Improving lung segmentation for higher coverage of clinically-relevant findings (September 2023).

• Here we describe and evaluate a lung segmentation algorithm that yields high accuracy despite disease patterns thanks to a 3D architecture and diverse training data set. We compare our algorithm with the established state-of-the-art algorithm.

Why HU may not be the best approach to emphysema quantification: a contextflow whitepaper. (October 2022). Emphysema quantification with contextflow SEARCH Lung CT

• AI is able to measure the extent of emphysema more accurately, shows more accurate detection results when quantified in AI than HU.

Prayer, F., Röhrich, S., Pan, J. et al. Künstliche Intelligenz in der Bildgebung der Lunge. Radiologe 60, 42–47 (2020). https://doi.org/10.1007/s00117-019-00611-2

Röhrich, S., Schlegl, T., Bardach, C. et al. Deep learning detection and quantification of pneumothorax in heterogeneous routine chest computed tomography. European Radiology Exp 4, 26 (2020).
https://doi.org/10.1186/s41747-020-00152-7

Röhrich, S., Heidinger, B.H., Prayer, F. et al. Impact of a content-based image retrieval system on the interpretation of chest CTs of patients with diffuse parenchymal lung disease. European Radiology (2022). https://doi.org/10.1007/s00330-022-08973-3

This study shows 30% less time consumed when reading chest CTs with AI
Adams, S., Madtes, K., Burbridge, B. et al. Clinical Impact and Generalizability of a Computer-Assisted Diagnostic Tool to Risk-Stratify Lung Nodules With CT. JACR, in press (2022).
https://doi.org/10.1016/j.jacr.2022.08.006

Des conférences ont présenté l’utilisation clinique du CFA lors du congrès européen de radiologie (ECR) 2023 à Vienne.

ECR 2023 (March, Vienna)

• Halfmann, M. et al. Pre-interventional AI-supported automated lung parenchyma quantification predicts post-interventional complications in CT-guided lung biopsies (2023)
  • Results: AI-based quantification of lung texture has potentially a predictive value for complications. Greater amount of pre-interventional consolidations (p=0.03) and smaller lesion size (p=0.04) were predictors for post-interventional pneumothorax.
• Pan, J. et al. Deep learning quantifies disease patterns in lung CT associated with individual outcome in idiopathic pulmonary fibrosis (2023)
  • Results: Patterns honeycombing and reticulation have a predictive power for individual patient outcome (death) in IPF patients.
• Perkonigg, M. et al. Comparing Emphysema Detection based on a Threshold and Deep Learning (2023)

European Society of Thoracic Imaging/ESTI 2022 (June, Oxford)

• Röhrich, S. et al. Evaluation of diagnosing diffuse parenchymal lung disease in pulmonary CTs (2022).
• Pan, J. et al. Prediction of disease severity in COVID-19 patients identifies predictive disease patterns in lung CT (2022).

ECR 2022 (July, Vienna)

• Pieler, M. et al. Evaluation of automatic volumetry of honeycombing and ground glass opacity patterns in lung CT scans (2022).
• Röhrich, S. Impact of a content-based image retrieval system on the interpretation of chest CTs of patients with diffuse parenchymal lung disease (2022).
• Agarwal, P. The next generation of reference books: Combining Content-Based Image Retrieval with a knowledge-based diagnostic decision support system in chest-CT (2022).

World Conference on Lung Cancer – IASLC 2022 (August, Vienna)

• Calhoun, M. E. et al. Combining automated malignancy risk estimation with lung nodule detection may reduce physician effort and increase diagnostic accuracy (2022).

Pending

• Dermatomyositis – description of the cohort, in terms of quantitative profiles, Dr F. Preyer et al. (Med Uni Vienna)
• Comparing predictive values for patterns in patients with/without ICU treatment, J. Pan/G. Langs (Med Uni Vienna)

NL/Leiden University Hospital (LUMC)
A study on nodule detection performance with vs. without CFS, and how AI assistance contributes to the wellbeing of radiologists.

SE/Jönköping Hospitals
Assessment of the quality of nodule and pattern detecion, and quality of reports with vs. without CFS.

DE/Medical University Mainz
Our long-term partner on research projects on ILDs and clinical use of AI.

Cambridge University (UK)
Long-term research on lung volume quantification for COVID and other ILDs prognosis.

Medical University Vienna (AT)
Our long-term partner on research studies on ILDs and pulmonary nodules.
A study on quantification of lung patterns relevant to dermatomyositis for disease detection
and prediction.

Charité Berlin (DE)
Assessing quantified lung profiles of patients with HFpEF.

Kepler University Hospital Linz (AT)
MILD study, Prof. Kaveh Akbari: a prospective study on comparing information from CT vs. MRI for fibrosis patients

In the imaging diagnostics of interstitial lung diseases, the study situation indicates that MRI can provide additional information compared to CT – for example with regard to the differentiation of inflammatory versus fibrotic genesis, early diagnosis of lung involvement in the context of collagen vascular disease or also with regard to the quantification of the disease during the course of therapy.

With the MILD study they want to analyze these points in a prospective setting. They will also examine to what extent the quantitative CT analysis using contextflow can also provide this additional information.

Study: Lung radionomics vs. lung function test: quantitative profiling for treatment response vs. manual lung pattern scores and lung function for 6/12/24 months.

Medical University Düsseldorf (DE)
A retrospective study on evaluating the correlation between the % of GGO and fibrosis in lungs.