List of Papers Browse by Subject Areas Author List
Abstract
Osteosarcopenia, characterized by concurrent deterioration of bone and muscle, is increasingly recognized as a major contributor to frailty, fractures, and mortality, particularly in prostate cancer. However, quantitative longitudinal assessment of osteosarcopenia remains limited by existing manual methods based on 2D analysis at a single time point, which are impractical for large-scale monitoring and cannot detect subtle changes over time due to slice-selection variability.
This study introduces an AI-enabled pipeline to quantify musculoskeletal (MSK) changes from 3D serial CT scans in prostate cancer. The pipeline integrates deep learning–based vertebrae detection and 3D vertebral body and psoas muscle segmentation. Lumbar vertebral bone density, psoas muscle volume and density were retrospectively quantified from 5,966 CT scans acquired from a cohort of 916 metastatic prostate cancer patients (2008–2024).
The pipeline demonstrated strong agreement with a manually contoured test set for vertebral density (R^2=0.874-0.986), psoas density (R^2=0.955), and psoas volume (R^2=0.830). In repeated short-interval scans, the proposed 3D biomarkers showed higher reliability than conventional 2D measurements for psoas density (ICC(3,1)=0.86 vs. 0.72) and psoas volume versus cross-sectional area (ICC(3,1)=0.94 vs. 0.87). Longitudinal mixed-effects modeling showed that declining muscle and bone biomarkers were associated with increasing lumbar metastatic burden (p<0.05). The combined muscle model explained the greatest proportion of variance (marginal R^2=0.424).
These findings demonstrate that automated 3D CT-based biomarkers provide reliable and clinically meaningful measures of progressive MSK decline. This framework enables scalable opportunistic monitoring of osteosarcopenia and may support earlier identification of patients at risk for treatment-related deterioration, metastatic progression, and fracture.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/6617_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{TabSal_Quantification_MICCAI2026,
author = { Tabatabaei, Saleh AND Ross, Tayler D. AND Emmenegger, Urban AND Whyne, Cari M. AND Hardisty, Michael},
title = { { Quantification of 3D Musculoskeletal CT Image-Based Biomarkers in the Lumbar Spine: Application in Metastatic Prostate Cancer } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16885},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
The study proposed an automated pipeline to quantify changes in psoas muscle volume/density and vertebrae density by evaluating 3D image-based biomarkers through longitudinal opportunistic CT scans to demonstrate association with prostate cancer progression for patients suffering from osteosarcopenia.
- 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.
An automated pipeline was presented to quantify changes in psoas muscle properties and vertebrae density using longitudinal CT scans. The generated 3D biomarkers outperformed traditional 2D biomarkers/clinical/manual results, and showed association with prostate cancer progression.
- 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.
The method was well-described, but with some flaws. Details should be included to describe the CT protocol, in-plane resolution, slice thickness, mA, kVP, kernels, etc. , as these will all affect muscle and bone volume and density estimation. Three scanners were mentioned (GE, Toshiba, Siemens), but there was no information on how density was calibrated for the muscle or bone. This is a main weakness. Both muscle and bone densities were used as biomarkers, but insufficient evidence was provided to show that these numbers were reliably or accurately estimated. This information is also critical if another group wants to reproduce the study. In addition, please clarify how was CT scans at different time points integrated into the pipeline, or were they not (i. e. the longitudinal scans need to be fed into the pipeline separately)? The study compared 3D biomarkers with 2D and manual measurements, but are there plans to further validate these 3D biomarkers in a cohort with known fracture incidence to ascertain the predictive ability of the pipeline for real fractures in prostate cancer patients?
- 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 provide sufficient information for 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.
(3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
Further clarification is needed in terms of density calibration procedure and CT protocols. The paper lacked sufficient evidence to support its outcome without these crucial information.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
Author provided detailed and specific information in their response. The clinical need is clear to better risk stratify prostate cancer patients. There is also a clear future application to other types of cancer that impacts skeletal health.
Review #2
- Please describe the contribution of the paper
This paper presents a fully automated, deep learning-enabled pipeline for longitudinal quantification of musculoskeletal (MSK) biomarkers (lumbar vertebral bone density, psoas muscle volume, and psoas muscle density) from opportunistic 3D CT scans. The authors address a critical limitation in current MSK assessment: the high variability and low reliability of 2D single-slice measurements for tracking progressive conditions like osteosarcopenia. The pipeline utilizes a multi-stage approach including 3D ResNet-50/FPN for vertebra detection, a 3D U-Net for vertebral body segmentation, and the TotalSegmentator model for psoas muscle segmentation. Validated on a large longitudinal cohort of 906 metastatic prostate cancer patients (5,966 scans), the 3D biomarkers demonstrated superior reliability compared to traditional 2D metrics and a significant clinical association with metastatic burden
- 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.Clinical relevance and motivation
- The problem of longitudinal MSK monitoring in cancer patients is well motivated, particularly in the context of osteosarcopenia and treatment-related deterioration.
2.Large-scale dataset
- The study leverages a substantial dataset with longitudinal follow-up, which is a strong asset.
3.End-to-end automatic pipeline
- The integration of detection, segmentation, and physiologically guided cropping is well designed and addresses key limitations of prior 2D approaches.
4.Clear quantitative evaluation
- The use of ICC for reliability and mixed-effects models for clinical association is appropriate and convincingly demonstrates the benefits of 3D biomarkers.
5.Demonstrated improvement over 2D methods The gains in reliability (ICC improvements) are clearly reported and support the main claims.
- 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.Limited novelty in methodology
- While the pipeline is well engineered, most components (3D CNNs, nnU-Net/TotalSegmentator, vertebra detection) are based on existing methods. The novelty lies more in integration than in methodological innovation.
2.Insufficient technical details
- Key implementation aspects are missing or anonymized (model training details, hyperparameters, inference time), which limits reproducibility.
3.Evaluation scope
- The validation focuses primarily on reliability and correlation with metastatic burden. Additional evaluation (external datasets, robustness to acquisition variability, ablation studies) would strengthen the paper.
4.Ground truth limitation
- Manual annotations are limited (40–50 scans), which may not fully capture variability, especially given the dataset size.
5.Clinical interpretation
- While associations are shown, the clinical impact (predictive value, decision-making utility) is not deeply explored. The link with prosate cancer is not very clear and this method should be employed for all cancer related patients.
- 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 provide sufficient information for 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.
(3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper presents a robust and well-engineered automated pipeline that effectively integrates existing deep learning techniques for musculoskeletal biomarker quantification. While the pipeline demonstrates strong clinical utility and reliability, its technical contributions are primarily incremental rather than fundamentally novel. The innovation lies in the thoughtful combination and optimization of established methods (e.g., 3D segmentation, physiologically guided cropping) to address a specific clinical challenge—longitudinal tracking of musculoskeletal health in metastatic prostate cancer.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
The authors have satisfactorily addressed the concerns raised from the 3 reviewers during the review process. Overall, the rebuttal significantly strengthens the manuscript by improving clarity, contextualizing contributions, and addressing methodological concerns.
Review #3
- Please describe the contribution of the paper
The method follows a “vertebra localization followed by segmentation” pipeline, similar to (https://pmc.ncbi.nlm.nih.gov/articles/PMC8589948/ ) and uses TotalSegmentator CT. The measurement is standardized by cutting planes defined by the vertebral geometry.
This study builds on a prior medical paper by applying existing methods, validating the accuracy of the derived data, and subsequently extracting clinically relevant statistics.
The results are evaluated against manual delineations, with reliability assessed using the intraclass correlation coefficient (ICC). In addition, the approach is compared with slice-based and other 2D methods.
The main outcome is the analysis of the relationship between bone and muscle density, muscle volume, and the number of metastatic vertebrae.
- 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 provides an automated, clear, and simple pipeline for correlating known features with metastasis.
The paper uses appropriate methods to evaluate the medical claims.
- 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.
TotalSegmentator is ill-suited for muscle volume estimation, as it commonly undersegments muscle tissue and was primarily designed for signal estimation. While this strategy helps to reduce partial-volume effects, it can lead to inaccurate volume measurements.
In my experience, localization-based methods may fail to detect individual vertebrae. Because the cited method is blinded, I am unable to assess whether this issue was systematically analyzed and appropriately mitigated.
In addition, there is no discussion of how translation anomalies are identified and handled.
- 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
Please revise the figure so that important details are clearly visible without requiring excessive zooming. At its current resolution, the figure is difficult to inspect and even caused noticeable lag while scrolling on my laptop. A more balanced resolution and improved scaling of labels/details would greatly enhance readability.
Regarding the statement “deep learning model previously developed in our laboratory [Anonymized]”, this should be phrased more neutrally and accompanied by an appropriate citation. Even in blinded form, it should read as though the method originated from prior independent work that is being reimplemented or applied in the current study, rather than as an internal laboratory development.
- 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?
Looking at the reviewer guidelines, this paper must convincingly address the following question: “Is the paper of sufficiently high clinical impact to outweigh a lower degree of methodological innovation?” The pipeline presented so far appears straightforward and primarily composed of established processing steps. Since no qualitative visualizations are provided, it is difficult to judge how subtle, robust, or clinically meaningful the extracted changes actually are, even though such a pipeline may be a necessary technical step. The significance of this work, therefore, depends strongly on the medical application, and the authors should use their rebuttal to clearly explain which Key Features for CAI and Clinical Translation Papers (https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html ) are fulfilled and why these aspects are noteworthy rather than routine. In particular, the manuscript should more specifically discuss the challenges associated with metastatic disease and explain why these measurements are clinically important. Be specific and do not generally say that this is important! For example, whether they enable early prediction of metastasis, capture features that are difficult to observe reliably by humans, address a need explicitly raised in prior clinical studies for rigorous extraction of these metrics, or replace an existing, more expensive technique.
I currently lean toward a weak accept in order to give the authors an opportunity to strengthen this clinical argument during rebuttal, especially since MICCAI aims to encourage more translational submissions. I do not see major methodological issues at this stage. However, the authors must make a convincing case that this application addresses an important unmet need and is genuinely relevant to the medical community.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Reject
- [Post rebuttal] Please justify your final decision from above.
The authors only addressed the missing details.
The major weakness, missing clinical impact and utility beyond their blinded previous work, remained.
Author Feedback
Reviewer1: CT protocol details (resolution, slice thickness, kVp, etc) will be summarized in the revision. All imaging was acquired at the same institution on a limited number of scanners, with routine weekly phantom calibration. This is a study limitation and users of the proposed pipeline may need to normalize images or ensure uniform acquisition parameters. As the reviewer points out, differences in x-ray spectra associated with kVp may affect HU radiological densities(Nhila 2022), with smaller effects on volume measurements. We expect more modest changes due to slice thickness(1-5mm) because all measurements are made over larger volumes with resampling to a consistent discretization. Restricting analysis to the consistent 120kVp GE scanner protocol (84% of scans) did not alter study conclusions. In prior blinded work, these biomarkers were associated with fracture incidence in a younger(<55 years) subset of these patients. Future work will evaluate fracture risk prediction. Reviewer2: 1-This primarily integrates existing models with evaluation targeting clinical need and demonstrated utility. The main innovation is application to enable longitudinal musculoskeletal assessment in metastatic prostate cancer, where current clinical approaches(e.g. DEXA) lack sensitivity to monitor progression. This work is the first to show that 3D muscle and bone biomarkers were more reliable than 2D assessments. 2-For vertebral detection, a 3D ResNet-50+FPN backbone was trained on the VerSe dataset using AdamW (lr=1e-4, batch=4) for 1500 epochs on 4 NVIDIA Tesla V100 GPUs. Scans were resampled to 1.75mm³ spacing and cropped to 128×128×384 voxels. For vertebral body segmentation, a 6-layer 3D U-Net with 32 base filters was trained using AdamW(lr=1e-4) for 200 epochs on 530 training and 130 validation segmentations. Scans were resampled to 1mm³ spacing and cropped to 128×128×64 patches. 3-The study is limited to one dataset, but contains variability with multiple vendors and scanning protocols. Our recent work (blinded) evaluated this analysis protocol in radiotherapy treatment-planning CT scans, demonstrating robustness across acquisition settings. External validation on datasets from 2 partner institutions is ongoing. 4-Manually annotated scans were chosen to capture variability in scanner vendor, IV contrast, scan type, and patient age. Observed power showed sufficient QC sample size, with low SEM (tables 2, 3) relative to MAE and clinically meaningful differences. 5-The pipeline can be used for other cancer populations undergoing opportunistic CT imaging. Advanced prostate cancer patients undergo regular abdominal imaging, have long survival, and high prevalence of OSP related to both cancer and its treatment (ADT), making this an excellent population. Reviewer3: Our prior work showed that TotalSeg systematically undersegmented muscle volumes relative to clinician ground truth but with a consistent proportional bias. This factor was applied to correct muscle volume biomarker. Prior blinded work (detection model) addressed localization failures using a GCN classification branch, Euclidean distance constraints, and graph-based post-processing enforcing anatomically consistent vertebral ordering, improving robustness in ablation studies. In this study, the vertebral detection failure rate was 1.59% (473/29,830 vertebrae). Ensemble methods may further reduce errors. Quantitative evaluation of bone and muscle health enables early detection of OSP through subtle longitudinal changes. Opportunistic CT imaging may further obviate the need for existing clinical methods(e.g. DEXA) reducing cost. Study reliability testing demonstrated improved consistency as these changes are difficult to assess using conventional visual or 2D assessments. Early OSP detection can predict disparate patient outcomes, allowing earlier interventions to address OSP with metastatic prostate cancer, mitigating impacts on morbidity, mortality, and improving quality of life.
Meta-Review
Meta-review #1
- Your recommendation
Invite for Rebuttal
- 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.
The reviews for this submission were mixed. The authors are invited to submit a rebuttal addressing the primary concerns and major points raised by the reviewers
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
I am aligning with the two accept decisions. I believe that once the blinded information is disclosed in the camera-ready version, several of the current concerns will be adequately addressed. The clinical investigation, although still at an early stage, is sufficiently strong and represents an important contribution that is likely to stimulate valuable discussion during the meeting.
Meta-review #2
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
The reviewers showed some enthusiasm for this work being presented at the conference. Several concerns remain but the overall contributions are strong enough to justify acceptance.
Meta-review #3
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
Following the rebuttal, two reviewers were happy with the authors’ providing detailed CT protocol information, clarifying scanner variability and calibration considerations, describing implementation details, and strengthening the clinical motivation for monitoring osteosarcopenia in prostate cancer patients. However, one reviewer expressed concerns, arguing that the rebuttal mainly addressed missing technical details while the broader clinical impact and novelty beyond prior related work remained insufficiently demonstrated. Overall, while the methodological novelty is largely integrative rather than algorithmic, the rebuttal addressed several key reproducibility and robustness concerns, and the application has clear clinical translational value. I therefore recommend acceptance.