Lukas Fischer is Research Manager for Data Science at the Software Competence Center Hagenberg (SCCH).He did his MSc in Medical Informatics at the TU Wien (TUW) with focus on medical image segmentation, statistical shape models, image registration, bio-inspired optimization algorithms and machine learning. He continued his PhD studies and research in the domain of medical imaging/medical physics as a research assistant at the Computational Imaging Research Lab (CIR) at the Medical University of Vienna (MUW). His research focus was on the computer vision based quantification of trabecular microarchitecture in patients suffering from severe osteoporosis after lung transplantation.His current main research interests are in Machine Learning, with special focus on Deep Learning, Generative Models (e.g. GANs, VAEs, AEs), Computer Vision, especially on medical data (e.g. segmentation, registration, classification, and tracking), Statistical Shape Models as well as Transfer Learning and Privacy Preserving Learning aspects.In addition to his research activities, he is an experienced project manager of various national and international research projects.
Poster: "ESSR 2013 / P-0078 / Reproducibility of cortical measures assessed with the thresho... more Poster: "ESSR 2013 / P-0078 / Reproducibility of cortical measures assessed with the threshold independent segmentation tool (TIST) in HR-pQCT images" by: "A. Valentinitsch1, L. Fischer2, M. D. DiFranco2, J. M. Patsch2, G. M. Gruber2, F. Kainberger2, G. Langs2; 1Wien/AT, 2Vienna/AT"
In the last decade, industry’s demand for deep learning (DL) has increased due to its high perfor... more In the last decade, industry’s demand for deep learning (DL) has increased due to its high performance in complex scenarios. Due to the DL method’s complexity, experts and non-experts rely on blackbox software packages such as Tensorflow and Pytorch. The frameworks are constantly improving, and new versions are released frequently. As a natural process in software development, the released versions contain improvements/changes in the methods and their implementation. Moreover, versions may be bug-polluted, leading to the model performance decreasing or stopping the model from working. The aforementioned changes in implementation can lead to variance in obtained results. This work investigates the effect of implementation changes in different major releases of these frameworks on the model performance. We perform our study using a variety of standard datasets. Our study shows that users should consider that changing the framework version can affect the model performance. Moreover, th...
In order to develop machine learning and deep learning models that take into account the guidelin... more In order to develop machine learning and deep learning models that take into account the guidelines and principles of trustworthy AI, a novel information theoretic approach is introduced in this article. A unified approach to privacy-preserving interpretable and transferable learning is considered for studying and optimizing the trade-offs between the privacy, interpretability, and transferability aspects of trustworthy AI. A variational membership-mapping Bayesian model is used for the analytical approximation of the defined information theoretic measures for privacy leakage, interpretability, and transferability. The approach consists of approximating the information theoretic measures by maximizing a lower-bound using variational optimization. The approach is demonstrated through numerous experiments on benchmark datasets and a real-world biomedical application concerned with the detection of mental stress in individuals using heart rate variability analysis.
The main challenges along with lessons learned from ongoing research in the application of machin... more The main challenges along with lessons learned from ongoing research in the application of machine learning systems in practice are discussed, taking into account aspects of theoretical foundations, systems engineering, and human-centered AI postulates. The analysis outlines a fundamental theory-practice gap which superimposes the challenges of AI system engineering at the level of data quality assurance, model building, software engineering and deployment.
Illustration of the cross-sectional cut planes for Âą1SD of the mean shape models. SD: Standard d... more Illustration of the cross-sectional cut planes for Âą1SD of the mean shape models. SD: Standard deviation; Distal: Distal plane; Middle: Middle plane; Proximal: Proximal plane. (TIF 9568 kb)
Zsfassung in dt. SpracheIn den letzten Jahren lieferten Segmentierungsansätze basierend auf seque... more Zsfassung in dt. SpracheIn den letzten Jahren lieferten Segmentierungsansätze basierend auf sequenzielle Monte Carlo Methoden vielversprechende Ergebnisse bei der Lokalisierung und Beschreibung anatomischer Strukturen in medizinisch relevanten Bildern. Auch bekannt unter der Bezeichnung Shape Particle Filter wurden diese Methoden für die Segmentierung von Wirbelkörpern, Lungenflügeln und Herzen eingesetzt. Ihr großer Vorteil liegt darin, dass sie auch bei Bildern mit starkem Rauschen wie zum Beispiel MR Aufnahmen, sowie bei sich überlagernden Strukturen, bei denen eine eindeutige Unterscheidung der Objekte schwierig ist, noch sehr gute Segmentierungsergebnisse liefern. Shape Particle Filter benötigen eine Maske, welche auf dem Mean Shape eines Shape Models basiert und in existierenden Implementierungen immer manuell definiert wird. Während der Suche nach einem Objekt wird die Wahrscheinlichkeit für jeden Pixel zu einer gewissen Region der Maske zu gehören durch das Klassifizieren vo...
Animated illustrations of the first five modes of all radius models. A: Female left radii model; ... more Animated illustrations of the first five modes of all radius models. A: Female left radii model; B: Female right radii model; C: Male left radii model; D: Male right radii model. (ZIP 9547 kb)
Poster: "ECR 2010 / B-566 / Towards automatic medical image segmentation using shape particl... more Poster: "ECR 2010 / B-566 / Towards automatic medical image segmentation using shape particle filters" by: "L. Fischer, R. Donner, F. Kainberger, G. Langs; Vienna/AT"
Poster: "ESSR 2013 / P-0078 / Reproducibility of cortical measures assessed with the thresho... more Poster: "ESSR 2013 / P-0078 / Reproducibility of cortical measures assessed with the threshold independent segmentation tool (TIST) in HR-pQCT images" by: "A. Valentinitsch1, L. Fischer2, M. D. DiFranco2, J. M. Patsch2, G. M. Gruber2, F. Kainberger2, G. Langs2; 1Wien/AT, 2Vienna/AT"
In the last decade, industry’s demand for deep learning (DL) has increased due to its high perfor... more In the last decade, industry’s demand for deep learning (DL) has increased due to its high performance in complex scenarios. Due to the DL method’s complexity, experts and non-experts rely on blackbox software packages such as Tensorflow and Pytorch. The frameworks are constantly improving, and new versions are released frequently. As a natural process in software development, the released versions contain improvements/changes in the methods and their implementation. Moreover, versions may be bug-polluted, leading to the model performance decreasing or stopping the model from working. The aforementioned changes in implementation can lead to variance in obtained results. This work investigates the effect of implementation changes in different major releases of these frameworks on the model performance. We perform our study using a variety of standard datasets. Our study shows that users should consider that changing the framework version can affect the model performance. Moreover, th...
In order to develop machine learning and deep learning models that take into account the guidelin... more In order to develop machine learning and deep learning models that take into account the guidelines and principles of trustworthy AI, a novel information theoretic approach is introduced in this article. A unified approach to privacy-preserving interpretable and transferable learning is considered for studying and optimizing the trade-offs between the privacy, interpretability, and transferability aspects of trustworthy AI. A variational membership-mapping Bayesian model is used for the analytical approximation of the defined information theoretic measures for privacy leakage, interpretability, and transferability. The approach consists of approximating the information theoretic measures by maximizing a lower-bound using variational optimization. The approach is demonstrated through numerous experiments on benchmark datasets and a real-world biomedical application concerned with the detection of mental stress in individuals using heart rate variability analysis.
The main challenges along with lessons learned from ongoing research in the application of machin... more The main challenges along with lessons learned from ongoing research in the application of machine learning systems in practice are discussed, taking into account aspects of theoretical foundations, systems engineering, and human-centered AI postulates. The analysis outlines a fundamental theory-practice gap which superimposes the challenges of AI system engineering at the level of data quality assurance, model building, software engineering and deployment.
Illustration of the cross-sectional cut planes for Âą1SD of the mean shape models. SD: Standard d... more Illustration of the cross-sectional cut planes for Âą1SD of the mean shape models. SD: Standard deviation; Distal: Distal plane; Middle: Middle plane; Proximal: Proximal plane. (TIF 9568 kb)
Zsfassung in dt. SpracheIn den letzten Jahren lieferten Segmentierungsansätze basierend auf seque... more Zsfassung in dt. SpracheIn den letzten Jahren lieferten Segmentierungsansätze basierend auf sequenzielle Monte Carlo Methoden vielversprechende Ergebnisse bei der Lokalisierung und Beschreibung anatomischer Strukturen in medizinisch relevanten Bildern. Auch bekannt unter der Bezeichnung Shape Particle Filter wurden diese Methoden für die Segmentierung von Wirbelkörpern, Lungenflügeln und Herzen eingesetzt. Ihr großer Vorteil liegt darin, dass sie auch bei Bildern mit starkem Rauschen wie zum Beispiel MR Aufnahmen, sowie bei sich überlagernden Strukturen, bei denen eine eindeutige Unterscheidung der Objekte schwierig ist, noch sehr gute Segmentierungsergebnisse liefern. Shape Particle Filter benötigen eine Maske, welche auf dem Mean Shape eines Shape Models basiert und in existierenden Implementierungen immer manuell definiert wird. Während der Suche nach einem Objekt wird die Wahrscheinlichkeit für jeden Pixel zu einer gewissen Region der Maske zu gehören durch das Klassifizieren vo...
Animated illustrations of the first five modes of all radius models. A: Female left radii model; ... more Animated illustrations of the first five modes of all radius models. A: Female left radii model; B: Female right radii model; C: Male left radii model; D: Male right radii model. (ZIP 9547 kb)
Poster: "ECR 2010 / B-566 / Towards automatic medical image segmentation using shape particl... more Poster: "ECR 2010 / B-566 / Towards automatic medical image segmentation using shape particle filters" by: "L. Fischer, R. Donner, F. Kainberger, G. Langs; Vienna/AT"
Uploads
Papers by Lukas Fischer