/abc8d4},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
journal = {Journal of Neural Engineering},
abstract = {Objective. Advancements in electrode design have resulted in micro-electrode arrays with hundreds of channels for single cell recordings. In the resulting electrophysiological recordings, each implanted electrode can record spike activity (SA) of one or more neurons along with background activity (BA). The aim of this study is to isolate SA of each neural source. This process is called spike sorting or spike classification. Advanced spike sorting algorithms are time consuming because of the human intervention at various stages of the pipeline. Current approaches lack generalization because the values of hyperparameters are not fixed, even for multiple recording sessions of the same subject. In this study, a fully automatic spike sorting algorithm called “SpikeDeep-Classifier” is proposed. The values of hyperparameters remain fixed for all the evaluation data. Approach. The proposed approach is based on our previous study (SpikeDeeptector) and a novel background activity rejector (BAR), which are both supervised learning algorithms and an unsupervised learning algorithm (K-means). SpikeDeeptector and BAR are used to extract meaningful channels and remove BA from the extracted meaningful channels, respectively. The process of clustering becomes straight-forward once the BA is completely removed from the data. Then, K-means with a predefined maximum number of clusters is applied on the remaining data originating from neural sources only. Lastly, a similarity-based criterion and a threshold are used to keep distinct clusters and merge similar looking clusters. The proposed approach is called cluster accept or merge (CAOM) and it has only two hyperparameters (maximum number of clusters and similarity threshold) which are kept fixed for all the evaluation data after tuning. Main Results. We compared the results of our algorithm with ground-truth labels. The algorithm is evaluated on data of human patients and publicly available labeled non-human primates (NHPs) datasets. The average accuracy of BAR on datasets of human patients is 92.3% which is further reduced to 88.03% after (K-means + CAOM). In addition, the average accuracy of BAR on a publicly available labeled dataset of NHPs is 95.40% which reduces to 86.95% after (K-mean + CAOM). Lastly, we compared the performance of the SpikeDeep-Classifier with two human experts, where SpikeDeep-Classifier has produced comparable results. Significance. The results demonstrate that “SpikeDeep-Classifier” possesses the ability to generalize well on a versatile dataset and henceforth provides a generalized well on a versatile dataset and henceforth provides a generalized and fully automated solution to offline spike sorting.},
keywords = {BCI, CNN, Machine Learning, Spike Sorting},
pubstate = {published},
tppubtype = {article}
}
Objective. Advancements in electrode design have resulted in micro-electrode arrays with hundreds of channels for single cell recordings. In the resulting electrophysiological recordings, each implanted electrode can record spike activity (SA) of one or more neurons along with background activity (BA). The aim of this study is to isolate SA of each neural source. This process is called spike sorting or spike classification. Advanced spike sorting algorithms are time consuming because of the human intervention at various stages of the pipeline. Current approaches lack generalization because the values of hyperparameters are not fixed, even for multiple recording sessions of the same subject. In this study, a fully automatic spike sorting algorithm called “SpikeDeep-Classifier” is proposed. The values of hyperparameters remain fixed for all the evaluation data. Approach. The proposed approach is based on our previous study (SpikeDeeptector) and a novel background activity rejector (BAR), which are both supervised learning algorithms and an unsupervised learning algorithm (K-means). SpikeDeeptector and BAR are used to extract meaningful channels and remove BA from the extracted meaningful channels, respectively. The process of clustering becomes straight-forward once the BA is completely removed from the data. Then, K-means with a predefined maximum number of clusters is applied on the remaining data originating from neural sources only. Lastly, a similarity-based criterion and a threshold are used to keep distinct clusters and merge similar looking clusters. The proposed approach is called cluster accept or merge (CAOM) and it has only two hyperparameters (maximum number of clusters and similarity threshold) which are kept fixed for all the evaluation data after tuning. Main Results. We compared the results of our algorithm with ground-truth labels. The algorithm is evaluated on data of human patients and publicly available labeled non-human primates (NHPs) datasets. The average accuracy of BAR on datasets of human patients is 92.3% which is further reduced to 88.03% after (K-means + CAOM). In addition, the average accuracy of BAR on a publicly available labeled dataset of NHPs is 95.40% which reduces to 86.95% after (K-mean + CAOM). Lastly, we compared the performance of the SpikeDeep-Classifier with two human experts, where SpikeDeep-Classifier has produced comparable results. Significance. The results demonstrate that “SpikeDeep-Classifier” possesses the ability to generalize well on a versatile dataset and henceforth provides a generalized well on a versatile dataset and henceforth provides a generalized and fully automated solution to offline spike sorting.
2019
Muhammad Saif-ur-Rehman; Robin Lienkämper; S Dyck; A Rayana; Y Parpaley; J Wllner; C Liu; B Lee; S Kellis; D Manahan-Vaughn; O Güntürkün; R A Andersen; Ioannis Iossifidis; T Galmachers; C Klaes
Universal SpikeDeeptector Miscellaneous
2019.
Abstract | BibTeX | Tags: BCI, CNN, Machine Learning, Spike Detection, Spike Sorting
@misc{ur-reimann2019a,
title = {Universal SpikeDeeptector},
author = {Muhammad Saif-ur-Rehman and Robin Lienkämper and S Dyck and A Rayana and Y Parpaley and J Wllner and C Liu and B Lee and S Kellis and D Manahan-Vaughn and O Güntürkün and R A Andersen and Ioannis Iossifidis and T Galmachers and C Klaes},
year = {2019},
date = {2019-01-01},
urldate = {2019-01-01},
publisher = {SfN 2019},
abstract = {State-of-the-art microelectrode array technology enables simultaneous, large-scale single unit recordings from hundreds of channels. Identification of channels recording neural data as compared to noise is the first step for all further analyses. Automatizing this process aims at minimizing the human involvement and time for manual curation. In our previous study, we introduced the “SpikeDeeptector” (SD), which enables us to automatically detect and track channels containing neural data from different human patients implanted with different types of microelectrodes across different brain areas. SD works on human data and to some extent on the data of non-human primates (NHPs). However, to make SD more versatile we proposed a more generalized method called “Universal SpikeDeeptector (USD)”, which is an extended version of SD. USD intends to detect and track the channels containing neural data recorded from four different species (rats, ravens, NHPs and humans) using different kinds of microelectrodes and different recording sites. To our knowledge, there is no method that can simultaneously detect and track neural data of multiple species. To enable contextual learning, USD constructs a feature vector from a batch of waveforms. The constructed feature vectors are then fed into a deep-learning algorithm, which learns contextualized, temporal and spatial patterns. USD is a supervised learning method. Therefore, it requires labeled data for training. It is mainly trained on data from a single human tetraplegic patient, and a small but equal portion of data from the remaining three species. The trained model is then evaluated on a test dataset collected from several humans, NHPs, rats, and birds. The results show that the USD performed consistently well across data collected from each species.},
keywords = {BCI, CNN, Machine Learning, Spike Detection, Spike Sorting},
pubstate = {published},
tppubtype = {misc}
}
State-of-the-art microelectrode array technology enables simultaneous, large-scale single unit recordings from hundreds of channels. Identification of channels recording neural data as compared to noise is the first step for all further analyses. Automatizing this process aims at minimizing the human involvement and time for manual curation. In our previous study, we introduced the “SpikeDeeptector” (SD), which enables us to automatically detect and track channels containing neural data from different human patients implanted with different types of microelectrodes across different brain areas. SD works on human data and to some extent on the data of non-human primates (NHPs). However, to make SD more versatile we proposed a more generalized method called “Universal SpikeDeeptector (USD)”, which is an extended version of SD. USD intends to detect and track the channels containing neural data recorded from four different species (rats, ravens, NHPs and humans) using different kinds of microelectrodes and different recording sites. To our knowledge, there is no method that can simultaneously detect and track neural data of multiple species. To enable contextual learning, USD constructs a feature vector from a batch of waveforms. The constructed feature vectors are then fed into a deep-learning algorithm, which learns contextualized, temporal and spatial patterns. USD is a supervised learning method. Therefore, it requires labeled data for training. It is mainly trained on data from a single human tetraplegic patient, and a small but equal portion of data from the remaining three species. The trained model is then evaluated on a test dataset collected from several humans, NHPs, rats, and birds. The results show that the USD performed consistently well across data collected from each species.
Muhammad Saif-ur-Rehman; Robin Lienkämper; Yaroslav Parpaley; Jörg Wellmer; Charles Liu; Brian Lee; Spencer Kellis; Richard Andersen; Ioannis Iossifidis; Tobias Glasmachers; Christian Klaes
SpikeDeeptector: a deep-learning based method for detection of neural spiking activity Journal Article
In: Journal of Neural Engineering, vol. 16, no. 5, pp. 056003, 2019.
Abstract | Links | BibTeX | Tags: BCI, CNN, Data Reduction, Machine Learning, Spike Sorting
@article{Saif-ur-Rehman2019,
title = {SpikeDeeptector: a deep-learning based method for detection of neural spiking activity},
author = {Muhammad Saif-ur-Rehman and Robin Lienkämper and Yaroslav Parpaley and Jörg Wellmer and Charles Liu and Brian Lee and Spencer Kellis and Richard Andersen and Ioannis Iossifidis and Tobias Glasmachers and Christian Klaes},
url = {https://iopscience.iop.org/article/10.1088/1741-2552/ab1e63/meta},
doi = {10.1088/1741-2552/ab1e63},
year = {2019},
date = {2019-01-01},
urldate = {2019-01-01},
journal = {Journal of Neural Engineering},
volume = {16},
number = {5},
pages = {056003},
abstract = {Objective . In electrophysiology, microelectrodes are the primary source for recording neural data (single unit activity). These microelectrodes can be implanted individually or in the form of arrays containing dozens to hundreds of channels. Recordings of some channels contain neural activity, which are often contaminated with noise. Another fraction of channels does not record any neural data, but only noise. By noise, we mean physiological activities unrelated to spiking, including technical artifacts and neural activities of neurons that are too far away from the electrode to be usefully processed. For further analysis, an automatic identification and continuous tracking of channels containing neural data is of great significance for many applications, e.g. automated selection of neural channels during online and offline spike sorting. Automated spike detection and sorting is also critical for online decoding in brain–computer interface (BCI) applications, in which on...},
keywords = {BCI, CNN, Data Reduction, Machine Learning, Spike Sorting},
pubstate = {published},
tppubtype = {article}
}
Objective . In electrophysiology, microelectrodes are the primary source for recording neural data (single unit activity). These microelectrodes can be implanted individually or in the form of arrays containing dozens to hundreds of channels. Recordings of some channels contain neural activity, which are often contaminated with noise. Another fraction of channels does not record any neural data, but only noise. By noise, we mean physiological activities unrelated to spiking, including technical artifacts and neural activities of neurons that are too far away from the electrode to be usefully processed. For further analysis, an automatic identification and continuous tracking of channels containing neural data is of great significance for many applications, e.g. automated selection of neural channels during online and offline spike sorting. Automated spike detection and sorting is also critical for online decoding in brain–computer interface (BCI) applications, in which on...