A5065893131- COVID-19 diagnosis using AI
- Radiomics and Machine Learning in Medical Imaging
- Lung Cancer Diagnosis and Treatment
- Artificial Intelligence in Healthcare and Education
- Mineral Processing and Grinding
- Digital Media Forensic Detection
- Advanced Neural Network Applications
- Electrical and Bioimpedance Tomography
- Advanced X-ray and CT Imaging
- Advanced Steganography and Watermarking Techniques
- Medical Imaging Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- AI in cancer detection
- Geochemistry and Geologic Mapping
- Medical Imaging and Analysis
- Colorectal Cancer Screening and Detection
- Cell Image Analysis Techniques
- Artificial Intelligence in Healthcare
- Soil Geostatistics and Mapping
- Machine Learning in Healthcare
- Image Processing Techniques and Applications
Technical University of Darmstadt
2021-2023
German Cancer Research Center
2022-2023
Heidelberg University
2023
Fraunhofer Institute for Secure Information Technology
2019-2022
DKFZ-ZMBH Alliance
2022
Deep Learning (DL) has the potential to optimize machine learning in both scientific and clinical communities. However, greater expertise is required develop DL algorithms, variability of implementations hinders their reproducibility, translation, deployment. Here we present community-driven Generally Nuanced Framework (GaNDLF), with goal lowering these barriers. GaNDLF makes mechanism development, training, inference more stable, reproducible, interpretable, scalable, without requiring an...
The segmentation of pelvic fracture fragments in CT and X-ray images is crucial for trauma diagnosis, surgical planning, intraoperative guidance. However, accurately efficiently delineating the bone remains a significant challenge due to complex anatomy imaging limitations. PENGWIN challenge, organized as MICCAI 2024 satellite event, aimed advance automated by benchmarking state-of-the-art algorithms on these tasks. A diverse dataset 150 scans was collected from multiple clinical centers,...
M3d-CAM is an easy to use library for generating attention maps of CNN-based PyTorch models improving the interpretability model predictions humans. The can be generated with multiple methods like Guided Backpropagation, Grad-CAM, Grad-CAM and Grad-CAM++. These visualize regions in input data that influenced prediction most at a certain layer. Furthermore, supports 2D 3D task classification as well segmentation. A key feature also cases only single line code required making basically plug play.
Fake news have been a problem for multiple years now and in addition to this "fake images" that accompany them are becoming increasingly too. The aim of such fake images is back up the message itself make it appear authentic. For purpose, more as photo-montages used, which spliced from several images. This can be used defame people by putting unfavorable situations or other way around propaganda making important. In addition, montages may altered with noise manipulations an automatic...
Traditionally, segmentation algorithms require dense annotations for training, demanding significant annotation efforts, particularly within the 3D medical imaging field. Scribble-supervised learning emerges as a possible solution to this challenge, promising reduction in efforts when creating large-scale datasets. Recently, plethora of methods optimized from scribbles have been proposed, but so far failed position scribble beneficial alternative. We relate shortcoming two major issues: 1)...
MSPaCMAn is the recently developed workflow that does mineralogical quantification of individual particles using its histograms while considering effects partial volume artefacts in interphases at particle level detail. This paper demonstrates and validates new developments workflow, aiming to minimize user bias enhance accuracy MSPaCMAn. Here, firstly, deep learning method, namely ParticleSeg3D, employed distinguish from background. Secondly, particle's size shape information are considered...
Abstract Type 2 diabetes (T2D) is a chronic disease currently affecting around 500 million people worldwide with often severe health consequences. Yet, histopathological analyses are still inadequate to infer the glycaemic state of person based on morphological alterations linked impaired insulin secretion and β-cell failure in T2D. Giga-pixel microscopy can capture subtle changes, but data complexity exceeds human analysis capabilities. In response, we generated dataset pancreas whole-slide...
Fake news have been a problem for multiple years now and in addition to this “fake images” that accompany them are becoming increasingly too. The aim of such fake images is back up the message itself make it appear authentic. For purpose, more as photo-montages used, which spliced from several images. This can be used defame people by putting unfavorable situations or other way around propaganda making important. In addition, montages may altered with noise manipulations an automatic...
Minerals, metals, and plastics are indispensable for a functioning modern society. Yet, their supply is limited causing need optimizing ore extraction recuperation from recyclable materials.Typically, those processes must be meticulously adapted to the precise properties of processed materials. Advancing our understanding these materials thus vital can achieved by crushing them into particles micrometer size followed characterization. Current imaging approaches perform this analysis based on...
To counter the ever increasing flood of image forgeries in form spliced images social media and web general, we propose novel splicing localization CNN NoiseSeg. NoiseSeg fuses statistical CNN-based methods separate branches to leverage benefits both. Unique anomalies that can be identified by its coarse noise separation branch, fine-grained feature branch error level analysis all get combined a segmentation fusion head predict precise regions. Experiments on DSO-1, CASIAv2, DEFACTO, IMD2020...
Automatic segmentation of ground glass opacities and consolidations in chest computer tomography (CT) scans can potentially ease the burden radiologists during times high resource utilisation. However, deep learning models are not trusted clinical routine due to failing silently on out-of-distribution (OOD) data. We propose a lightweight OOD detection method that leverages Mahalanobis distance feature space seamlessly integrates into state-of-the-art pipelines. The simple approach even...
Automatic segmentation of lung lesions in computer tomography has the potential to ease burden clinicians during Covid-19 pandemic. Yet predictive deep learning models are not trusted clinical routine due failing silently out-of-distribution (OOD) data. We propose a lightweight OOD detection method that exploits Mahalanobis distance feature space. The proposed approach can be seamlessly integrated into state-of-the-art pipelines without requiring changes model architecture or training...