Unsupervised automatic online spike sorting using reward-based online clustering

From End to End: Gaining, Sorting, and Employing High-Density Neural Single Unit Recordings

“…Sequentially constructed algorithms, such as those building upon multiple basic dense layers (Mahallati et al, 2019 ; Yeganegi et al, 2020 ) and convolutional (Li et al, 2020b ) and recurrent layers (Rácz et al, 2020 ) require an expansive repository, although by weights' and activation functions' binarization, complexity may be cut back (Valencia and Alimohammad, 2021 ), or parallelization by graphical processing units may take place (Tam and Yang, 2018 ). These layers may be constructed in different ways, mainly in order to mitigate or abandon the need for hand-labeled neural data throughout training: autoencoders (Weiss, 2019 ; Radmanesh et al, 2021 ; Rokai et al, 2021 ) or networks generated by adversarial (Wu et al, 2019 ; Ciecierski, 2020 ) or reinforcement learning paradigms (Salman et al, 2018 ; Moghaddasi et al, 2020 ) have successfully clustered features originating from noisiest datasets. Likewise, a more sophisticated learning-based method may even incorporate multiple steps of spike sorting, resolving detection, feature extraction, and clustering as a close-packed solution (Eom et al, 2021 ; Rokai et al, 2021 ), although manual curation is advisable (Horváth et al, 2021 ).…”

Functional Clustering of Neuronal Signals with FMM Mixture Models

“…In this work, ten different datasets have been analyzed. They were selected because they encompass a wide variety of spike waveforms -real and simulated-; the real class labels (groundtruth) are known; and they are widely used as benchmarking data in the literature, as in [43,47,25,45], among others. Table 1 gives information on the number of clusters and signals, and the main references.…”

“…Spike Sorting (SS) is the collection of techniques to identify spikes corresponding to different neurons. The correct identification of spikes is crucial for studying the connectivity patterns between close-by neurons [6], relating the firing of certain neurons to the memory process [33], the treatment of epileptic patients [42], or the development of high-accuracy brainmachine interfaces [25], among many other questions. However, SS remains one of the most challenging open problems in neuroscience.…”

Statistical oscillatory models to solve problems in neuroscience

“…Algunos retos que surgen de analizar los datos neuronales son los siguientes: (a) las señales suelen tener bastante ruido, especialmente si han sido tomadas de forma no-invasiva; (b) los spikes son altamente asimétricos, lo que imposibilita el uso de ciertos métodos; (c) las señales neuronales requieren ser caracterizadas con medidas de escala, localización y forma debido a la amplia variedad de patrones que exhiben; (d) si dos neuronas se disparan simultáneamente, las formas de sus spikes se superponen y su separación no es sencilla; (e) en los análisis en tiempo real, el alto volumen de datos restringe el uso de modelos complejos. Algunos avances recientes en neurociencia gracias a los modelos de spikes son los siguientes: entender el funcionamiento del sistema nervioso, principalmente mediante la definición de subtipos neuronales (Zeng y Sanes, 2017); análisis de los procesos de aprendizaje y memorización (Donato, Rompani y Caroni, 2013;Rey, Pedreira y Quiroga, 2015), estudio de la conectividad entre neuronas (Buzsáki, 2004), tratamiento de pacientes con epilepsia (Sharma et al, 2017), e incluso el desarrollo de interfaces cerebro-computadora de alta precisión (Moghaddasi et al, 2020).…”

“…Some challenges that arise from the analysis of neuronal spike data are the following: (a) signals have high levels of noise, especially if recorded non-invasively; (b) spikes are highly asymmetrical, making unsuitable some traditionally popular models; (c) neuronal signals display a wide variety of patterns which require a precise characterization of the waveforms, not just timing and amplitude; (d) if two neurons fire simultaneously, their spikes are recorded overlapping, and its separation is not trivial; (e) in real-time analyses, the high data volume limits the use of complex models. Some recent advances in neuroscience due to spike models are the following: understanding of the neuronal system functioning, mainly by defining neuronal subpopulations (Zeng and Sanes, 2017), comprehension of learning and memory processes (Donato, Rompani, and Caroni, 2013;Rey, Pedreira, and Quiroga, 2015), study of connectivity between close-by neurons (Buzsáki, 2004), treatment of epileptic patients (Sharma et al, 2017), and even development of high-accuracy brain-machine interfaces (Moghaddasi et al, 2020). Nowadays, there are still many open problems in the neuroscience literature.…”