SA-Profile: Automated Sulcus Angle Profiling from Super-Resolution MRI
Michael Wehrli, Leo Widmer, Edwin Li, Noel Fiechter, Lorenzo Pettinari, Sidaty El Hadramy, Carol C. Hasler, Philippe C. Cattin
cs.CV · cs.AI
Quick verdict
The paper presents continuous sulcus-angle profiling from super-resolved MRI; a feasible but exploratory method: 11.6° MAE, no baselines, tiny cohorts.
Summary
The paper addresses the slice-selection sensitivity of the sulcus angle (SA), a measure of trochlear dysplasia conventionally read from a single axial MRI slice about 3 cm above the joint line. SA-Profile is a three-stage pipeline: an implicit neural representation fuses axial, coronal, and sagittal clinical acquisitions into an isotropic 256³ super-resolved volume; two 2D U-Nets localize the sagittal bounds (S1, S2) and axial trochlear landmarks (T1–T3) as heatmaps; maximum-likelihood spline fitting through the heatmaps yields a continuous SA profile along the trochlea. Models were trained on 159 labeled fastMRI/in-house volumes (111/16/32 patient-level split) and compared with single-rater manual SA on 25 fastMRI and 25 in-house TD patients. Overall mean absolute error was 11.6° (14.2° fastMRI, 8.9° in-house TD), with automated means of 139.8° versus manual 150.9° on fastMRI, suggesting systematic underestimation; landmark errors were 2.59–4.74 mm, above the sub-millimeter figures of prior single-slice work. Population profiles over 1,566 fastMRI cases and 32 TD patients differed visually around 40–80% of normalized position, but without statistical testing. Main limitations: small pathological cohort, single-rater reference, no comparison with existing automated SA methods, and unquantified super-resolution smoothing effects; the authors themselves frame the method as exploratory decision support.
Review complete
Assessment at a glance
4
risk flags identified
● high 0
● medium 3
● low 1
4 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- An INR-based super-resolution reconstruction fusing axial, coronal, and sagittal clinical knee MRI enables continuous sulcus angle measurement across the trochlea instead of at a single manually chosen axial slice.
- The automated profile agrees with single-rater manual SA at the corresponding slice with an overall MAE of 11.6° (14.2° fastMRI, 8.9° in-house TD), interpreted as agreement with current practice rather than accuracy against ground truth.
- Maximum-likelihood spline fitting through full landmark heatmaps enforces anatomically continuous landmark trajectories and reduces slice-to-slice variability.
- Population-level mean SA profiles of 1,566 fastMRI cases and 32 TD patients are distinct over roughly 40–80% of normalized trochlear position, suggesting profile-based assessment can characterize TD.
Strengths
Well-motivated formulation shift from single-slice to continuous profiling
The Introduction grounds the problem in cited evidence: slice choice alters trochlear assessment (Tscholl et al. [14]), observer variability of several degrees [10,17], and 18° mean cartilage-vs-bone SA differences [15]; the profile output directly answers these documented failure modes rather than a hypothetical one.
Principled use of full landmark heatmaps via MLE spline fitting
Section 2.3 fits a 3D spline (x(z), y(z), z) by maximizing summed heatmap log-likelihood (Eq. 1, L-BFGS-B, adaptive control points), enforcing proximal-distal continuity instead of independent per-slice argmax — the paper's most distinctive technical contribution.
Honest interpretive framing throughout
Feasibility shown at scale with released code
Weaknesses
Large MAE with an unexamined systematic bias on the public cohort
No baselines or component ablation
Population-level claim is untested
Landmark accuracy regresses vs prior art
Single-rater reference chain with no reliability data
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
SA-Profile targets a real, documented fragility: the sulcus angle (SA) is read manually from one axial slice "approximately 3 cm above the joint line" with no unique slice definition; slice selection changes trochlear assessment (Tscholl et al. [14]), observer variability reaches several degrees [10,17], and cartilage vs. subchondral bone references shift mean SA by up to 18° [15]. The paper claims: (1) a fully automatic pipeline — implicit-neural-representation (INR) super-resolution fusing axial/coronal/sagittal acquisitions (Sec. 2.1), two 2D U-Net landmark detectors (Sec. 2.2), and maximum-likelihood spline fitting (Sec. 2.3) — producing a continuous SA profile along the trochlea; (2) agreement with single-rater manual SA of MAE 11.6°; (3) population-level profile separation between 1,566 filtered fastMRI cases and 32 in-house TD patients (Fig. 5). The authors deserve credit for restraint: fastMRI is explicitly not a healthy reference, the MAE is framed as "agreement with current practice rather than... a definitive ground truth," and the method is labeled "exploratory decision-support."
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 4 strengths & 5 weaknesses, analyzed in detail
- 4 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What is the sulcus angle and why does it matter for trochlear dysplasia?
The sulcus angle measures the opening angle between the medial and lateral facets of the femoral trochlear groove on an axial knee MRI slice. Values above 145° are commonly associated with trochlear dysplasia, a malformation linked to anterior knee pain and patellar instability. Conventionally it is measured manually on one slice about 3 cm above the joint line, which makes the result sensitive to slice choice and landmark placement.
How accurate is the automated sulcus angle measurement in SA-Profile?
Compared with single-rater manual measurements at the same slice location, the automated method achieved a mean absolute error of 11.6° overall — 14.2° on 25 fastMRI cases and 8.9° on 25 in-house trochlear dysplasia patients. The authors interpret this as agreement with routine clinical practice, noting that manual sulcus angle itself varies by several degrees (up to about 9°) between observers.
Is the SA-Profile code or data publicly available?
The code is publicly available on GitHub (github.com/wehrlimi/SA_Profile). Experiments use the public fastMRI knee dataset (10,012 patients, filtered to 1,566 with the required axial sequences) plus a small private in-house cohort of 32 trochlear dysplasia patients, which is not publicly available.
Does SA-Profile replace the standard single-slice sulcus angle measurement?
No. The authors position it as an exploratory decision-support tool that extends the single-slice measurement into a continuous profile along the trochlea, remaining conceptually linked to the conventional measurement. They explicitly call for validation in larger clinically phenotyped cohorts and against annotations from multiple expert raters before clinical use.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figure panels (Figs. 1–5) survive in the extraction only as scattered axis labels and legend text; visual properties such as the separation and overlap of the ±1 SD bands in Fig. 5 could not be verified from the text.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.