List of Papers Browse by Subject Areas Author List
Abstract
Drug-resistant epilepsy (DRE) is often treated with resective surgery, yet a central clinical challenge remains: given a planned (or completed) resection, what is the likelihood the patient will achieve seizure freedom? Answering this is difficult because the optimal surgical target—the epileptogenic zone (EZ), the brain tissue that generates seizures—is latent and must be inferred from biomarkers derived from intracranial EEG (iEEG). Biomarkers at both channel and network level have been developed and evaluated as independent sources of data stream and thus have shown limited promise in predicting surgical outcomes. We hypothesize that the channel and network level biomarkers are mechanistically connected, and introduce NetCorr, a platform that makes use of this underlying correlation to better predict seizure freedom. Towards this end, we first introduce a regression framework that predicts a channel’s smoothed high-frequency-oscillation with spike (spkHFO) rate from its participation properties in a ripple-band synchronization network, quantifying network-spkHFO coupling via R^2. Secondly, we model resection as a perturbation of the preserved channel coupling scores and construct a 2-dimensional patient signature that summarizes resection-induced shifts in coupling statistics. In a multi-institutional retrospective cohort of 159 pediatric-onset focal DRE patients, this 2-dimensional signature predicts post-surgical outcome with a cross-validated mean F1 score of 95%, significantly outperforming ratio-based baselines using SOZ resection (F1 = 78%) and spkHFO resection (F1 = 79%).
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/5888_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
N/A
Link to the Dataset(s)
N/A
BibTex
@InProceedings{MonTon_NetCorr_MICCAI2026,
author = { Monsoor, Tonmoy AND Kanai, Sotaro AND Zhang, Lina AND Daida, Atsuro AND Sinha, Prateik AND Nariai, Hiroki AND Roychowdhury, Vwani},
title = { { NetCorr: A Resection-Induced Network Perturbation Framework for Predicting Seizure Freedom in Drug-Resistant Epilepsy } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16887},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
This paper presents a framework to predict seizure freedom in DREs. It is clinically relevant, well-motivated and also effective.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
1.The paper targets on a clinically relevant challenge: how to predict the surgical outcome in DRES. 2.The framework is neat and effective. It first uses a regression framework to model synchronization network–spkHFO coupling based on iEEG data, where a stable, interpretable statistics could be predicted to compare across channels and patients, based on which network perturbation features could be obtained to predict surgical outcome. 3.The paper is easy to follow, although several details are missing.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.The signal length seems variable across patients. It is not clear how, and to what extent, signal length affects the calculation of R2 and the network perturbation features. Is there any instruction of choose IEEG length when applied this method clinically? 2.In section 2.1, it says artifact-free interictal epochs were used for analysis. How robust the method is to noise remain to be explored.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The framework is novel and effective. But noise robustness should be explored to strengthen its soundness.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #2
- Please describe the contribution of the paper
The paper introduces NetCorr, a framework that predicts epilepsy surgery outcomes by modeling surgery as a network perturbation and measuring how removing brain regions changes the relationship between network dynamics and pathological signals (spkHFOs), showing improved prediction of seizure freedom compared to existing methods.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
- The paper proposes a novel formulation that models surgery as a network perturbation, offering a principled way to link pre- and post-resection brain dynamics.
-
It introduces an original integration of channel-level (spkHFO) and network-level features, which are typically studied separately in prior work.
-
The approach demonstrates strong empirical performance, significantly outperforming standard baselines
- The evaluation is conducted on a relatively large multi-institutional cohort, supporting the robustness of the findings.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
- Several methodological choices, such as retaining only the top 10% of network edges and using specific network features (degree and centrality), appear heuristic
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
I am not an expert in this specific domain, so my evaluation is based primarily on the clarity of the paper and the evidence presented rather than deep domain-specific judgment. From my reading, the method appears to offer a substantial improvement over existing approaches, with a clear and intuitive formulation and strong reported performance. However, given my limited expertise and some concerns about validation and generalizability, I am not fully confident in assessing the broader impact and robustness of the method. For these reasons, I lean toward a weak accept, acknowledging both the promising results and my uncertainty in fully evaluating the work.
- Reviewer confidence
Not confident (1)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #3
- Please describe the contribution of the paper
This paper proposes NetCorr, a framework for predicting postsurgical seizure freedom in drug-resistant epilepsy by linking channel-level pathological activity and network-level interictal synchronization dynamics. The method first models per-channel spkHFO rate as a function of network participation features in ripple-band synchronization networks, yielding channel-wise coupling scores. It then treats surgery as a network perturbation, recomputes preserved-channel coupling after removing resected channels, and summarizes the induced shift with a compact two-dimensional patient-level signature. Using this perturbation feature, the paper predicts seizure-freedom outcome in a multi-institutional retrospective cohort and reports strong performance relative to ratio-based SOZ and spkHFO resection baselines.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
1.The paper addresses a clinically important problem: predicting seizure freedom after epilepsy surgery is highly relevant to presurgical planning and postoperative assessment. 2.The central hypothesis is scientifically interesting. Rather than treating channel-level biomarkers and network-level biomarkers as unrelated signals, the paper proposes a mechanistic relationship between them and uses this relationship to define a perturbation-based outcome signature. 3.The method is conceptually coherent and relatively interpretable. The preserved-channel pre/post perturbation idea is intuitive and offers a clinically meaningful way to think about whether surgery disrupts the network organization underlying epileptiform activity. 4.The paper uses interictal data, which is practically valuable because it avoids relying solely on sparse ictal segments. 5.The patient-level representation is compact and elegant. Reducing the effect of resection to a two-dimensional feature summarizing changes in preserved-channel coupling is an appealing and understandable design choice. 6.The reported empirical performance is strong, and the comparison to spkHFO- and SOZ-based baselines supports the claim that the proposed perturbation feature captures more informative structure. 7.The discussion section is thoughtful and goes beyond simple performance reporting by considering translational implications such as in-silico resection-margin planning and signatures of undersampling.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.The paper’s framing should more carefully distinguish what is novel from what builds on established prior work. Interictal channel-level biomarkers such as HFO/spike-ripple–based outcome prediction [7,9,11,16,19,24] and network-level epilepsy modeling such as source-sink and fragility approaches [10,14,17] are already well developed. The present novelty lies in linking these two views through a perturbation-based coupling framework, rather than in introducing either family of biomarkers from scratch. 2.The reported performance is very strong for a compact 2D representation, which is exciting but also raises questions about robustness and possible optimism in evaluation. The manuscript would benefit from more detail on how design choices (e.g., smoothing kernel, sparsification threshold, k in kNN, and any model selection) were tuned and whether all tuning was strictly nested within cross-validation. 3.The baseline set is relatively narrow. The main comparisons are to ratio-based SOZ resection and spkHFO resection features, but the paper would be stronger if it also compared against stronger multivariable surgical outcome models or richer predictive baselines that reflect the broader outcome-prediction literature [12]. 4.The mechanistic interpretation is plausible but still inferential. The observation that preserved-channel coupling remains high in NSF patients is consistent with incomplete disruption of epileptogenic network structure, but this remains correlational evidence rather than direct proof of causal mechanism. 5.The framework depends on several upstream processing steps—artifact selection, spkHFO detection, network construction, and coupling regression—and the paper does not yet fully quantify how sensitive the final outcome prediction is to variability in these steps. 6.The paper would benefit from additional stratified analysis, for example by center, implant type, pathology, or seizure outcome subtype, to better understand where the method is strongest or weakest.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
I found this paper interesting and clinically relevant. The central idea—viewing surgery as a perturbation to the residual network and quantifying how that perturbation alters the relationship between synchronization structure and pathological event dynamics—is both intuitive and potentially impactful. This is a good example of a method that is not only predictive, but also offers a mechanistically meaningful lens for thinking about surgical sufficiency. I particularly appreciated that the paper tries to bridge two lines of work that are often treated separately: local event-based biomarkers such as spkHFOs and network-level interictal structure. The resulting 2D representation is elegant and easy to interpret, which is a real strength. My main suggestions are intended to strengthen confidence in the very strong reported performance. First, I would encourage the authors to provide more detail on tuning and validation choices, to rule out any concern about optimistic evaluation. Second, the comparison set could be broadened to include stronger outcome-prediction baselines beyond ratio-style features. Third, more sensitivity analysis around upstream choices (e.g., detection pipeline, smoothing, network sparsification) would help establish robustness. Finally, the translational discussion is promising, and I would welcome even more explicit analysis of how the proposed perturbation signature might support practical surgical planning. Overall, I think this is a thoughtful and promising contribution.
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
My recommendation is based on the paper’s clinically important problem, coherent mechanistic hypothesis, and strong empirical performance, balanced against some remaining concerns about validation depth and baseline breadth. I think the main contribution is meaningful: the paper does not merely introduce another predictive biomarker, but rather proposes a way to connect channel-level pathological activity and network-level synchronization through a resection-induced perturbation perspective. This is conceptually appealing and clinically interpretable. The use of interictal data is also practical, and the translational implications discussed by the authors are compelling. My main reservations concern how strongly the current evidence supports the reported level of performance and generalization. The baseline comparisons are somewhat limited, the cohort remains retrospective and relatively specific, and additional robustness or sensitivity analyses would help increase confidence in the method. Nonetheless, I believe the paper is above the acceptance threshold because it contributes a clear new perspective, reports promising results, and is stronger than many purely incremental prediction papers. For these reasons, I place it slightly above the acceptance threshold.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Author Feedback
We thank the meta-reviewer and reviewers for their thoughtful and constructive assessment, and for recognizing the clinical relevance, interpretability, and translational potential of NetCorr. We are encouraged that the reviewers found the central idea useful: modeling resection as a perturbation to the residual epileptogenic network and linking channel-level spkHFO activity with network-level synchronization dynamics. We agree that the novelty of the work should be framed precisely. In the camera-ready version, we will clarify that NetCorr does not claim to introduce spkHFO biomarkers or interictal network biomarkers independently. Rather, the main contribution is to connect these two established biomarker families through a regression-based network–spkHFO coupling score (R²), and then quantify how this coupling changes in preserved tissue after resection. We will revise the Introduction and Discussion to make this distinction clearer with respect to prior HFO/spike-ripple, source–sink, neural-fragility, and interictal synchronization-network biomarker approaches. We also appreciate the comments regarding validation, robustness, and heuristic design choices. In the revised manuscript, we will clarify the validation protocol, including patient-level cross-validation, training-fold-only normalization, and the fact that held-out patients were used only for out-of-fold evaluation. We will also better motivate the main design choices. Specifically, the top-10% adaptive sparsification was used to retain the strongest synchronization relationships while normalizing graph density across windows and patients, and the selected network features capture complementary aspects of local incoming synchronization, outgoing synchronization, and global network participation. Regarding variable recording duration and noise robustness, we will clarify that coupling is estimated from within-patient 1-second windows and summarized using patient-level preserved-channel distributional changes, rather than raw event counts alone. We will expand the limitations section to explicitly discuss sensitivity to recording duration, sleep and medication state, and upstream detection and sparsification choices. We will also temper mechanistic language to emphasize that persistent preserved-channel coupling is consistent with residual epileptogenic network organization, but remains correlational rather than direct causal proof. Finally, we agree that broader baselines are an important future direction. We will explicitly acknowledge that future work should include stronger single-modality baselines — including data-driven pathological HFO-only and interictal synchronization-network-only approaches — to further isolate the added value of NetCorr’s central hypothesis: coupling channel-level pathological activity with network-level synchronization dynamics and measuring how this coupling is disrupted by resection.
Meta-Review
Meta-review #1
- Your recommendation
Provisional Accept
- Please justify your decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
All of the reviewers found this manuscript a useful, relevant contribution to MICCAI with some weaknesses in the evaluation of the model. This manuscript has a clear clinical application and translation. Open questions around how robust the model is to inclusion of noise, heuristics used in the model can be addressed in the revised manuscript.