Embedded Files
Home • Publications • Software • Student offers • Funding • News • People
AI in Medical Applications - Software
Our research projects cover interdisciplinary cardiovascular and oncology applications, as well as assessment of patient's general health, for example the body composition analysis, based on the muscle and fat mass at the L3 vertebra, geometric deep learning approaches for the prediction of a liver resection zone for a specific patient cohort or the analysis of lung thrombi for assessment of acute pulmonary embolism.
GitHub of the AI in Medical Applications (AIMed) Group
L3BOCA - L3-Body Composition Analysis: Patient condition strongly influences treatment outcomes. The BMI is widely used but fails to distinguish muscle, bone, and fat, so we focus on CT-based body composition analysis that enables precise measurement of skeletal muscle mass (SMM), visceral (VAT), subcutaneous (SAT), and intramuscular adipose tissue (IMAT). Instead of manual segmentation, we developed an offline GUI with a single-step, multilabel segmentation network that produces 3 adipose tissue and 5 muscle masks from L3 CT slices. The method works with both conventional and Photon-Counting CT, validated using Dice, IoU, and pixel accuracy. See specifically: https://github.com/AIinMedicalApplications/L3BOCA
Our liver resection zone prediction within the LIZARD project: Liver resection zone prediction using image-based and geometric deep learning (DFG Project No. 547369510) enables patient-specific resection planning and is ongoing work focussing on liver cancer patients.
See specifically: https://github.com/AIinMedicalApplications/LIZARD-RandLA-Net
For deep learning-based segmentation of acute pulmonary embolism in thorax CT image data, we trained a network and the detailed description of the computation can be found in the article “Deep Learning-Based Segmentation of Acute Pulmonary Embolism in Cardiac CT images" See specifically: https://github.com/AIinMedicalApplications/APES_ThrombusSegmenter
Page updated
Google Sites
Report abuse