Michael Püsken
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Michael Püsken has a total of 89 co-authors. Here are the top co-authors by number of shared publications:
- Andra-Iza Iuga
Institute of Diagnostic and Interventional Radiology, Faculty of Medicine and University Hospital Cologne, University of Cologne, 50937 Cologne, Germany
2 shared publications - Bettina Baeßler
Department of Diagnostic and Interventional Radiology, University Hospital Würzburg, Würzburg, Germany
2 shared publications David Maintz
Institute for Diagnostic and Interventional Radiology, University Cologne, Faculty of Medicine and University Hospital Cologne, Kerpener Straße 62, 50937, Cologne, Germany
2 shared publications- David Maintz
Institute for Diagnostic and Interventional Radiology, Faculty of Medicine, University Hospital Cologne, University of Cologne, 50937 Cologne, Germany
2 shared publications Eva M. Fallenberg
2 shared publicationsKerstin Rhiem
2 shared publications
Latest Publications
- Top Cited Papers
- 1 May 2023
- Published by Radiological Society of North America (RSNA) in Radiology
- Vol. 307 (4) , e222176
Automation Bias in Mammography: The Impact of Artificial Intelligence BI-RADS Suggestions on Reader Performance
Background Automation bias (the propensity for humans to favor suggestions from automated decision-making systems) is a known source of error in human-machine interactions, but its implications regarding artificial intelligence (AI)-aided mammography reading are unknown. Purpose To determine how automation bias can affect inexperienced, moderately experienced, and very experienced radiologists when reading mammograms with the aid of an artificial intelligence (AI) system. Materials and Methods In this prospective experiment, 27 radiologists read 50 mammograms and provided their Breast Imaging Reporting and Data System (BI-RADS) assessment assisted by a purported AI system. Mammograms were obtained between January 2017 and December 2019 and were presented in two randomized sets. The first was a training set of 10 mammograms, with the correct BI-RADS category suggested by the AI system. The second was a set of 40 mammograms in which an incorrect BI-RADS category was suggested for 12 mammograms. Reader performance, degree of bias in BI-RADS scoring, perceived accuracy of the AI system, and reader confidence in their own BI-RADS ratings were assessed using analysis of variance (ANOVA) and repeated-measures ANOVA followed by post hoc tests and Kruskal-Wallis tests followed by the Dunn post hoc test. Results The percentage of correctly rated mammograms by inexperienced (mean, 79.7% ± 11.7 [SD] vs 19.8% ± 14.0; P < .001; r = 0.93), moderately experienced (mean, 81.3% ± 10.1 vs 24.8% ± 11.6; P < .001; r = 0.96), and very experienced (mean, 82.3% ± 4.2 vs 45.5% ± 9.1; P = .003; r = 0.97) radiologists was significantly impacted by the correctness of the AI prediction of BI-RADS category. Inexperienced radiologists were significantly more likely to follow the suggestions of the purported AI when it incorrectly suggested a higher BI-RADS category than the actual ground truth compared with both moderately (mean degree of bias, 4.0 ± 1.8 vs 2.4 ± 1.5; P = .044; r = 0.46) and very (mean degree of bias, 4.0 ± 1.8 vs 1.2 ± 0.8; P = .009; r = 0.65) experienced readers. Conclusion The results show that inexperienced, moderately experienced, and very experienced radiologists reading mammograms are prone to automation bias when being supported by an AI-based system. This and other effects of human and machine interaction must be considered to ensure safe deployment and accurate diagnostic performance when combining human readers and AI. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Baltzer in this issue.
- 1 May 2023
- Published by Radiological Society of North America (RSNA) in Radiology
- Vol. 307 (4) , e222176
Top Cited PapersAutomation Bias in Mammography: The Impact of Artificial Intelligence BI-RADS Suggestions on Reader Performance
Background Automation bias (the propensity for humans to favor suggestions from automated decision-making systems) is a known source of error in human-machine interactions, but its implications regarding artificial intelligence (AI)-aided mammography reading are unknown. Purpose To determine how automation bias can affect inexperienced, moderately experienced, and very experienced radiologists when reading mammograms with the aid of an artificial intelligence (AI) system. Materials and Methods In this prospective experiment, 27 radiologists read 50 mammograms and provided their Breast Imaging Reporting and Data System (BI-RADS) assessment assisted by a purported AI system. Mammograms were obtained between January 2017 and December 2019 and were presented in two randomized sets. The first was a training set of 10 mammograms, with the correct BI-RADS category suggested by the AI system. The second was a set of 40 mammograms in which an incorrect BI-RADS category was suggested for 12 mammograms. Reader performance, degree of bias in BI-RADS scoring, perceived accuracy of the AI system, and reader confidence in their own BI-RADS ratings were assessed using analysis of variance (ANOVA) and repeated-measures ANOVA followed by post hoc tests and Kruskal-Wallis tests followed by the Dunn post hoc test. Results The percentage of correctly rated mammograms by inexperienced (mean, 79.7% ± 11.7 [SD] vs 19.8% ± 14.0; P < .001; r = 0.93), moderately experienced (mean, 81.3% ± 10.1 vs 24.8% ± 11.6; P < .001; r = 0.96), and very experienced (mean, 82.3% ± 4.2 vs 45.5% ± 9.1; P = .003; r = 0.97) radiologists was significantly impacted by the correctness of the AI prediction of BI-RADS category. Inexperienced radiologists were significantly more likely to follow the suggestions of the purported AI when it incorrectly suggested a higher BI-RADS category than the actual ground truth compared with both moderately (mean degree of bias, 4.0 ± 1.8 vs 2.4 ± 1.5; P = .044; r = 0.46) and very (mean degree of bias, 4.0 ± 1.8 vs 1.2 ± 0.8; P = .009; r = 0.65) experienced readers. Conclusion The results show that inexperienced, moderately experienced, and very experienced radiologists reading mammograms are prone to automation bias when being supported by an AI-based system. This and other effects of human and machine interaction must be considered to ensure safe deployment and accurate diagnostic performance when combining human readers and AI. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Baltzer in this issue.
- Open Access
- 28 December 2022
- Published by MDPI AG in Journal of Personalized Medicine
- Vol. 13 (1) , 64
Preoperative Perforator Mapping in DIEP Flaps for Breast Reconstruction. The Impact of New Contrast-Enhanced Ultrasound Techniques
Deep inferior epigastric artery flaps (DIEP) represent the gold standard of autologous breast reconstruction. Due to significant variations in vascular anatomy, preoperative perforator mapping (PM) is mandatory in order to ensure the presence of a sufficient perforator within the flap. In this regard, CT angiography (CTA) is currently the method of choice. Therefore, we investigated the value of contrast-enhanced ultrasound (CEUS) techniques for preoperative PM in comparison to CTA. Patients underwent PM, utilizing both CTA and CEUS techniques. Documentation included the course of the vascular pedicle through the rectus muscle (M), fascial penetration (F), the subcutaneous plexus (P) and the skin point (SP) on either side of the abdomen. Thus, contrast-enhanced B-Flow (BCEUS), B-Flow ultrasound (BUS), CEUS, color Doppler ultrasound (CDUS) and CTA were evaluated in terms of the diagnostic consistency and effectiveness of PM. Precision (∆L) was then calculated in relation to the actual intraoperative location. Statistical analysis included Kruskall–Wallis, Levene and Bonferroni tests, as well as Spearman correlations. A total of 39 DIEP flaps were analyzed. Only CTA (∆L = 2.85 mm) and BCEUS (∆L = 4.57 mm) enabled complete PM, also including P and SP, whereas CDUS, CEUS and BUS enabled clear PM throughout M and F only. Regarding the number of detected perforators, PM techniques are ranked from high to low as follows: CTA, BCEUS, BUS, CEUS and CDUS. CTA and BCEUS showed sufficient diagnostic consistency for SP, P and F, while CDUS and CTA had a superior performance for M. BCEUS offers precise image-controlled surface tags and dynamic information for PM without imposing radiation and may, therefore, be considered a feasible add-on or alternative to CTA. However, BCEUS requires an experienced examiner and is more time-consuming.
- 28 December 2022
- Published by MDPI AG in Journal of Personalized Medicine
- Vol. 13 (1) , 64
Open AccessPreoperative Perforator Mapping in DIEP Flaps for Breast Reconstruction. The Impact of New Contrast-Enhanced Ultrasound Techniques
Deep inferior epigastric artery flaps (DIEP) represent the gold standard of autologous breast reconstruction. Due to significant variations in vascular anatomy, preoperative perforator mapping (PM) is mandatory in order to ensure the presence of a sufficient perforator within the flap. In this regard, CT angiography (CTA) is currently the method of choice. Therefore, we investigated the value of contrast-enhanced ultrasound (CEUS) techniques for preoperative PM in comparison to CTA. Patients underwent PM, utilizing both CTA and CEUS techniques. Documentation included the course of the vascular pedicle through the rectus muscle (M), fascial penetration (F), the subcutaneous plexus (P) and the skin point (SP) on either side of the abdomen. Thus, contrast-enhanced B-Flow (BCEUS), B-Flow ultrasound (BUS), CEUS, color Doppler ultrasound (CDUS) and CTA were evaluated in terms of the diagnostic consistency and effectiveness of PM. Precision (∆L) was then calculated in relation to the actual intraoperative location. Statistical analysis included Kruskall–Wallis, Levene and Bonferroni tests, as well as Spearman correlations. A total of 39 DIEP flaps were analyzed. Only CTA (∆L = 2.85 mm) and BCEUS (∆L = 4.57 mm) enabled complete PM, also including P and SP, whereas CDUS, CEUS and BUS enabled clear PM throughout M and F only. Regarding the number of detected perforators, PM techniques are ranked from high to low as follows: CTA, BCEUS, BUS, CEUS and CDUS. CTA and BCEUS showed sufficient diagnostic consistency for SP, P and F, while CDUS and CTA had a superior performance for M. BCEUS offers precise image-controlled surface tags and dynamic information for PM without imposing radiation and may, therefore, be considered a feasible add-on or alternative to CTA. However, BCEUS requires an experienced examiner and is more time-consuming.
- 11 March 2022
- Published by Georg Thieme Verlag KG in Onkologie up2date
- Vol. 04 (01) , 17-31
Brustkrebsfrüherkennung bei Frauen mit erhöhten Erkrankungsrisiken
Zusammenfassung: Für Patientinnen und gesunde Ratsuchende mit einem erhöhten Erkrankungsrisiko für Brustkrebs besteht ein langjährig etabliertes Programm zur intensivierten Früherkennung und Nachsorge in den spezialisierten Zentren des Deutschen Konsortiums Familiärer Brust- und Eierstockkrebs (https://www.konsortium-familiaerer-brustkrebs.de/). Dabei kommt der MR-Mammografie aufgrund der sehr hohen Sensitivität beim Tumornachweis eine besondere Bedeutung zu.
- 11 March 2022
- Published by Georg Thieme Verlag KG in Onkologie up2date
- Vol. 04 (01) , 17-31
Brustkrebsfrüherkennung bei Frauen mit erhöhten Erkrankungsrisiken
Zusammenfassung: Für Patientinnen und gesunde Ratsuchende mit einem erhöhten Erkrankungsrisiko für Brustkrebs besteht ein langjährig etabliertes Programm zur intensivierten Früherkennung und Nachsorge in den spezialisierten Zentren des Deutschen Konsortiums Familiärer Brust- und Eierstockkrebs (https://www.konsortium-familiaerer-brustkrebs.de/). Dabei kommt der MR-Mammografie aufgrund der sehr hohen Sensitivität beim Tumornachweis eine besondere Bedeutung zu.
- Open Access
- 19 July 2021
- Published by S. Karger AG in Breast Care
- Vol. 17 (2) , 199-207
Consensus Recommendations of the German Consortium for Hereditary Breast and Ovarian Cancer
Background: The German Consortium for Hereditary Breast and Ovarian Cancer (GC-HBOC) has established a multigene panel (TruRisk®) for the analysis of risk genes for familial breast and ovarian cancer. Summary: An interdisciplinary team of experts from the GC-HBOC has evaluated the available data on risk modification in the presence of pathogenic mutations in these genes based on a structured literature search and through a formal consensus process. Key Messages: The goal of this work is to better assess individual disease risk and, on this basis, to derive clinical recommendations for patient counseling and care at the centers of the GC-HBOC from the initial consultation prior to genetic testing to the use of individual risk-adapted preventive/therapeutic measures.
- 19 July 2021
- Published by S. Karger AG in Breast Care
- Vol. 17 (2) , 199-207
Open AccessConsensus Recommendations of the German Consortium for Hereditary Breast and Ovarian Cancer
Background: The German Consortium for Hereditary Breast and Ovarian Cancer (GC-HBOC) has established a multigene panel (TruRisk®) for the analysis of risk genes for familial breast and ovarian cancer. Summary: An interdisciplinary team of experts from the GC-HBOC has evaluated the available data on risk modification in the presence of pathogenic mutations in these genes based on a structured literature search and through a formal consensus process. Key Messages: The goal of this work is to better assess individual disease risk and, on this basis, to derive clinical recommendations for patient counseling and care at the centers of the GC-HBOC from the initial consultation prior to genetic testing to the use of individual risk-adapted preventive/therapeutic measures.
- 1 June 2021
- Published by Elsevier in European Journal of Radiology
- Vol. 139, 109718
Automated mapping and N-Staging of thoracic lymph nodes in contrast-enhanced CT scans of the chest using a fully convolutional neural network
- 1 June 2021
- Published by Elsevier in European Journal of Radiology
- Vol. 139, 109718
Automated mapping and N-Staging of thoracic lymph nodes in contrast-enhanced CT scans of the chest using a fully convolutional neural network
- 27 April 2021
- Published by Georg Thieme Verlag KG in Radiologie up2date
- Vol. 21 (03) , 211-223
Brustkrebsfrüherkennung bei Frauen mit erhöhten Erkrankungsrisiken
Zusammenfassung: Für Patientinnen und gesunde Ratsuchende mit einem erhöhten Erkrankungsrisiko für Brustkrebs besteht ein langjährig etabliertes Programm zur intensivierten Früherkennung und Nachsorge in den spezialisierten Zentren des Deutschen Konsortiums Familiärer Brust- und Eierstockkrebs (https://www.konsortium-familiaerer-brustkrebs.de/). Dabei kommt der MR-Mammografie aufgrund der sehr hohen Sensitivität beim Tumornachweis eine besondere Bedeutung zu.
- 27 April 2021
- Published by Georg Thieme Verlag KG in Radiologie up2date
- Vol. 21 (03) , 211-223
Brustkrebsfrüherkennung bei Frauen mit erhöhten Erkrankungsrisiken
Zusammenfassung: Für Patientinnen und gesunde Ratsuchende mit einem erhöhten Erkrankungsrisiko für Brustkrebs besteht ein langjährig etabliertes Programm zur intensivierten Früherkennung und Nachsorge in den spezialisierten Zentren des Deutschen Konsortiums Familiärer Brust- und Eierstockkrebs (https://www.konsortium-familiaerer-brustkrebs.de/). Dabei kommt der MR-Mammografie aufgrund der sehr hohen Sensitivität beim Tumornachweis eine besondere Bedeutung zu.
- Open Access
- 13 April 2021
- Published by Springer Nature in BMC Medical Imaging
- Vol. 21 (1) , 1-12
Automated detection and segmentation of thoracic lymph nodes from CT using 3D foveal fully convolutional neural networks
Background: In oncology, the correct determination of nodal metastatic disease is essential for patient management, as patient treatment and prognosis are closely linked to the stage of the disease. The aim of the study was to develop a tool for automatic 3D detection and segmentation of lymph nodes (LNs) in computed tomography (CT) scans of the thorax using a fully convolutional neural network based on 3D foveal patches.Methods: The training dataset was collected from the Computed Tomography Lymph Nodes Collection of the Cancer Imaging Archive, containing 89 contrast-enhanced CT scans of the thorax. A total number of 4275 LNs was segmented semi-automatically by a radiologist, assessing the entire 3D volume of the LNs. Using this data, a fully convolutional neuronal network based on 3D foveal patches was trained with fourfold cross-validation. Testing was performed on an unseen dataset containing 15 contrast-enhanced CT scans of patients who were referred upon suspicion or for staging of bronchial carcinoma.Results: The algorithm achieved a good overall performance with a total detection rate of 76.9% for enlarged LNs during fourfold cross-validation in the training dataset with 10.3 false-positives per volume and of 69.9% in the unseen testing dataset. In the training dataset a better detection rate was observed for enlarged LNs compared to smaller LNs, the detection rate for LNs with a short-axis diameter (SAD) ≥ 20 mm and SAD 5–10 mm being 91.6% and 62.2% (p < 0.001), respectively. Best detection rates were obtained for LNs located in Level 4R (83.6%) and Level 7 (80.4%).Conclusions: The proposed 3D deep learning approach achieves an overall good performance in the automatic detection and segmentation of thoracic LNs and shows reasonable generalizability, yielding the potential to facilitate detection during routine clinical work and to enable radiomics research without observer-bias.
- 13 April 2021
- Published by Springer Nature in BMC Medical Imaging
- Vol. 21 (1) , 1-12
Open AccessAutomated detection and segmentation of thoracic lymph nodes from CT using 3D foveal fully convolutional neural networks
Background: In oncology, the correct determination of nodal metastatic disease is essential for patient management, as patient treatment and prognosis are closely linked to the stage of the disease. The aim of the study was to develop a tool for automatic 3D detection and segmentation of lymph nodes (LNs) in computed tomography (CT) scans of the thorax using a fully convolutional neural network based on 3D foveal patches.Methods: The training dataset was collected from the Computed Tomography Lymph Nodes Collection of the Cancer Imaging Archive, containing 89 contrast-enhanced CT scans of the thorax. A total number of 4275 LNs was segmented semi-automatically by a radiologist, assessing the entire 3D volume of the LNs. Using this data, a fully convolutional neuronal network based on 3D foveal patches was trained with fourfold cross-validation. Testing was performed on an unseen dataset containing 15 contrast-enhanced CT scans of patients who were referred upon suspicion or for staging of bronchial carcinoma.Results: The algorithm achieved a good overall performance with a total detection rate of 76.9% for enlarged LNs during fourfold cross-validation in the training dataset with 10.3 false-positives per volume and of 69.9% in the unseen testing dataset. In the training dataset a better detection rate was observed for enlarged LNs compared to smaller LNs, the detection rate for LNs with a short-axis diameter (SAD) ≥ 20 mm and SAD 5–10 mm being 91.6% and 62.2% (p < 0.001), respectively. Best detection rates were obtained for LNs located in Level 4R (83.6%) and Level 7 (80.4%).Conclusions: The proposed 3D deep learning approach achieves an overall good performance in the automatic detection and segmentation of thoracic LNs and shows reasonable generalizability, yielding the potential to facilitate detection during routine clinical work and to enable radiomics research without observer-bias.
- 1 March 2020
- Published by Wolters Kluwer Health in Journal of Computer Assisted Tomography
- Vol. 44 (2) , 236-241
Semiautomated Renal Cortex Volumetry in Multislice Computed Tomography: Effect of Slice Thickness and Iterative Reconstruction Algorithms
Objective The aim of the study was to evaluate the effect of slice thickness, iterative reconstruction (IR) algorithm, and kernel selection on measurement accuracy and interobserver variability for semiautomated renal cortex volumetry (RCV) with multislice computed tomography (CT). Methods Ten patients (62.4 +/- 17.2 years) undergoing abdominal biphasic multislice computed tomography were enrolled in this retrospective study. Computed tomography data sets were reconstructed at 1-, 2-, and 5-mm slice thickness with 2 different IR algorithms (iDose, IMRST) and 2 different kernels (IMRS and IMRR) (Philips, the Netherlands). Two readers independently performed semiautomated RCV for each reconstructed data set to calculate left kidney volume (LKV) and split renal function (SRF). Statistics were calculated using analysis of variance with Geisser-Greenhouse correction, followed by Tukey multiple comparisons post hoc test. Statistical significance was defined as P <= 0.05. Results Semiautomated RCV of 120 data sets (240 kidneys) was successfully performed by both readers. Semiautomated RCV provides comparable results for LKV and SRF with 3 different slice thicknesses, 2 different IR algorithms, and 2 different kernels. Only the 1-mm slice thickness showed significant differences for LKV between IMRR and IMRS (P = 0.02, mean difference = 4.28 bb) and IMRST versus IMRS (P = 0.02, mean difference = 4.68 cm(3)) for reader 2. Interobserver variability was low between both readers irrespective of slice thickness and reconstruction algorithm (0.82 >= P >= 0.99). Conclusions Semiautomated RCV measurements of LKV and SRF are independent of slice thickness, IR algorithm, and kernel selection. These findings suggest that comparisons between studies using different slice thicknesses and reconstruction algorithms for RCV are valid.
- 1 March 2020
- Published by Wolters Kluwer Health in Journal of Computer Assisted Tomography
- Vol. 44 (2) , 236-241
Semiautomated Renal Cortex Volumetry in Multislice Computed Tomography: Effect of Slice Thickness and Iterative Reconstruction Algorithms
Objective The aim of the study was to evaluate the effect of slice thickness, iterative reconstruction (IR) algorithm, and kernel selection on measurement accuracy and interobserver variability for semiautomated renal cortex volumetry (RCV) with multislice computed tomography (CT). Methods Ten patients (62.4 +/- 17.2 years) undergoing abdominal biphasic multislice computed tomography were enrolled in this retrospective study. Computed tomography data sets were reconstructed at 1-, 2-, and 5-mm slice thickness with 2 different IR algorithms (iDose, IMRST) and 2 different kernels (IMRS and IMRR) (Philips, the Netherlands). Two readers independently performed semiautomated RCV for each reconstructed data set to calculate left kidney volume (LKV) and split renal function (SRF). Statistics were calculated using analysis of variance with Geisser-Greenhouse correction, followed by Tukey multiple comparisons post hoc test. Statistical significance was defined as P <= 0.05. Results Semiautomated RCV of 120 data sets (240 kidneys) was successfully performed by both readers. Semiautomated RCV provides comparable results for LKV and SRF with 3 different slice thicknesses, 2 different IR algorithms, and 2 different kernels. Only the 1-mm slice thickness showed significant differences for LKV between IMRR and IMRS (P = 0.02, mean difference = 4.28 bb) and IMRST versus IMRS (P = 0.02, mean difference = 4.68 cm(3)) for reader 2. Interobserver variability was low between both readers irrespective of slice thickness and reconstruction algorithm (0.82 >= P >= 0.99). Conclusions Semiautomated RCV measurements of LKV and SRF are independent of slice thickness, IR algorithm, and kernel selection. These findings suggest that comparisons between studies using different slice thicknesses and reconstruction algorithms for RCV are valid.