A5027935660- Radiomics and Machine Learning in Medical Imaging
- Pancreatic and Hepatic Oncology Research
- Advanced X-ray and CT Imaging
- Artificial Intelligence in Healthcare and Education
- MRI in cancer diagnosis
- Cancer Genomics and Diagnostics
- Lung Cancer Diagnosis and Treatment
- Colorectal Cancer Screening and Detection
- Topic Modeling
- Medical Imaging Techniques and Applications
- Radiation Dose and Imaging
- Renal cell carcinoma treatment
- COVID-19 diagnosis using AI
- AI in cancer detection
- Venous Thromboembolism Diagnosis and Management
- Pulmonary Hypertension Research and Treatments
- Biomedical Text Mining and Ontologies
- Atomic and Subatomic Physics Research
- Photoacoustic and Ultrasonic Imaging
- Hepatocellular Carcinoma Treatment and Prognosis
- Dental Radiography and Imaging
- Diverticular Disease and Complications
- Neuroendocrine Tumor Research Advances
- Advanced MRI Techniques and Applications
- Colorectal Cancer Surgical Treatments
Klinikum rechts der Isar
2018-2025
Technical University of Munich
2019-2025
Development of a supervised machine-learning model capable predicting clinically relevant molecular subtypes pancreatic ductal adenocarcinoma (PDAC) from diffusion-weighted-imaging-derived radiomic features.The retrospective observational study assessed 55 surgical PDAC patients. Molecular were defined by immunohistochemical staining KRT81. Tumors manually segmented and 1606 features extracted with PyRadiomics. A gradient-boosted-tree algorithm was trained on 70% the patients (N = 28) tested...
Abstract Background To develop a supervised machine learning (ML) algorithm predicting above- versus below-median overall survival (OS) from diffusion-weighted imaging-derived radiomic features in patients with pancreatic ductal adenocarcinoma (PDAC). Methods One hundred two histopathologically proven PDAC were retrospectively assessed as training cohort, and 30 prospectively accrued enrolled served independent validation cohort (IVC). Tumors segmented on preoperative apparent diffusion...
To bridge the translational gap between recent discoveries of distinct molecular phenotypes pancreatic cancer and tangible improvements in patient outcome, there is an urgent need to develop strategies tools informing improving clinical decision process. Radiomics machine learning approaches can offer non-invasive whole tumor analytics for imaging data-based classification. The retrospective study assessed baseline computed tomography (CT) from 207 patients with proven ductal adenocarcinoma...
The differentiation of autoimmune pancreatitis (AIP) and pancreatic ductal adenocarcinoma (PDAC) poses a relevant diagnostic challenge can lead to misdiagnosis consequently poor patient outcome. Recent studies have shown that radiomics-based models achieve high sensitivity specificity in predicting both entities. However, radiomic features only capture low level representations the input image. In contrast, convolutional neural networks (CNNs) learn extract more complex which been used for...
Lung cancer screening is already implemented in the USA and strongly recommended by European Radiological Thoracic societies as well. Upon implementation, total number of thoracic computed tomographies (CT) likely to rise significantly. As shown previous studies, modern artificial intelligence-based algorithms are on-par or even exceed radiologist's performance lung nodule detection classification. Therefore, aim this study was evaluate cost-effectiveness an AI-based system context baseline...
The purpose of this retrospective study was to evaluate the value contrast-enhanced computed tomography (CE-CT) image features at baseline and after neoadjuvant chemotherapy in predicting histopathological response patients with adenocarcinoma gastroesophageal junction (GEJ). A total 105 a diagnosis GEJ were examined by CE-CT preoperatively chemotherapy. All underwent surgical resection. Histopathological parameters tumor regression grading according Becker et al. collected 93 patients. Line...
Purpose To evaluate the performance of LLMs in extracting data from stroke CT reports presence and absence an annotation guideline. Methods In this study, GPT-4o Llama-3.3-70B ten imaging findings was assessed two datasets a single academic center. Dataset A (n = 200) stratified cohort including various pathological findings, whereas B 100) consecutive cohort. Initially, guideline providing clear extraction instructions designed based on review cases with inter-annotator disagreements...
Abstract This study aims to investigate the feasibility, usability, and effectiveness of a Retrieval-Augmented Generation (RAG)-powered Patient Information Assistant (PIA) chatbot for pre-CT information counseling compared standard physician consultation informed consent process. prospective comparative included 86 patients scheduled CT imaging between November December 2024. Patients were randomly assigned either PIA group ( n = 43), who received via chat app, or control with doctor-led...
Purpose To evaluate the potential of LLMs to generate sequence-level brain MRI protocols. Methods A dataset 150 cases was derived from imaging request forms obtained local institution. For each case, a reference protocol established by two board-certified neuroradiologists, with discrepancies resolved through consensus. GPT-4o, o3-mini, DeepSeek-R1 and Qwen-2.5-72B were employed protocols based on case descriptions. model, generation conducted under conditions: 1) additional in-context...
To assess the performance of prospectively accelerated and deep learning (DL) reconstructed T2-weighted (T2w) imaging in volunteers patients with histologically proven prostate cancer (PCa).Prospectively undersampled T2w datasets were acquired acceleration factors 1.7 (reference), 3.4 4.8 10 healthy 23 PCa. Image reconstructions using compressed SENSE (C-SENSE) a combination C-SENSE DL-based artificial intelligence (C-SENSE AI) analyzed. Qualitative image comparison was performed 6-point...
Exploring the generative capabilities of multimodal GPT-4, our study uncovered significant differences between radiological assessments and automatic evaluation metrics for chest x-ray impression generation revealed bias.
Importance Differentiating between malignant and benign etiology in large-bowel wall thickening on computed tomography (CT) images can be a challenging task. Artificial intelligence (AI) support systems improve the diagnostic accuracy of radiologists, as shown for variety imaging tasks. Improvements performance, particular reduction false-negative findings, may useful patient care. Objective To develop evaluate deep learning algorithm able to differentiate colon carcinoma (CC) acute...
Abstract We present a method to generate synthetic thorax radiographs with realistic nodules from CT scans, and perfect ground truth knowledge. evaluated the detection performance of nine radiologists two convolutional neural networks in reader study. Nodules were artificially inserted into lung volume obtained by forward-projecting volume. Hence, our framework allowed for detailed evaluation CAD systems’ radiologists’ due availability accurate ground-truth labels data. Radiographs network...
Imaging phantoms were scanned twice on 3 computed tomography scanners from 2 different manufactures with varying tube voltages and currents. Phantoms segmented, features extracted using PyRadiomics a pretrained CNN. After standardization the concordance correlation coefficient (CCC), mean feature variance, range, of variant calculated to assess robustness. In addition, cosine similarity was for vectorized activation maps an exemplary phantom. For in vivo comparison, radiomics CNN 30 patients...
Especially patients with aortic aneurysms and multiple computed tomography angiographies (CTA) might show medical conditions which oppose the use of iodine-based contrast agents. CTA using monoenergetic reconstructions from dual layer CT gadolinium (Gd-)based agents be a feasible alternative in these patients. Therefore, purpose this study was to evaluate feasibility clinical spectral Gd-based agent aneurysms.
Pancreatic ductal adenocarcinoma (PDAC) remains a tumor entity of exceptionally poor prognosis, and several biomarkers are under current investigation for the prediction patient prognosis. Many studies focus on promoting newly developed imaging without rigorous comparison to other established parameters. To assess true value leverage potential all efforts in this field, multi-parametric evaluation available PDAC survival is warranted. Here we present multiparametric analysis predictive...
The success of deep learning in recent years has arguably been driven by the availability large datasets for training powerful predictive algorithms. In medical applications however, sensitive nature data limits collection and exchange large-scale datasets. Privacy-preserving collaborative systems can enable successful application machine medicine. However, protocols such as federated require frequent transfer parameter updates over a network. To deployment to wide range with varying...
Abstract Purpose In this prospective exploratory study, we evaluated the feasibility of [ 18 F]fluorodeoxyglucose ([ F]FDG) PET/MRI-based chemotherapy response prediction in pancreatic ductal adenocarcinoma at two weeks upon therapy onset. Material and methods a mixed cohort, seventeen patients treated with neoadjuvant or palliative intent were enrolled. All imaged by F]FDG PET/MRI before after onset chemotherapy. Response per RECIST1.1 was then assessed 3 months PET/MRI-derived parameters...
To evaluate the perception of different types AI-based assistance and interaction radiologists with algorithm's predictions certainty measures.In this retrospective observer study, four were asked to classify Breast Imaging-Reporting Data System 4 (BI-RADS4) lesions (n = 101 benign, n 99 malignant). The effect (occlusion-based interpretability map, classification, certainty) on radiologists' performance (sensitivity, specificity, questionnaire) measured. influence Big Five personality traits...
Background: PDAC remains a tumor entity with poor prognosis and 5-year survival rate below 10%. Recent research has revealed invasive biomarkers, such as distinct molecular subtypes, predictive for therapy response patient survival. Non-invasive prediction of individual outcome however an unresolved task. Methods: Discrete cellularity regions resection specimen (n = 43) were analyzed by routine histopathological work up. Regional CT-derived Hounsfield Units (HU, n 66) well iodine...
Abstract Purpose Development of a supervised machine-learning model capable predicting clinically relevant molecular subtypes pancreatic ductal adenocarcinoma (PDAC) from diffusion-weighted-imaging-derived radiomic features. Methods The retrospective observational study assessed 55 surgical PDAC patients. Molecular were defined by immunohistochemical staining KRT81. Tumors manually segmented and 1606 features extracted with PyRadiomics . A gradient-boosted-tree algorithm (XGBoost) was...