A5078497529- Medical Imaging and Analysis
- Dental Radiography and Imaging
- Dental Implant Techniques and Outcomes
- Spectroscopy Techniques in Biomedical and Chemical Research
- Medical Imaging Techniques and Applications
- Remote Sensing and LiDAR Applications
- Radiomics and Machine Learning in Medical Imaging
- AI-based Problem Solving and Planning
- Advanced Chemical Sensor Technologies
- Cognitive Science and Education Research
- Cutaneous Melanoma Detection and Management
- Digital Imaging for Blood Diseases
- Advanced X-ray and CT Imaging
- Advanced Causal Inference Techniques
- Ethics and Social Impacts of AI
- Remote Sensing in Agriculture
- Advanced Neural Network Applications
- Forensic Anthropology and Bioarchaeology Studies
- AI in cancer detection
- Data Visualization and Analytics
- Explainable Artificial Intelligence (XAI)
- Shoulder Injury and Treatment
- Semantic Web and Ontologies
- Spectroscopy and Chemometric Analyses
University of Basel
2020-2024
Diagnosing and mitigating changes in model fairness under distribution shift is an important component of the safe deployment machine learning healthcare settings. Importantly, success any mitigation strategy strongly depends on structure shift. Despite this, there has been little discussion how to empirically assess a that one encountering practice. In this work, we adopt causal framing motivate conditional independence tests as key tool for characterizing shifts. Using our approach two...
Countless low-surface brightness objects - including spiral galaxies, dwarf and noise patterns have been detected in recent large surveys. Classically, astronomers visually inspect those detections to distinguish between real galaxies artefacts. Employing the Dark Energy Survey (DES) machine learning techniques, Tanoglidis et al. (2020) shown how this task can be automatically performed by computers. Here, we build upon their pioneering work further separate into spirals, ellipticals,...
Automated distinct bone segmentation from CT scans is widely used in planning and navigation workflows. U-Net variants are known to provide excellent results supervised semantic segmentation. However, upper-body CTs a large field of view computationally taxing 3D architecture required. This leads low-resolution lacking detail or localisation errors due missing spatial context when using high-resolution inputs.We propose solve this problem by end-to-end trainable networks that combine several...
During the diagnostic process, doctors incorporate multimodal information including imaging and medical history - similarly AI development has increasingly become multimodal. In this paper we tackle a more subtle challenge: take targeted to obtain only most pertinent pieces of information; how do enable same? We develop wrapper method named MINT (Make your model INTeractive) that automatically determines what are valuable at each step, ask for useful information. demonstrate efficacy...
In the above article <xref ref-type="bibr" rid="ref1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</xref> , we found major issues with data used. Specifically, received from rid="ref2" xmlns:xlink="http://www.w3.org/1999/xlink">[2]</xref> for five distinct tissues, ten specimens per tissue. However, upon closer examination, realized that these were not unique; rather, they scaled variations derived a single specimen. As result, our training,...
While multimodal foundation models can now natively work with data beyond text, they remain underutilized in analyzing the considerable amounts of multi-dimensional time-series fields like healthcare, finance, and social sciences, representing a missed opportunity for richer, data-driven insights. This paper proposes simple but effective method that leverages existing vision encoders these to "see" via plots, avoiding need additional, potentially costly, model training. Our empirical...
Today's mechanical tools for bone cutting (osteotomy) lead to trauma that prolong the healing process. Medical device manufacturers continuously strive improve their minimize such trauma. One example of a new tool and procedure is minimally invasive surgery with laser as element. This setup allows tissue ablation using light instead tools, which reduces post-surgery time. During surgery, reliable feedback system crucial avoid collateral damage surrounding tissues. Therefore, we propose...
Abstract Purpose: Automated distinct bone segmentation has many applications in planning and navigation tasks. 3D U-Nets have previously been used to segment bones the upper body, but their performance is not yet optimal. Their most substantial source of error lies confusing one for another, background with bone-tissue. Methods: In this work, we propose binary-prediction-enhanced multi-class (BEM) inference, which takes into account an additional binary background/bone-tissue prediction,...
Using a laser for cutting bones instead of the traditional saws has been shown to improve patient's healing process. Additionally, potential reduce collateral damage surrounding tissue if appropriately applied. This can be achieved by building additional sensing elements besides itself into an endoscope. To this end, we use microsecond pulsed Erbium-doped Yttrium Aluminium Garnet (Er:YAG) cut bones. During ablation, each pulse emits acoustic shock wave that is captured air-coupled...
Purpose: Automated distinct bone segmentation from CT scans is widely used in planning and navigation workflows. U-Net variants are known to provide excellent results supervised semantic segmentation. However, upper body CTs a large field of view computationally taxing 3D architecture required. This leads low-resolution lacking detail or localisation errors due missing spatial context when using high-resolution inputs. Methods: We propose solve this problem by end-to-end trainable networks...
Purpose: The localisation and segmentation of individual bones is an important preprocessing step in many planning navigation applications. It is, however, a time-consuming repetitive task if done manually. This true not only for clinical practice but also the acquisition training data. We therefore present end-to-end learnt algorithm that capable segmenting 125 distinct upper-body CT, provide ensemble-based uncertainty measure helps to single out scans enlarge dataset with. Methods create...