Maximilian Lindholz

Maximilian Lindholz

Auto-generated profile, based on publicly available information

Maximilian Lindholz

Department of Radiology, Charité Universitätsmedizin Berlin, Berlin, Germany

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 November 2022 in Critical Care
Verified
Journal Article
Open access
PDF file
MLMaximilian Lindholz
Maximilian Lindholz
CSClara M. Schellenberg
Clara M. Schellenberg
JGJulius J. Grunow
Julius J. Grunow
SKSimone Kagerbauer
Simone Kagerbauer
AMAnnette Milnik
Annette Milnik
SSStefan J Schaller
Stefan J Schaller
Background: Mobilisation and exercise intervention in general are safe and feasible in critically ill patients. For patients requiring catecholamines, however, doses of norepinephrine safe for mobilisation in the intensive care unit (ICU) are not defined. This study aimed to describe mobilisation practice in our hospital and identify doses of norepinephrine that allowed a safe mobilisation. Methods: We conducted a retrospective single-centre cohort study of 16 ICUs at a university hospital in Germany with patients admitted between March 2018 and November 2021. Data were collected from our patient data management system. We analysed the effect of norepinephrine on level (ICU Mobility Scale) and frequency (units per day) of mobilisation, early mobilisation (within 72 h of ICU admission), mortality, and rate of adverse events. Data were extracted from free-text mobilisation entries using supervised machine learning (support vector machine). Statistical analyses were done using (generalised) linear (mixed-effect) models, as well as chi-square tests and ANOVAs. Results: A total of 12,462 patients were analysed in this study. They received a total of 59,415 mobilisation units. Of these patients, 842 (6.8%) received mobilisation under continuous norepinephrine administration. Norepinephrine administration was negatively associated with the frequency of mobilisation (adjusted difference -0.07 mobilisations per day; 95% CI − 0.09, − 0.05; p ≤ 0.001) and early mobilisation (adjusted OR 0.83; 95% CI 0.76, 0.90; p ≤ 0.001), while a higher norepinephrine dose corresponded to a lower chance to be mobilised out-of-bed (adjusted OR 0.01; 95% CI 0.00, 0.04; p ≤ 0.001). Mobilisation with norepinephrine did not significantly affect mortality (p > 0.1). Higher compared to lower doses of norepinephrine did not lead to a significant increase in adverse events in our practice (p > 0.1). We identified that mobilisation was safe with up to 0.20 µg/kg/min norepinephrine for out-of-bed (IMS ≥ 2) and 0.33 µg/kg/min for in-bed (IMS 0–1) mobilisation. Conclusions: Mobilisation with norepinephrine can be done safely when considering the status of the patient and safety guidelines. We demonstrated that safe mobilisation was possible with norepinephrine doses up to 0.20 µg/kg/min for out-of-bed (IMS ≥ 2) and 0.33 µg/kg/min for in-bed (IMS 0–1) mobilisation.

15 July 2025 in European Journal of Radiology
Verified
Journal Article
Open access
MLMaximilian Lindholz
Maximilian Lindholz
ABAlina Burdenski
Alina Burdenski
RRRichard Ruppel
Richard Ruppel
SSSchulze-Weddige Schulze-Weddige
Schulze-Weddige Schulze-Weddige
Georg Lukas BaumgärtnerGeorg Lukas Baumgärtner
Georg Lukas Baumgärtner
TPTobias Penzkofer
Tobias Penzkofer
Radiology reports can change during workflows, especially when residents draft preliminary versions that attending physicians finalize. We explored how large language models (LLMs) and embedding techniques can categorize these changes into textual, semantic, or clinically actionable types. We evaluated 400 adult CT reports drafted by residents against finalized versions by attending physicians. Changes were rated on a five-point scale from no changes to critical ones. We examined open-source LLMs alongside traditional metrics like normalized word differences, Levenshtein and Jaccard similarity, and text embedding similarity. Model performance was assessed using quadratic weighted Cohen's kappa (κ), (balanced) accuracy, F1, precision, and recall. Inter-rater reliability among evaluators was excellent (κ = 0.990). Of the reports analyzed, 1.3 % contained critical changes. The tested methods showed significant performance differences (P < 0.001). The Qwen3-235B-A22B model using a zero-shot prompt, most closely aligned with human assessments of changes in clinical reports, achieving a κ of 0.822 (SD 0.031). The best conventional metric, word difference, had a κ of 0.732 (SD 0.048), the difference between the two showed statistical significance in unadjusted post-hoc tests (P = 0.038) but lost significance after adjusting for multiple testing (P = 0.064). Embedding models underperformed compared to LLMs and classical methods, showing statistical significance in most cases. Large language models like Qwen3-235B-A22B demonstrated moderate to strong alignment with expert evaluations of the clinical significance of changes in radiology reports. LLMs outperformed embedding methods and traditional string and word approaches, achieving statistical significance in most instances. This demonstrates their potential as tools to support peer review.

14 November 2023 in Annals of Intensive Care
Verified
Journal Article
Open access
PDF file
PKPhilipp Kloss
Philipp Kloss
MLMaximilian Lindholz
Maximilian Lindholz
AMAnnette Milnik
Annette Milnik
EAElie Azoulay
Elie Azoulay
MCMaurizio Cecconi
Maurizio Cecconi
SSStefan J Schaller
Stefan J Schaller
Early mobilisation (EM) is an intervention that may improve the outcome of critically ill patients. There is limited data on EM in COVID-19 patients and its use during the first pandemic wave. This is a pre-planned subanalysis of the ESICM UNITE-COVID, an international multicenter observational study involving critically ill COVID-19 patients in the ICU between February 15th and May 15th, 2020. We analysed variables associated with the initiation of EM (within 72 h of ICU admission) and explored the impact of EM on mortality, ICU and hospital length of stay, as well as discharge location. Statistical analyses were done using (generalised) linear mixed-effect models and ANOVAs. Mobilisation data from 4190 patients from 280 ICUs in 45 countries were analysed. 1114 (26.6%) of these patients received mobilisation within 72 h after ICU admission; 3076 (73.4%) did not. In our analysis of factors associated with EM, mechanical ventilation at admission (OR 0.29; 95% CI 0.25, 0.35; p = 0.001), higher age (OR 0.99; 95% CI 0.98, 1.00; p ≤ 0.001), pre-existing asthma (OR 0.84; 95% CI 0.73, 0.98; p = 0.028), and pre-existing kidney disease (OR 0.84; 95% CI 0.71, 0.99; p = 0.036) were negatively associated with the initiation of EM. EM was associated with a higher chance of being discharged home (OR 1.31; 95% CI 1.08, 1.58; p = 0.007) but was not associated with length of stay in ICU (adj. difference 0.91 days; 95% CI − 0.47, 1.37, p = 0.34) and hospital (adj. difference 1.4 days; 95% CI − 0.62, 2.35, p = 0.24) or mortality (OR 0.88; 95% CI 0.7, 1.09, p = 0.24) when adjusted for covariates. Our findings demonstrate that a quarter of COVID-19 patients received EM. There was no association found between EM in COVID-19 patients' ICU and hospital length of stay or mortality. However, EM in COVID-19 patients was associated with increased odds of being discharged home rather than to a care facility. Trial registration ClinicalTrials.gov: NCT04836065 (retrospectively registered April 8th 2021).

5 January 2025 in Radiography
Verified
Journal Article
Open access
MLM. Lindholz
M. Lindholz
RRR. Ruppel
R. Ruppel
SSSchulze-Weddige Schulze-Weddige
Schulze-Weddige Schulze-Weddige
Georg Lukas BaumgärtnerGeorg Lukas Baumgärtner
Georg Lukas Baumgärtner
ISIsabel Schobert
Isabel Schobert
TPT. Penzkofer
T. Penzkofer
Facial recognition technology in medical imaging, particularly with head scans, poses privacy risks due to identifiable facial features. This study evaluates the use of facial recognition software in identifying facial features from head CT scans and explores a defacing pipeline using TotalSegmentator to reduce re-identification risks while preserving data integrity for research. 1404 high-quality renderings from the UCLH EIT Stroke dataset, both with and without defacing were analysed. The performance of defacing with the face mask created by TotalSegmentator was compared to a state-of-the-art CT defacing algorithm. Face detection was performed using deep learning models. The cosine similarity between facial embeddings for intra- and inter-patient images was compared. A Support Vector Machine was trained on cosine similarity values to assess defacing performance, determining if two renderings came from the same patient. This analysis was conducted on defaced and non-defaced images using 5-fold cross-validation. Faces were detected in 76.5 % of non-defaced images. Intra-patient images exhibited a median cosine similarity of 0.65 (IQR: 0.47-0.80), compared to 0.50 (IQR: 0.39-0.62) for inter-patient images. A binary classifier performed moderately on non-defaced images, achieving a ROC-AUC of 0.69 (SD = 0.01) and an accuracy of 0.65 (SD = 0.01) in distinguishing whether a scan belonged to the same or a different individual. Following defacing, performance declined markedly. Defacing with the TotalSegmentator decreased the ROC-AUC to 0.55 (SD = 0.02) and the accuracy to 0.56 (SD = 0.01), whereas the CTA-DEFACE algorithm brought the performance down to a ROC-AUC of 0.60 (SD = 0.02) and an accuracy of 0.59 (SD = 0.01). These results demonstrate the effectiveness of defacing algorithms in mitigating re-identification risks, with the TotalSegmentator providing slightly superior privacy protection. Facial recognition software can identify facial features from partial and complete head CT scan renderings. However, using the TotalSegmentator to deface images reduces re-identification risks to a near-chance level. We offer code to implement this privacy-preserving pipeline. Utilizing the TotalSegmentator framework, the proposed pipeline efficiently removes facial features from CT images, making it ideal for multi-site research and data sharing. It is a useful tool for radiographers and radiologists who must comply with medico-legal requirements necessitating the removal of facial features.

8 February 2025 in RMD Open
Verified
Journal Article
Open access
CACarolina Dominguez Aleixo
Carolina Dominguez Aleixo
KZKatharina Ziegeler
Katharina Ziegeler
SUSevtap Tugce Ulas
Sevtap Tugce Ulas
Torsten DiekhoffTorsten Diekhoff
Torsten Diekhoff
JGJuliane Greese
Juliane Greese
FPFabian Proft
Fabian Proft
Objectives: Anatomical variation of the sacroiliac (SI) joints is common and specific variants are associated with erosions and bone marrow oedema on imaging. Our investigation aims to evaluate whether anatomical variations influence the clinical presentation of axial spondyloarthritis (axSpA). Methods: In this propensity score matched post hoc analysis documented clinical data from four prospective clinical cohorts was assessed. Classification of back pain as inflammatory (=IBP), human leucocyte antigen-B27 positivity, family history, disease activity according to Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), symptom duration, elevated acute phase reactants, peripheral and extramusculoskeletal manifestations were evaluated. Statistical analyses were done using (generalised) linear models, t-tests, χ2 tests and analysis of variances. Multiple testing was corrected according to Bonferroni. Results: A total of 165 patients (86 women) were included. Atypical SI joints, defined by the presence of accessory joint facets, iliosacral complex or crescent-shaped ilii on MRI, were identified in 61 out of 165 patients with axSpA. Disease activity, assessed by BASDAI and symptom duration were similar in both groups (adjusted ß=−0.118 (95% CI -0.713, 0.476), p=0.696 and 120.0 (107.4) vs 116.5 (98.3) months, p=0.838, respectively). There was no significant difference in IBP between the groups (adjusted OR=0.614 (95% CI 0.274, 1.377), p=0.236). Sex-stratified analysis revealed no statistically significant results. Conclusion: Our analysis suggests that clinical phenotypes do not significantly differ between patients with axSpA with and without atypical joints.

26 June 2026 in JMIR AI
Verified
Journal Article
Open access
SESemil Eminovic
Semil Eminovic
RSRobin Schmidt
Robin Schmidt
BLBogdan Levita
Bogdan Levita
MLMaximilian Lindholz
Maximilian Lindholz
AHAnna-Maria Haack
Anna-Maria Haack
TPTobias Penzkofer
Tobias Penzkofer
Background: Radiology trainees require efficient, accurate, and accessible resources to master complex imaging techniques and identify findings that guide clinical decision-making. Large language models (LLMs) are emerging as promising tools for medical education and clinical workflows, offering the potential to enhance learning by providing instant feedback, aiding in diagnostic accuracy, and offering personalized learning experiences. However, systematic comparisons of LLMs for radiology education and clinical support remain limited, particularly regarding differences across subspecialties and resident experience levels. Objective: This study aimed to evaluate and compare the response quality of 2 state-of-the-art reasoning-based LLMs, namely DeepSeek-R1 and ChatGPT-o1, as clinical and radiology residency support tools, comparing performance across clinical and didactic dimensions, including text- and image-based responses. Methods: Overall, 27 radiology questions covering 9 radiological subspecialties were answered by both LLMs. Additionally, 6 image-based questions were presented only to ChatGPT-o1 due to its image processing capabilities. Responses were independently rated by 7 radiology residents (postgraduate years 2‐5) across 9 rating criteria grouped into 3 dimensions (factual accuracy, clinical practicality, and didactic value), using a 5-point Likert scale. Statistics compared LLMs, reader experience, and response types for text- as well as image-based for ChatGPT-o1 queries. Results: DeepSeek-R1 consistently outperformed ChatGPT-o1 across all rating dimensions, with highly significant differences across all criteria (mean ratings: DeepSeek-R1 4.51, SD 0.73 vs ChatGPT-o1 3.73, SD 0.98; P <.001). In an exploratory subspecialty-level analysis, DeepSeek-R1 descriptively outperformed ChatGPT-o1 across all subspecialties. For both LLMs accumulated, junior residents tended to rate slightly higher than seniors in 7 of 9 criteria, although differences were not statistically significant. However, for ChatGPT-o1, junior residents rated significantly higher in overall score across all criteria (juniors 3.81, SD 0.64 vs seniors 3.63, SD 0.65; P =.02). Image-based responses by ChatGPT-o1 scored significantly lower than text-based (mean 3.19, SD 1.42; P =.007), particularly in factual accuracy (mean 2.75, SD 1.45; P <.001) and clinical practicality (mean 3.11, SD 1.47; P =.03). Conclusions: Both DeepSeek-R1 and ChatGPT-o1 demonstrate promising potential on simulated radiology question sets designed for educational and clinical contexts, with DeepSeek-R1 outperforming ChatGPT-o1 across all evaluated criteria. These results emphasize the value of open-source models for educational use and provide early evidence that LLMs may support radiology resident training under controlled conditions; however, their real-world educational and clinical effects require further investigation. Future research should prospectively evaluate how LLMs can be integrated into radiology training, assess their impact alongside conventional teaching methods, and investigate multimodal capabilities to better reflect realistic clinical scenarios.