HHHelge Haarmann

Helge Haarmann

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Helge Haarmann has a total of 19 co-authors. Here are the top co-authors by number of shared publications:

  • Eva Rasenack

    Department of Cardiology and Pneumology University Medical Center Göttingen (UMG) Göttingen Germany

    3 shared publications
  • Leonard Bergau

    Department of Cardiology and Pneumology University Medical Center Göttingen (UMG) Göttingen Germany

    3 shared publications
  • Markus Zabel

    Department of Cardiology and Pneumology University Medical Center Göttingen (UMG) Göttingen Germany

    3 shared publications
  • Nibras Soubh

    Department of Cardiology and Pneumology University Medical Center Göttingen (UMG) Göttingen Germany

    3 shared publications
  • Constanze Schmidt

    Department of Cardiology and Pneumology, University Medical Center Göttingen (UMG), Robert-Koch-Str. 40, Göttingen, 37075, Germany, 49 15114609645

    2 shared publications
  • Felix Wiedmann

    Department of Cardiology, Angiology and Pneumology, University Hospital Heidelberg Heidelberg, Germany; DZHK (German Center for Cardiovascular Research), Partner Site Heidelberg/Mannheim, University of Heidelberg Heidelberg Germany

    2 shared publications

Latest Publications

  • Open Access

    First Experience With Extravascular Implantable Cardioverter‐Defibrillator Under Deep Sedation

    by Nibras Soubh, Bernhard C. Danner, Ulrich Krause, Matthias J. Müller, Markus Zabel, Gerd Hasenfuß, Constanze Schmidt, Helge Haarmann, Eva Rasenack, Leonard Bergau

    Background: The Extravascular Implantable Cardioverter‐Defibrillator (EV‐ICD) utilizes a substernal lead to provide defibrillation and anti‐tachycardia pacing (ATP) while avoiding transvenous complications. General anesthesia (GA) was applied for implantation procedures in the EV‐ICD pivotal trial and is currently recommended by the manufacturer. However, GA carries specific risks and consumes significant resources. This study evaluates the feasibility, safety, and procedural efficiency of EV‐ICD implantation performed under cardiologist‐administered deep sedation with noninvasive ventilation (DS‐NIV) compared to standard GA. Methods: We retrospectively analyzed 24 consecutive patients undergoing EV‐ICD implantation in our center. Patients received either GA ( n = 14) or DS‐NIV ( n = 10) using a propofol‐ketamine protocol delivered by cardiologists. Analyses focused on peri‐procedural feasibility and safety, procedural workflow and anesthesia characteristics, and early device electrical performance. Results: No anesthesia‐related complications occurred in either group. The DS‐NIV group demonstrated significantly shorter wheels‐to‐incision times (median: 45 vs. 70 min, p = 0.022), whereas the procedural duration did not differ significantly from the GA group. Patients receiving DS‐NIV required fewer vasopressors (60% vs. 100%, p = 0.024) despite higher propofol infusion rates (800 vs. 350 mg/h, P <0.001). Defibrillation testing success and electrical parameters were comparable. In a median follow‐up of 182 days two patients received appropriate EV‐ICD therapies and no inappropriate ATP or shocks were delivered. Conclusion: EV‐ICD implantation under cardiologist‐administered DS‐NIV appears to be feasible and safe, offering improved workflow efficiency compared to GA. These findings support deep sedation as a practical alternative in experienced centers, potentially expanding access to EV‐ICD therapy.


  • Open Access

    Performance of Vision-Enabled Large Language Models in Image-Based Electrocardiogram Interpretation: Exploratory Evaluation

    by Nibras Soubh, Eva Rasenack, Helge Haarmann, Felix Wiedmann, Markus Zabel, Constanze Schmidt, Rayan Suliman, Leonard Bergau

    Background: Vision-enabled large language models (VE-LLMs) have the potential to provide flexible and explainable medical image interpretation. However, their real-world performance on clinical data, such as 12-lead electrocardiograms (ECGs), has not been systematically assessed. Objective: This study aimed to evaluate the diagnostic accuracy and reliability of state-of-the-art generalist VE-LLMs in interpreting real-world ECG images. Methods: We tested 6 generalist VE-LLMs (ChatGPT-5, ChatGPT-4, Gemini 2.5, Copilot, Claude Sonnet-4, and Claude Opus-4.1) using 70 deidentified ECG images. A standardized prompt requested 9 determinations: rhythm, first-degree atrioventricular (AV) block, intraventricular conduction block and pattern, corrected QT (QTc) prolongation, premature atrial and ventricular contractions, ischemic ST-segment deviation, and axis deviation. An expert consensus served as the reference standard. Moreover, 2 image-based ECG-specialized LLMs (PULSE-7B and ECG-Instruct-Llama-3.2-11B-Vision) were tested for exploratory comparison. Model outputs were evaluated using overall and per-category diagnostic metrics. Results: Overall balanced accuracy across generalist models ranged from 50.1% to 61.8% (Cochran Q, P <.001). ChatGPT-5 achieved the highest balanced accuracy (61.8%) but had the slowest response time (median 276, IQR 110-407 s), whereas Copilot responded within a median of 3 (IQR 2-4) seconds. Balanced accuracy for rhythm classification ranged from 38.6% to 55.8%, but sensitivity for atrial fibrillation among generalist models was ≤11.1%, detecting either none or only 1 of the 9 cases. Detection of first-degree AV block (sensitivity 0%‐22%; 0/9 to 2/9) and QTc prolongation (sensitivity 0%‐45.5%; 0/22 to 10/22) was poor. Intraventricular block was identified with up to 67.8% balanced accuracy, but correct subtype assignment was ≤44% (≤11/25). ST-segment deviation sensitivity was <25% for all generalist models (highest 3/14). Agreement with expert interpretation was low, with Cohen κ indicating poor-to-fair concordance (κ≤0.39). Specialized models achieved overall balanced accuracy of 56.5% (ECG-Instruct-Llama-3.2-11B-Vision) and 64.4% (PULSE-7B), with PULSE-7B showing higher task-specific balanced accuracy in rhythm classification and ectopic beats detection (up to 86.3% and 89.2%, respectively). Conclusions: VE-LLMs showed moderate overall performance but mostly low sensitivity and limited agreement with expert ECG interpretation. Current performance remains inconsistent across models and diagnostic categories and is insufficient to support clinical deployment.


  • Open Access

    Performance of Vision-Enabled Large Language Models in ECG Interpretation: Exploratory Evaluation (Preprint)

    by Nibras Soubh, Eva Rasenack, Helge Haarmann, Felix Wiedmann, Markus Zabel, Constanze Schmidt, Rayan Suliman, Leonard Bergau

    BACKGROUND Vision-enabled large language models (VE-LLMs) have the potential to provide flexible and explainable medical image interpretation. However, their real-world performance on clinical data such as 12-lead electrocardiograms (ECGs) has not been systematically assessed. OBJECTIVE This study aimed to evaluate the diagnostic accuracy and reliability of state-of-the-art VE-LLMs in interpreting real-world ECG images. METHODS We tested eight VE-LLMs (ChatGPT-5, ChatGPT-4, Gemini 2.5 Pro, Copilot, Grok-4, Perplexity, Claude Sonnet-4, and Claude Opus-4.1) using 70 de-identified ECG images. A standardized prompt requested nine determinations: rhythm, first-degree atrioventricular (AV) block, intraventricular conduction block and pattern, corrected QT (QTc) prolongation, premature atrial and ventricular contractions, ischemic ST-segment deviation, and axis deviation. An expert consensus served as the reference standard. Model outputs were evaluated using overall and per-class diagnostic metrics. RESULTS Overall accuracy across models varied significantly from 68.1% to 78.3% (429/630 to 493/630, Cochran’s Q, P<.001). ChatGPT-5 achieved the highest accuracy (78.3%) but had the slowest response time (median 276 seconds), whereas Perplexity and Copilot responded within a median of 2 and 3 seconds, respectively. Rhythm classification reached 72.9%–82.9% accuracy (51/70 to 58/70), but sensitivity for atrial fibrillation was ≤22% (≤2/9). Detection of first-degree AV block was poor (sensitivity 0%–33%; 0/9 to 3/9), and QTc prolongation was also poor (sensitivity 0%–45.5%; 0/22 to 10/22). Intraventricular block was identified with up to 70% accuracy (49/70), but correct subtype assignment was ≤44% (≤11/25). ST-segment deviation sensitivity was <25% for all models (highest 3/14). Agreement with expert interpretation was low, with Cohen’s kappa (κ) indicating poor-to-fair concordance (κ≤.37). CONCLUSIONS VE-LLMs showed moderate overall accuracy but low sensitivity and limited agreement with expert ECG interpretation. Current performance is inconsistent and insufficient for clinical deployment. Future development should focus on domain-specific training and hybrid approaches combining LLM reasoning with established ECG algorithms before use in patient care.

  • First experience with a new tool for automatic mapping of fragmented signals in a case report of cardioneuroablation

    by Helge Haarmann, Besir Hasan, Nibras Soubh, Eva Rasenack, Simon Schlögl, Markus Zabel, Leonard Bergau

    Cardioneuroablation is a treatment option for patients with recurrent vasovagal syncope (VVS). Ablation targets of parasympathetic ganglionated plexi (GP) adjacent to the right and left atrial walls and distal endocardial inputs of these GP can be identified both by their anatomical localization and by intracardiac mapping of fragmented electrogram signals. In this case of a successful cardioneuroablation of a 22-year-old patient suffering from recurrent VVS, a new algorithm for automatic mapping of fragmented signals (CARTO Elevate Module, Biosense Webster) was used to identify areas of GP. In this first experience, automatic tags of fragmented signals were matching well with anatomically guided ablation points. This new tool for automatic identification of fragmented signals may facilitate and improve cardioneuroablation procedures.