ITR Atlas
← All interfaces

BrainGate2 Cursor BCI (ReFIT-KF), pointing (Pandarinath et al., 2017)

Measured by Pandarinath, Nuyujukian, Henderson, Shenoy et al. · eLife 6:e18554 (2017)

Intracortical Cursor invasive 2017

Inputs

The measured or assumed values behind the calculations, each with its source.

B = 3.7 bits/s
Mean achieved bitrate, best participant (T5), grid target task (Results). T6: 2.2 bits/s; T7: 1.4 bits/s.
method = log2(targets)/selection, net of errors
Standard achieved-bitrate for 2D target acquisition, reported directly by the authors. This paper credits log2(N−1) per net-correct selection (Methods, citing Nuyujukian et al. 2015); the 2015 source paper's own Eq. 1 uses log2(N).

Strictest ITR

Each scoring method is an upper bound on the channel, so the headline is the strictest (smallest) one for this entry. Use the score selector on the home page to view any single method across entries.

Strictest Fitts' law Recomputed
Fitts' law throughput on the grid task
2D pointing channel (apples-to-apples with the mouse)
1.48 bits/s

  1. Grid geometry → movement difficulty

    6×6 board: cell width W = S/6; random targets → mean amplitude A ≈ 0.52·S, so A/W ≈ 3.1.  ID = log2(A/W + 1) = log2(4.1) ≈ 2.05 bits/movement.

    S = board span; 0.52 is the mean distance between two random points on a square. The target is cued, so each correct selection is a pointing movement carrying the Fitts index of difficulty (≈2.05 bits), not the log2(35) ≈ 5.13 bits the achieved-bitrate metric credits.

  2. Re-credit the reported achieved bitrate onto the Fitts basis

    ITR = B × ID / log2(N − 1) = 3.7 × 2.05 / 5.13 = 1.48 bits/s

    The paper reports only the achieved bitrate B = 3.7 bits/s (T5) and no independent selection timing, so the Fitts figure is that same measured throughput re-credited per movement at the Fitts ID instead of log2(N−1). This is a unit re-crediting of the reported score, not an independent measurement. Same correction applied to Neuralink Webgrid and the Card 2026 grid, putting all three intracortical cursors on the mouse's Fitts basis (mouse 4.5, stylus 4.9, trackball 3.3 bits/s).

What counts as a bit depends on the action space. The number of distinguishable actions and how likely each one is are design choices of the task, not the sensing hardware. The same modality can present a fixed set of targets, a set pruned per step by a grammar or language model, or a continuous control space. Each of these changes how many actions are live and how the probability mass is spread, and therefore the information per selection. Read the action space below before comparing headline numbers across entries.

Action space

What the user can produce at each step, and how those options are distributed.

Structure
Continuous control space
Size
Continuous
Prior
Uniform: all actions assumed equally likely
Notes
A 2D point-and-click cursor from the BrainGate2 pilot clinical trial (BrainGate Neural Interface System), driven by a ReFIT Kalman-filter decoder; throughput is measured by how fast and accurately it acquires randomly placed grid targets. ReFIT is the decoder, not the system. This is a continuous control channel, not a discrete speller. The authors' headline uses the field-standard achieved-bitrate metric (log2(N-1) per net-correct cued selection, as defined in this paper citing Nuyujukian et al. 2015; the 2015 original uses log2(N)), while the atlas pointing convention uses a Fitts-style movement difficulty on the same grid. The reference number here is therefore a Fitts throughput for comparability with mouse, trackball and stylus entries; the log2(N-1) achieved-bitrate is kept as a secondary, as-reported figure. This entry is the pointing channel; the same system's real-world copy-typing application is a separate entry (BrainGate2 Cursor BCI (ReFIT-KF), text entry), which is what the participants actually communicated.

Comparability The strictest bound here is Fitts throughput: the index of difficulty, log₂(A/W + 1), per movement. Directly comparable to the other continuous-pointing entries (mouse, trackball, stylus, gaze and the cursor BCIs). Set against the text entries (keyboards, spellers, speech) it crosses methods: both report bits/s, but one measures movement difficulty and the other text information, so compare within the family first.

Other score types

Bounds the atlas keeps out of the default strictest headline: as-reported figures, alternate task conditions, or raw-channel ceilings that shouldn't win the headline by default. Each still carries a score type, so the home-page selector ranks this entry on it when you choose that type. Read its derivation before comparing across entries.

Nuyujukian Author-reported · reproduced
Nuyujukian achieved bitrate (log2(N−1)), as reported by the authors
Achieved-bitrate task metric, shown for comparison
3.7 bits/s

  1. Achieved-bitrate metric

    6×6 grid → N = 36 targets; this paper credits log2(N − 1) = log2(35) ≈ 5.13 bits per net-correct cued selection (Pandarinath et al. 2017, Methods, citing Nuyujukian et al. 2015). The 2015 source paper's Eq. 1 uses log2(N) instead, so its 8-target radial task scores log2(8) = 3.0 bits per net selection, not log2(7) ≈ 2.8.

  2. Authors' reported score (taken as reported, not re-derived)

    B = 3.7 ± 0.4 bits/s  (T5, 6×6 grid; T6 2.2, T7 1.4; T5 reached 4.16 on a dense 9×9 grid).

    Reported directly in Results. The paper gives only the bitrate, with no separate selections/s or acquisition time. There is no more-primary quantity to derive it from: it is author-reported. Dividing by the 5.13-bit credit implies ~0.72 selections/s, but that rate is a consequence of the score, not an independent measurement.

Source

Authors
Pandarinath, Nuyujukian, Henderson, Shenoy et al.
Publication
eLife 6:e18554, 2017
Paper
10.7554/eLife.18554
Reference
Open-access full text (eLife)