Pith. sign in

REVIEW 2 major objections 5 minor 19 references

Towards Deep Learning-Based EEG Electrode Detection Using Automatically Generated Labels

T0 review · 2 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper demonstrates that EEG electrode positions can be learned by CNNs trained on automatically generated labels from a robot-mounted phantom, reaching 5.69 mm accuracy at real-time speed.

desk verdict The label-generation pipeline is a genuinely useful contribution, but the reported accuracy may simply reflect head-pose regression on a rigid phantom, not electrode detection. read the letter →

arxiv 1908.04186 v1 pith:TGQGXZY2 submitted 2019-08-12 cs.CV

Nils Gessert , Martin Gromniak , Marcel Bengs , Lars Matthäus , Alexander Schlaefer This is my paper · ORCID
classification cs.CV
keywords EEGelectrodedetectionelectroencephalographydeeplearningconvolutionalneuralnetworksRGBDcameraautomaticlabelgenerationrobotichand-eyecalibrationmulti-outputregression
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether convolutional neural networks can find EEG electrodes directly in RGBD camera images, and whether the training labels can be produced automatically instead of by hand. It answers yes to both on a robot-mounted head phantom: after one manual annotation is propagated through hand-eye calibration to thousands of poses, a CNN regressing electrode coordinates reaches a mean error of $5.69 \pm 6.10$ mm against manual labels, with inference times of 34 to 58 ms. A reader should care because conventional electrode localization with a stylus is slow and expensive, while a learned camera-based method could work in changing environments such as ambulances.

What carries the argument

The machinery is a label-propagation chain built on hand-eye calibration: one manual 2D annotation is converted to a camera-frame pose $^{K}T_{EL_i}$, moved into the robot end-effector frame via $^{EF}T_{EL_i} = (^{R}T_{EF})^{-1}\,^{R}T_K\,^{K}T_{EL_i}$, then re-projected to every other camera pose with $^{K}T^j_{EL_i} = (^{R}T_K)^{-1}\,^{R}T^j_{EF}\,^{EF}T_{EL_i}$. The same transformations generate both 3D labels for the EEG use case and 2D pixel labels for image-based training. The learner is a convolutional regression network with mean-squared-error loss and a fixed-size output of all $N$ electrode positions, which forces predictions even for occluded electrodes.

What would settle it

Run the trained network on a real human head wearing a comparable EEG cap and compare the predicted electrode positions against an optical digitizer or MRI ground truth; if the mean absolute error against real heads is far above the $5.69$ mm phantom result, the central feasibility claim for practical use fails.

Watch

Extended reading notes

Core claim

The central discovery is that a standard CNN regression model—DenseNet121 or SE-ResNeXt50 taking RGB, depth, or RGBD crops of 270×254 pixels and outputting $N \times 3$ coordinates—can learn electrode locations from automatically generated labels. Labels are created by annotating one image, lifting those pixels to 3D via the depth point cloud, and transforming the resulting poses to every other robot pose with hand-eye calibration; the $4.76$ mm mean gap between generated and manual labels sets an upper bound on achievable accuracy. Against manual labels, the best model reaches $5.69 \pm 6.10$ mm mean absolute error in 3D, with pixel-level 2D localization even more accurate (about 1.3 pixels). The paper also shows that depth information adds nothing on this fixed phantom, that both architectures perform similarly, and that the approach tolerates partially occluded electrodes because labels are still propagated for invisible positions.

Load-bearing premise

The whole demonstration uses a single head phantom with one EEG cap, and the hand-eye calibration that propagates one manual annotation stands in for real human heads; if CNNs trained on phantom images do not transfer to real patients, the practical benefit is not delivered.

Editorial extensions

If this is right

  • Real-time electrode localization becomes possible, since both networks run in tens of milliseconds compared with seconds to minutes for earlier photogrammetric approaches.
  • Training data can be generated at scale without manual annotation: one annotated image plus logged robot poses produces labels for thousands of camera views.
  • The automatically generated labels are accurate enough to serve as a training signal, and the $4.76$ mm gap between generated and manual labels bounds how well any model trained on them can do.
  • Depth input adds little on the fixed phantom, indicating that RGB alone captures most of the relevant information in that setting.
  • The same pipeline can produce arbitrarily large datasets for studying head shape, cap type, and pose variation before attempting real-world transfer.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If depth becomes important only when head shape and size vary, a direct experiment is training on several phantom heads and checking whether RGBD surpasses RGB; the paper's fixed-phantom results predict that crossover.
  • Because the model always outputs all electrode positions, even occluded ones, a natural extension is adding a visibility or confidence head and penalizing confident predictions on occluded electrodes.
  • A cap pre-segmentation step, as the authors suggest for future work, would decouple detection from the underlying head and could be the key to transferring phantom-trained models to real humans.
  • The calibration error directly sets the model error floor, so repeating the pipeline with more calibration poses or a higher-accuracy calibration method should lower the achievable MAE in a measurable way.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper proposes an automatic labeling pipeline for CNN-based EEG electrode detection. A robotic arm moves a head phantom with a fixed EEG cap while an RGBD camera records images; electrode positions are annotated manually in one image and propagated to all poses via hand-eye calibration. CNNs (DenseNet121 and SE-ResNext50) are trained to regress the 3D or 2D positions of 8 electrodes from RGB, depth, or RGBD crops. Evaluation on 150 manually annotated images reports a best 3D MAE of 5.69±6.1 mm and shows that the generated labels themselves differ from manual labels by 4.76±5.1 mm. The authors conclude that deep learning-based electrode detection is feasible and that the labeling scheme is a useful testbed.

Significance. The automatic labeling scheme is a sensible way to generate large training sets without per-image manual annotation, and the use of an independent manual test set is a sound evaluation protocol. The paper also reports calibration error explicitly, which is helpful. If the reported accuracy reflected genuine electrode detection, real-time electrode localization from a single RGBD image would be practically valuable for EEG source localization. However, as analyzed below, the current experimental design cannot separate electrode-specific learning from regression of the visible robot/head pose, so the central feasibility claim is not yet established.

major comments (2)
  1. [Section 2.2, Eqs. (1)-(2), Table 1] The 3D labels are deterministic functions of the logged end-effector pose and the hand-eye calibration, since the head phantom and cap are rigid. The robot arm and workspace are visible in the RGBD input, so a network can achieve low error by regressing head/end-effector pose and outputting the fixed 8-electrode layout without ever detecting electrodes. The "Gen. Labels" row in Table 1 shows that simply using the logged pose-derived labels as predictions gives MAE 4.76±5.1 mm against manual ground truth, which is better than the best CNN result (5.69±6.1 mm for SE-Resnext50 RGBD). Thus the reported accuracy does not demonstrate electrode-specific visual learning. Please add a pose-only baseline and ablations that break the pose-electrode correlation (e.g., different cap layouts, occluded electrodes, or randomized cap placement), or reframe the claim accordingly.
  2. [Section 2.3] The fixed-size output of N×3 (or N×2) forces the network to predict all electrode positions even when they are not visible, as the paper itself notes. This design makes the task a fixed-output regression from an image to eight 3D points, not detection/localization of an unknown set of visible electrodes. Combined with the pose-electrode correlation in Eq. (1)-(2), this further weakens the "detection" claim. Please evaluate with variable electrode visibility/count or use a detection formulation (e.g., heatmap or object detector) to demonstrate that individual electrodes are actually localized.
minor comments (5)
  1. [Section 2.2] The notation in Eqs. (1)-(2), such as "K T j ELi", is not fully defined; please specify the source and target frames for the transformation matrices and explicitly state the identity-orientation assumption for the electrode poses.
  2. [Table 1] The "Gen. Labels" row is not a network model but a calibration-based label-generation baseline; clarify this in the caption or text. Also, the rMAE columns appear to be scaled by 10^-3; please state this explicitly.
  3. [Abstract and Section 4] The claim that "deep learning-based electrode detection is feasible" should be qualified as "on a head phantom with a fixed EEG cap and known robot poses," given the acknowledged limitation in the Discussion that the method is currently limited to head phantoms.
  4. [Section 3] The manual annotations serve as ground truth, but no inter- or intra-annotator variability is reported, and it is unclear whether a single annotator labeled all 150 test images. Please report annotation reliability or justify the assumption that manual labels are accurately centered.
  5. [Section 2.3] The training/validation split is not specified; please state how many of the 3000 images are used for training versus validation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the robot-generated labels are a data-production pipeline, the test set is manually annotated, and no reported quantity reduces to a fitted parameter or to a self-citation.

full rationale

The paper's automatic label generation (Section 2.2, Eqs. 1-2) is a data-annotation procedure, not a fitted model: one manually labeled image is transformed through hand-eye calibration and logged robot poses to create training labels. The evaluation set of 150 images is manually annotated independently of this pipeline, and the reported CNN errors are computed against that manual ground truth. Thus the central claim, that CNNs can predict electrode locations with an MAE of 5.69 +/- 6.10 mm, is not equivalent by construction to the training labels or to any fitted parameter. The only self-citation is [13] for the QR24 calibration method, which is used as a standard algorithmic tool and is not load-bearing for the paper's quantitative conclusions. The skeptic's observation that the generated labels are deterministic functions of robot pose, and that a pose-only baseline might perform comparably, is a potential experimental confound or external-validity concern, not circularity: it does not make the paper's prediction equal to its input by definition. No equation in the paper reduces a reported result to a parameter fitted to that same result, and no load-bearing uniqueness claim is imported from the authors' prior work. Therefore the derivation chain is self-contained with respect to the manual test annotations, and the appropriate circularity score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

This paper is an empirical engineering study with no theoretical derivation. The central claim depends on the accuracy of the calibration and labeling pipeline, on the independent manual test labels, and on the assumption that the experimental setup represents a useful environment for developing electrode detection. The listed free parameters are standard training hyperparameters; no parameter was fitted to the test data.

free parameters (4)
  • Initial learning rate = 1e-5
    Chosen by hand for fine-tuning pretrained CNNs; no sensitivity analysis is reported.
  • Batch size = 10
    Chosen by hand; no justification or sensitivity analysis is given.
  • Training epochs = 300
    Fixed at 300 epochs with learning rate halved every 50 epochs; no early stopping or evidence that this is optimal.
  • Network input size = 270 x 254 pixels
    Determined by cropping to the extent of the ground-truth annotations plus a margin; affects the spatial scale of the regression task.
assumptions (5)
  • domain assumption The hand-eye calibration between robot and camera (computed with QR24) is unbiased, with position error 3.910 mm and rotation error 0.780 degrees as reported.
    All automatically generated labels are produced by transforming a single manual annotation through these calibration matrices (Eq. 1 and 2). If the calibration error were systematic, the training labels would be systematically biased.
  • domain assumption The Kinect V2 manufacturer-provided RGB-to-depth calibration accurately aligns the two sensors.
    The RGB images are mapped into the depth coordinate frame using manufacturer calibrations (Section 2.2); errors here propagate to both the 2D and 3D labels.
  • domain assumption The head phantom and EEG cap are rigid and do not deform between robot poses.
    The electrode poses are assumed fixed relative to the end effector (EF_T_ELi is computed once and reused for all poses); any slippage of the cap or flex of the phantom would introduce label errors not captured by the calibration error.
  • domain assumption A single manual annotation of the center pixel of each electrode in one image provides correct 3D positions when back-projected through the depth point cloud.
    All training labels derive from this one annotation; the only check is the comparison of generated labels to a separate manual labeling of 150 test images.
  • ad hoc to paper Identity orientation is assumed for electrode poses when deriving EF_T_ELi from the initial 3D positions (Section 2.2).
    The initial electrode poses KT_ELi are assigned identity rotation, which is unphysical but only affects orientation, not position; positions are what matter for the MAE metric.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Deep Learning-Based EEG Electrode Detection Using Automatically Generated Labels." pith.science (2026). https://pith.science/paper/TGQGXZY2

@misc{pith2026190804186,
  author       = {Pith},
  title        = {Pith review of: Towards Deep Learning-Based EEG Electrode Detection Using Automatically Generated Labels},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TGQGXZY2}},
  note         = {Machine review of arXiv:1908.04186}
}
read the original abstract

Electroencephalography (EEG) allows for source measurement of electrical brain activity. Particularly for inverse localization, the electrode positions on the scalp need to be known. Often, systems such as optical digitizing scanners are used for accurate localization with a stylus. However, the approach is time-consuming as each electrode needs to be scanned manually and the scanning systems are expensive. We propose using an RGBD camera to directly track electrodes in the images using deep learning methods. Studying and evaluating deep learning methods requires large amounts of labeled data. To overcome the time-consuming data annotation, we generate a large number of ground-truth labels using a robotic setup. We demonstrate that deep learning-based electrode detection is feasible with a mean absolute error of 5.69 +- 6.1mm and that our annotation scheme provides a useful environment for studying deep learning methods for electrode detection.

Figures

Figures reproduced from arXiv: 1908.04186 by the authors.

Figure 1
Figure 1. The experimental setup we use for data acquisition and label generation. Red crosses represent electrodes on the EEG cap. 2 Material and Methods 2.1 Experimental Setup Our experimental setup for data acquisition is shown in [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Left, a cropped RGB image with 2D electrode locations is shown. The black pixels are artifacts caused by the transformation of the RGB image to the depth sensor coordinate frame. Right, a point cloud with a zoom on the EEG cap with 3D electrode locations is shown. Green marks indicate the manually annotated electrode locations. Red marks indicate the automatically generated labels. Blue marks indicate the model pred… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [1]

    (2015) A framework to automate assessment of upper-limb motor function impairment: A feasibility study

    Otten, P., Kim, J., Son, S. (2015) A framework to automate assessment of upper-limb motor function impairment: A feasibility study. Sensors 15(8), 20,097–20,114

    work page 2015

  2. [2]

    (2008) Eeg source localization in focal epilepsy: where are we now? Epilepsia 49(2), 201–218

    Plummer, C., Harvey, A.S., Cook, M. (2008) Eeg source localization in focal epilepsy: where are we now? Epilepsia 49(2), 201–218

    work page 2008

  3. [3]

    (1958) Report of the committee on methods of clinical examination in electroencephalography

    Jasper, H. (1958) Report of the committee on methods of clinical examination in electroencephalography. Elec- troencephalogr Clin Neurophysiol 10, 370–375

    work page 1958

  4. [4]

    (2007) 10/20, 10/10, and 10/5 systems revisited: their validity as relative head- surface-based positioning systems

    Jurcak, V., Tsuzuki, D., Dan, I. (2007) 10/20, 10/10, and 10/5 systems revisited: their validity as relative head- surface-based positioning systems. Neuroimage 34(4), 1600–1611

    work page 2007

  5. [5]

    (1993) The spatial location of eeg electrodes: locating the best-fitting sphere relative to cortical anatomy

    Towle, V.L., Bola˜ nos, J., Suarez, D., Tan, K., Grzeszczuk, R., Levin, D.N., Cakmur, R., Frank, S.A., Spire, J.P. (1993) The spatial location of eeg electrodes: locating the best-fitting sphere relative to cortical anatomy. Electroencephalography and clinical neurophysiology 86(1), 1–6

    work page 1993

  6. [6]

    (1998) Scalp- recorded eeg localization in mri volume data

    Brinkmann, B.H., O’Brien, T.J., Dresner, M.A., Lagerlund, T.D., Sharbrough, F.W., Robb, R.A. (1998) Scalp- recorded eeg localization in mri volume data. Brain topography 10(4), 245–253

    work page 1998

  7. [7]

    (2011) A single camera photogrammetry system for multi-angle fast localization of eeg electrodes

    Qian, S., Sheng, Y. (2011) A single camera photogrammetry system for multi-angle fast localization of eeg electrodes. Annals of biomedical engineering 39(11), 2844

    work page 2011

  8. [8]

    (2015) Using a motion capture system for spatial localization of eeg electrodes

    Reis, P.M., Lochmann, M. (2015) Using a motion capture system for spatial localization of eeg electrodes. Frontiers in neuroscience 9, 130

    work page 2015

Show all 19 references
  1. [9]

    (2017) Photogrammetry-based head digitization for rapid and accurate localization of eeg electrodes and meg fiducial markers using a single digital slr camera

    Clausner, T., Dalal, S.S., Crespo-Garc´ ıa, M. (2017) Photogrammetry-based head digitization for rapid and accurate localization of eeg electrodes and meg fiducial markers using a single digital slr camera. Frontiers in neuroscience 11, 264

    2017

  2. [10]

    (2015) Faster r-cnn: Towards real-time object detection with region proposal networks

    Ren, S., He, K., Girshick, R., Sun, J. (2015) Faster r-cnn: Towards real-time object detection with region proposal networks. In: Advances in neural information processing systems, pp. 91–99

    2015

  3. [11]

    (2019) Hyperface: A deep multi-task learning framework for face detection, landmark localization, pose estimation, and gender recognition

    Ranjan, R., Patel, V.M., Chellappa, R. (2019) Hyperface: A deep multi-task learning framework for face detection, landmark localization, pose estimation, and gender recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence 41(1), 121–135

    2019

  4. [12]

    (2000) The OpenCV Library

    Bradski, G. (2000) The OpenCV Library. Dr. Dobb’s Journal of Software Tools

    2000

  5. [13]

    (2012) Non-orthogonal tool/flange and robot/world calibration

    Ernst, F., Richter, L., Matth¨ aus, L., Martens, V., Bruder, R., Schlaefer, A., Schweikard, A. (2012) Non-orthogonal tool/flange and robot/world calibration. The International Journal of Medical Robotics and Computer Assisted Surgery 8(4), 407–420

    2012

  6. [14]

    (2017) Densely connected convolutional networks

    Huang, G., Liu, Z., Van Der Maaten, L., Weinberger, K.Q. (2017) Densely connected convolutional networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 4700–4708

    2017

  7. [15]

    (2018) Squeeze-and-excitation networks

    Hu, J., Shen, L., Sun, G. (2018) Squeeze-and-excitation networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 7132–7141

    2018

  8. [16]

    (2017) Automatic differentiation in PyTorch

    Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L., Lerer, A. (2017) Automatic differentiation in PyTorch. In: NIPS Autodiff Workshop

    2017

  9. [17]

    (2015) A survey on multi-output regression

    Borchani, H., Varando, G., Bielza, C., Larra˜ naga, P. (2015) A survey on multi-output regression. Wiley Interdisci- plinary Reviews: Data Mining and Knowledge Discovery 5(5), 216–233

    2015

  10. [18]

    (2014) Learning and transferring mid-level image representations using convolutional neural networks

    Oquab, M., Bottou, L., Laptev, I., Sivic, J. (2014) Learning and transferring mid-level image representations using convolutional neural networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1717–1724

    2014

  11. [19]

    (2018) Learning to compare: Relation network for few-shot learning

    Sung, F., Yang, Y., Zhang, L., Xiang, T., Torr, P.H., Hospedales, T.M. (2018) Learning to compare: Relation network for few-shot learning. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1199–1208

    2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.