Dorus Paper Report
A referee's read of any paper, in under a minute.
How rigorous is it, how new, can it be rebuilt, and does it overclaim? Every answer points to the sentence behind it.
Citations with intent
Normalized results
Code & data checks
Patents that cite it
Contactless atrial fibrillation screening with an under-mattress piezoelectric sensor: a 120-night home study
ShareExport PDF
Rigor
7.8/10
Top 18% of 412 papers on this task
Novelty
Incremental
+0.03 AUROC over the prior best · 3 close prior works
Reproducibility
Partial
Data on request
Claims
2 of 9 flagged
1 overclaim1 contradicted
Headline result against earlier resultsAF detection · home setting · AUROC
Earlier papers
Best so far
This paper
Closest prior workcitations, claims and patents
- 2024Ballistocardiographic AF detection in a sleep laboratory0.81
- 2023Home ballistocardiography for arrhythmia screening: a feasibility study0.74
- 2023Bed-based arrhythmia monitoring system (patent, granted)0.69
Rigor checks4 of 6 passed
- Reference standard
- Subject-independent evaluation
- Multi-site
- Confidence intervals
- Cohort size and balance
- Pre-registered analysis plan
Artifactslinks followed · checked 2026-11-14
- Code repositoryrepo
- Overnight BCG datasetdataset
- Trained model weightsmodel
- Sensor and reference devicehardware
Reporting checklist · STARD 2015diagnostic accuracy study
21 / 30 items
- Participant flow diagram
- Sample size rationale
- Indeterminate results
Claim"Our system matches polysomnography for detecting atrial fibrillation."
EvidenceAUROC 0.93 (95% CI 0.89–0.96) against Holter · Table 2
OverclaimAgreement is reported, not equivalence. No equivalence test.
Claim"The first contactless AF screening validated in the home."
Prior workHome ballistocardiography feasibility study, 2023 (cited by this paper)
Contradicted by prior workAn earlier home validation exists in the atlas.
Claim"Detection remains robust with a bed partner."
EvidenceSubgroup of 11 subjects, AUROC 0.90 · Table 4
Supported, small samplen = 11; no confidence interval for the subgroup.
Writing checks
- Claims checked
- 9 · 7 supported
- "Significant" without a test
- 1 · §5, para 2
- Strong words
- "robust" ×4, "novel" ×3
- Limitations section
- Present · 3 limitations
- Conflicts of interest
- 2 authors employed by the sensor maker
"…demonstrating robust performance across all conditions…"
§5 · flagged: "all conditions" covers 3 tested conditions.
Quality & novelty
Novelty measured against real prior work, not word overlap.
- Rigor checks, each with the sentence that triggered it
- Headline result against the prior best on the same task and metric
- Closest prior work, cited or not, across papers and patents
Uncited prior artfound through the patent layer
PatentBed-based arrhythmia monitoring system
Granted 2025 · priority 2023 · assigned to a sensor company
Granted 2025 · priority 2023 · assigned to a sensor company
OverlapClaims 1 and 4 cover under-mattress AF detection with a piezoelectric film, the paper's main contribution.
Not citedSimilarity 0.69 · higher than 2 of the 3 papers it does cite
Reproducibility
Can it be rebuilt? Checked, not assumed.
- Live code and data, with the date they were checked
- STARD for diagnostic accuracy, TRIPOD for prediction models
- Subject split and leakage read from the methods
Reproducibility across the field412 papers on this task
This paper: live code and subject-independent split; data on request; no weights.
Claims & writing
Every claim, next to the number behind it.
- Claims linked to tables, figures and quotes
- "First" claims checked against the whole atlas
- Conflicts of interest and limitations surfaced
Claim"Our system matches polysomnography for detecting atrial fibrillation."
EvidenceAUROC 0.93 (95% CI 0.89–0.96) against Holter · Table 2
OverclaimAgreement is reported, not equivalence.
Claim"Detects AF episodes as short as 30 seconds."
EvidenceSensitivity 0.71 for 30-second episodes · Table 3
SupportedStated with its sensitivity.
Built on the atlas
One paper, read in the context of everything around it.
Citationswhat it builds on, and why it cites each paper
Resultsevery result, next to earlier ones on the same task
Code & datawhether the code and data are really there
Patentspatents that cite it, or cover it without citing it
Peoplewho did what, and where they work
Who it's for
For anyone who has to trust a paper.
Journal editor
"Tell me in a minute whether this deserves a reviewer's afternoon."
R&D lead
"Before we build on a result, I want to know it will hold."
Technical diligence
"Is the science behind this company as strong as the deck says?"
More on the same atlas
Get a report on any paper.
In development, starting with sleep and cardiac sensing research. Tell us the papers you need read.