Neurtex Manual Spike Sorting Contest

July 3, 2024 | News

Neurtex Manual Spike Sorting Contest

Announcing the 2024 manual spike sorting contest by the Neurtex Brain Research Institute. This contest offers a $5000 cash first prize and other prizes for the most accurate spike sorting of a 120 channel set of simulated and actual human intracranial microwire recordings.

This contest is designed to assess how well the spike sorting techniques normally used in this type of recording detect single neuron activity. By using a dataset with some simulated recordings, ground truth is known and spike sorting outputs can be compared to determine accuracy. Most techniques employed to date have used some form of manual adjustment or curation of the results. At the same time, the accuracy of these adjustments in typical recordings has not been established.

Prizes in this contest will be based on how accurately spike sorting outputs agree with known simulated spike times as well as the ability to identify real versus simulated recordings.

Timeframe

The recordings to be sorted will be available as a set of Neuralynx continuously sampled .Ncs files starting September 27, 2024 at http://steinmetz.org/peter/Set1.zip or https://osf.io/92z3n/ . Submissions are due by 11:59 PM Eastern Standard Time, January 30, 2025. Winners will be announced by March 30, 2025.

Dataset

The recordings to be sorted consist of 120 channels in the Neuralynx continuously sampled, .Ncs, format.  They are provided in a single zip archive. Each recording is 11 minutes long, sampled at 32,000 samples/s, and 43 MB long. The recordings are a mixture of actual and simulated recordings and may contain zero or a small number of action potential extracellular waveforms. The simulations are performed using background noise and activity per Steinmetz, 2024 with up to 4 waveform shapes added.

Judging

Accuracy of Spike Sorting

First ($5000) and second ($3000) prizes will be awarded for the spike sorting which results in the highest level of total agreement (AMIall, Steinmetz, 2017) between those output clusters identified as single unit activity and the simulated times of action potentials. This will be measured for only the simulated recording channels.

Distinguishing Real versus Simulated Recordings

A third prize ($2000) will be provided for the most accurate identification of those recordings which are simulated versus actual, based on the highest level of agreement between a provided set of labels and the actual type of recording.

In the unlikely event of ties for a prize, the cash value of the prize will be split. Each participant may win only one prize. All decisions regarding prizes and administration of the contest will be made by Neurtex scientific personnel and are final.

Eligible Sorting Techniques

Any desired sorting technique may be applied, whether involving manual steps or fully automated. Participants are encouraged to use whatever manual steps they normally would apply. The submission form requests descriptive details of the technique used. The descriptions must be complete enough to allow a worker in the field to thoroughly understand the steps taken.

Eligible Contestants

This contest is open to anyone in the world, but awards will be paid in US dollars and per US law.

Submission of Entries

A separate submission form is available at https://neurtex.org/manual-spike-sorting-contest-submission/ .

Limits on Entries and Early Closure

This contest is limited to 40 entries and may be closed early. Such early closure will be noted on this page and the submission form.

Publication of Results

The results of this contest may be published. Contestants agree to the use of their submissions in such publications and may participate as co-authors if desired.

Further Information

For questions or further information, please email contest24@neurtex.org .

A small single page flyer announcing the contest which can be used to notify departments and people about the contest is available at Flyer .