Medical AI safety researcher focused on memorisation bias, privacy, and longitudinal angiographic imaging methods
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-20) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Not yet scored: publication counts, venue data and authorship roles are insufficient in this snapshot. This axis would measure cumulative output and where and how often this researcher publishes.
Impact—
Not yet scored: citation counts, h-index and citing-context data are unavailable here. This axis would measure how often and how influentially the wider literature cites this researcher's work.
Focus100/100
Scores 100, reflecting complete concentration in the available data: the only field-tagged catalog paper sits in machine learning for healthcare (medical AI safety), covering memorisation, differential privacy and membership inference; the other 2026 preprint, on angiographic graph matching, is clinically adjacent. A deliberately narrow, safety-centred agenda.
Collaboration48/100
Scores 48 from 10 distinct coauthors in the snapshot. Lux appears in middle authorship positions without corresponding status on both catalog papers, consistent with a contributor role inside a dense, senior-led medical imaging AI network that includes Daniel Rueckert, Ben Glocker, Georgios Kaissis, Martin J. Menten and Mélanie Roschewitz.
Activity—
Not yet scored: publication cadence and yearly volume data are insufficient in this snapshot. This axis would measure research pace over time, including papers per year and continuity of output across the career record.
Reproducibility100/100
Scores 100 from one evaluated paper (reproMean 100 across reproPapers 1). The memorisation bias study is tagged high reproducibility readiness by Evolve, consistent with its large-scale, privacy-tool-anchored evaluation on ECG and chest radiograph benchmarks. A single-paper basis means this score should be read as promising, not yet settled.
Research summary
Laurin Lux is an early-stage researcher in machine learning for healthcare whose visible record, resolved from public bibliographic data and ORCID 0009-0003-7359-6212, consists of two 2026 arXiv preprints. The work spans two complementary directions: methods for temporally consistent graph extraction and matching in longitudinal angiographic images, and a safety study of memorisation bias in medical AI. The latter carries Evolve's field tag 'machine learning for healthcare (medical AI safety)' and a reviewed verdict: training on patients' past records biases AI predictions on their future data, backed by rigorous large-scale evidence with modest assessed clinical harm. Its toolkit draws on differential privacy and membership inference, evaluated through ECG deep learning and chest radiograph classification, and Evolve rates its reproducibility readiness as high. Radar scores mirror this picture: focus 100 from a single tagged field, reproducibility 100 from one evaluated paper, and collaboration 48 across 10 distinct coauthors. Output, impact, and activity remain unscored for lack of data. The co-author list — Daniel Rueckert, Ben Glocker, Georgios Kaissis, Martin J. Menten, Mélanie Roschewitz, and others — sits inside an established medical imaging AI community, though no institution or country is confirmed in the available data.
Collaboration pattern
Collaboration data show 10 distinct coauthors and a radar collaboration score of 48. Lux appears as a middle author without corresponding status on both catalog preprints, consistent with contributing inside a senior-led medical imaging AI group. The most frequent partners named in the data are Daniel Rueckert (listed twice), Martin J. Menten, Ben Glocker, Georgios Kaissis, Mélanie Roschewitz, Emma A. M. Stanley, Moritz A. Knolle, Linus Kreitner, and Carmen Baumann.
Research keywords
Highlights
Large-scale evidence of memorisation bias in medical AI
Temporally consistent graph methods for longitudinal angiography
This 2026 arXiv preprint addresses extraction and matching of graph structures across time points in longitudinal angiographic images, targeting temporal consistency in follow-up imaging analysis. Evolve has not yet issued a verdict, and no field tag or reproducibility assessment is available for this paper.
Timeline
- 2026Co-authored 'Temporally Consistent Graph Extraction and Matching for Longitudinal Angiographic Images' (arXiv preprint), contributing methods for temporal consistency in longitudinal vascular imaging analysis.
- 2026Co-authored 'Memorisation bias in medical AI' (arXiv preprint), field-tagged machine learning for healthcare (medical AI safety) and assessed by Evolve as rigorous large-scale evidence with high reproducibility readiness.
- 2026Both catalog papers appear in the same year, establishing an early 2026 footprint in medical imaging AI and AI safety alongside 10 distinct coauthors.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric dimensions are derived from OpenAlex-style public bibliographic data only; output, impact and activity axes lack sufficient data and are not scored.
- Institution, country and career details are not available in the input data and are deliberately omitted.
- Both catalog papers are 2026 arXiv preprints with middle authorship and no corresponding role; peer-review status is unknown.
- The reproducibility score rests on a single evaluated paper, and the focus score reflects one field-tagged paper only.