Decoding depression during subcallosal cingulate deep brain stimulation

Vineet Tiruvadi, Ashan Veerakumar, Otis Smart, Andrea Crowell, Patricio Riva‐Posse, Robert E. Gross, Cameron C. McIntyre, Christopher J. Rozell, Robert J. Butera, Helen D. MaybergbioRxiv 2022 · DOI

TL;DR

This paper introduces a novel method to decode depression severity from brain activity (local field potentials or LFPs) recorded during deep brain stimulation (DBS) of the subcallosal cingulate (SCC). The developed "Depression Readout from Subcallosal Cingulate Cortex" (DR-SCC) decoder can predict approximately 10% of depression symptom variance, offering a promising step towards objective, real-time monitoring of treatment response and adaptive DBS strategies.

What Problem Does This Paper Address?

Major Depressive Disorder (MDD) is a debilitating condition, and treatment-resistant depression (TRD) often requires invasive interventions like deep brain stimulation (DBS). While SCC-DBS has shown efficacy, assessing its effectiveness is challenging. Current methods rely heavily on subjective patient self-reports and clinical scales, which can be slow to capture changes and vary significantly between individuals. This makes it difficult to systematically study treatment mechanisms and optimize DBS parameters. The paper aims to bridge this gap by developing an objective, physiologically derived signal that can track depression severity over time, complementing existing subjective measures.

Methods

The researchers employed a multi-faceted approach:

  • Data Acquisition: Local Field Potentials (LFPs) were recorded multiple times daily from six patients undergoing SCC-DBS for TRD over a seven-month period using the Activa PC+S™ system.
  • Signal Preprocessing: A crucial step involved developing a pipeline to correct for "mismatch compression distortions" that arise from the stimulation itself, ensuring that the recorded LFPs accurately reflect brain activity rather than stimulation artifacts.
  • Oscillatory Analysis: Power spectral densities (PSDs) were calculated for different frequency bands (delta, theta, alpha, beta, gamma) to identify changes in brain activity over time.
  • Decoder Development: An elastic net regularized linear regression model was trained to predict depression symptom severity (normalized HDRS17 scores) from the processed LFP oscillatory data. This "Depression Readout from Subcallosal Cingulate Cortex" (DR-SCC) was trained on a portion of the data and tested on the remainder.
  • Clinical Simulation: The utility of the DR-SCC was evaluated by simulating its use in clinical decision-making tasks, such as predicting treatment response and guiding adjustments to DBS amplitude.

Datasets

The study utilized data from six patients with treatment-resistant depression who received subcallosal cingulate deep brain stimulation (SCC-DBS). LFP recordings were collected over seven months post-implantation. Depression severity was assessed using the Hamilton Depression Rating Scale (HDRS17), which was normalized to create nHDRS scores for inter-patient comparison. No new benchmark datasets were introduced; the study focused on analyzing existing clinical and physiological data.

Key Insights and Ideas

  • Stimulation Artifacts are a Major Hurdle: The paper highlights the significant challenge of separating true brain signals from stimulation artifacts in DBS recordings. Their developed "mismatch compression correction" pipeline was essential for obtaining meaningful data.
  • Asymmetric Brain Activity Matters: The DR-SCC decoder identified an asymmetry in the contribution of left and right SCC oscillations, with left-beta showing a positive correlation with depression severity and right-beta and right-delta showing negative correlations. This suggests that the brain's response to depression and treatment is not uniform across hemispheres.
  • Beyond Subjective Measures: The DR-SCC offers a glimpse into an objective, physiological marker for depression, moving beyond the limitations of self-report scales. This could revolutionize how we monitor treatment progress and personalize interventions.
  • A Step Towards Adaptive DBS: The simulated clinical scenarios suggest that the DR-SCC, while not perfect on its own, can augment existing clinical assessments to potentially inform more dynamic and responsive DBS therapies.

Key Results

  • Oscillatory Changes: Significant changes in oscillatory power were observed between early and late therapy, particularly in the beta band, even after accounting for stimulation artifacts.
  • Predictive Power: The DR-SCC decoder explained approximately 10% of the variance in depression symptom severity (nHDRS) over seven months, achieving statistical significance (R² ≈ 0.11).
  • Clinical Significance: In simulated clinical tasks, the DR-SCC demonstrated a statistically significant ability to predict treatment response (AU-ROC ≈ 0.7). When used in conjunction with nHDRS, it showed potential for improving predictions of stimulation parameter increases.

Limitations and Future Work

  • Small Sample Size: The study involved only six patients, limiting the generalizability of the findings. Future work should involve larger cohorts.
  • Limited Predictive Power: The DR-SCC currently explains only about 10% of the depression signal. This is attributed to the fact that depression is a network disorder, and SCC is only one node. Future research should explore incorporating data from other brain regions or using whole-brain models.
  • Technical Limitations: The Activa PC+S™ system has limitations in reliably measuring high-frequency gamma oscillations. Future studies with more advanced recording devices are needed.
  • Linear Model: The use of linear regression might not capture the full complexity of the relationship between brain activity and depression. Exploring non-linear models could be beneficial.
  • Future Directions: The authors suggest further validation of the DR-SCC in next-generation devices, exploring the role of oscillatory phase in addition to power, and developing more sophisticated models that can infer network-wide states. They also emphasize the need for larger datasets and cleaner recordings for more robust analysis.

Summary updated

Abstract

A bstract Deep brain stimulation of subcallosal cingulate white matter (SCCwm-DBS) alleviates symptoms of treatment resistant depression (TRD) over months of therapy. Readouts of depression symptom severity derived from neural recordings are needed for more systematic study and improvement of the therapy. In this study, we measured local field potentials (LFP) multiple times a day alongside seven months of therapy using the Activa PC+S™ in six patients treated with SCCwm-DBS. We found significant changes in oscillatory power between early and late therapy after accounting for stimulation-related distortions, particularly within the β band. We then used a decoder strategy to identify oscillatory activity that tracked with depression measurements over seven months, with asymmetric δ and β oscillations contributing to a statistically significant prediction of 10% of the measured depression signal. Simulating its use in clinical decision-making, we demonstrated that the DR-SCC yield clinically meaningful information that can augment other measures of depression state. Ultimately, this DR-SCC provides a data-driven first-step towards objectively tracking chronic recovery after antidepressant DBS implantation and developing adaptive DBS strategies in the presence of active stimulation.
BibTeX
@misc{tiruvadi2022decoding,
  author = {Vineet Tiruvadi and Ashan Veerakumar and Otis Smart and Andrea Crowell and Patricio Riva‐Posse and Robert E. Gross and Cameron C. McIntyre and Christopher J. Rozell and Robert J. Butera and Helen D. Mayberg},
  title = {{Decoding depression during subcallosal cingulate deep brain stimulation}},
  year = {2022},
  doi = {10.1101/2022.07.27.501778},
  howpublished = {bioRxiv},
}
Figure