Subspace Learning with Interval-Censored Likelihoods for Dequantizing Percept PC LFP Snapshots
Abstract
Implanted neurostimulators that sense local field potentials now enable chronic electrophysiology based biomarker tracking in patients at home. The Medtronic Percept PC, the only commercially available sensing-enabled deep brain stimulation (DBS) device, stores spectral amplitudes as 16-bit integers at approximately 0.1 μV per bit (quantum q ≈ 0.11 μVp). At frequencies where the true amplitude spans only a few quantization levels, consecutive bins round to the same stored value. Standard spectral parameterization (FOOOF, fitting oscillations and one over f), which separates periodic peaks from the aperiodic 1/f activity, treats every value as exact and fits oscillatory peaks to these plateaus. Because these spectra feed clinical biomarker pipelines and spectral foundation models for symptom decoding, spurious peaks can corrupt downstream inference. Across 9,438 spectra from 14 hemispheres in 7 subcallosal cingulate DBS patients, 20.6% of peaks detected at [2, 45] Hz have no match in ground truth synthesized by quantizing clean in-clinic BrainSense recordings, while aggregate beta band power and the aperiodic exponent are preserved. We formalize dequantization as interval-censored subspace estimation and compare five classes of correction methods. Quantized probabilistic PCA is the only tested method that reduces the spurious rate (20.6% to 18.3%) while preserving true peak detection and keeping noise floor RMSE below q/√(12).
Related
- DisKO: Deep Koopman Learning in Distribution Space from Unpaired Snapshots
- Learning hidden-state dynamics from unpaired snapshots with lift-project SDEs
- The Geometry of Multitask Learning: Sharp Rates for Shared Subspace Estimation
- Optimal Personalized Subspace Learning for Multi-view Tensor Observations
- Spectral and Subspace Dynamics of LoRA in Supervised Fine-Tuning and Reinforcement Learning
BibTeX
@misc{karjagi2026subspace,
title = {{Subspace Learning with Interval-Censored Likelihoods for Dequantizing Percept PC LFP Snapshots}},
author = {Shreesh Karjagi and Elif Ceren Fitoz and Maryam Khalid and Tanya Nauvel and Parisa Sarikhani and Helen S. Mayberg and Christopher J. Rozell and Sankaraleengam Alagapan},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.22658},
eprint = {2609.22658},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.22658},
}
RIS
TY - UNPB
TI - Subspace Learning with Interval-Censored Likelihoods for Dequantizing Percept PC LFP Snapshots
AU - Shreesh Karjagi
AU - Elif Ceren Fitoz
AU - Maryam Khalid
AU - Tanya Nauvel
AU - Parisa Sarikhani
AU - Helen S. Mayberg
AU - Christopher J. Rozell
AU - Sankaraleengam Alagapan
PY - 2026
DA - 2026/09/19
PB - arXiv
DO - 10.48550/arxiv.2609.22658
UR - https://arxiv.org/abs/2609.22658
AB - Implanted neurostimulators that sense local field potentials now enable chronic electrophysiology based biomarker tracking in patients at home. The Medtronic Percept PC, the only commercially available sensing-enabled deep brain stimulation (DBS) device, stores spectral amplitudes as 16-bit integers at approximately 0.1 μV per bit (quantum q ≈ 0.11 μVp). At frequencies where the true amplitude spans only a few quantization levels, consecutive bins round to the same stored value. Standard spectral parameterization (FOOOF, fitting oscillations and one over f), which separates periodic peaks from the aperiodic 1/f activity, treats every value as exact and fits oscillatory peaks to these plateaus. Because these spectra feed clinical biomarker pipelines and spectral foundation models for symptom decoding, spurious peaks can corrupt downstream inference. Across 9,438 spectra from 14 hemispheres in 7 subcallosal cingulate DBS patients, 20.6% of peaks detected at [2, 45] Hz have no match in ground truth synthesized by quantizing clean in-clinic BrainSense recordings, while aggregate beta band power and the aperiodic exponent are preserved. We formalize dequantization as interval-censored subspace estimation and compare five classes of correction methods. Quantized probabilistic PCA is the only tested method that reduces the spurious rate (20.6% to 18.3%) while preserving true peak detection and keeping noise floor RMSE below q/√(12).
ER -