A5049450234- Radiomics and Machine Learning in Medical Imaging
- Advanced X-ray and CT Imaging
- Medical Imaging Techniques and Applications
- AI in cancer detection
- Artificial Intelligence in Healthcare and Education
- Acute Ischemic Stroke Management
- Colorectal Cancer Screening and Detection
- Microscopic Colitis
- Inflammatory Bowel Disease
- Radiation Dose and Imaging
- Venous Thromboembolism Diagnosis and Management
- Sarcoma Diagnosis and Treatment
- Bone Tumor Diagnosis and Treatments
- Cardiac Imaging and Diagnostics
- Traumatic Brain Injury and Neurovascular Disturbances
- Orthopedic Infections and Treatments
- Diagnosis and treatment of tuberculosis
- Hepatocellular Carcinoma Treatment and Prognosis
- MRI in cancer diagnosis
- Advanced Multi-Objective Optimization Algorithms
- Lung Cancer Diagnosis and Treatment
- Colorectal Cancer Surgical Treatments
- Glioma Diagnosis and Treatment
- COVID-19 diagnosis using AI
- Infective Endocarditis Diagnosis and Management
Essen University Hospital
2018-2025
University of Duisburg-Essen
2018-2024
Institut für Medizinische Informatik, Biometrie und Epidemiologie
2019-2022
Ruhr University Bochum
2016
University of Tübingen
2014
Abstract Purpose To propose a new quality scoring tool, METhodological RadiomICs Score (METRICS), to assess and improve research of radiomics studies. Methods We conducted an online modified Delphi study with group international experts. It was performed in three consecutive stages: Stage#1, item preparation; Stage#2, panel discussion among EuSoMII Auditing Group members identify the items be voted; Stage#3, four rounds exercise by panelists determine eligible for METRICS their weights. The...
Abstract Objectives In radiomics, different feature normalization methods, such as z-Score or Min–Max, are currently utilized, but their specific impact on the model is unclear. We aimed to measure effect predictive performance and selection. Methods employed fifteen publicly available radiomics datasets compare seven methods. Using four selection classifier we used cross-validation area under curve (AUC) of resulting models, agreement selected features, calibration. addition, assessed...
Our purpose was to investigate differences between PET/MRI and PET/CT in lesion detection classification oncologic whole-body examinations radiation exposure the 2 modalities. <b>Methods:</b> In this observational single-center study, 1,003 (918 patients; mean age, 57.8 ± 14.4 y) were included. Patients underwent subsequent (149.8 49.7 min after tracer administration). Examinations reviewed by radiologists nuclear medicine physicians consensus. Additional findings, characterization of...
Abstract Background Many studies in radiomics are using feature selection methods to identify the most predictive features. At same time, they employ cross-validation estimate performance of developed models. However, if is performed before cross-validation, data leakage can occur, and results be biased. To measure extent this bias, we collected ten publicly available datasets conducted two experiments. First, models were by incorrectly applying prior cross-validation. Then, experiment was...
Objectives A critical problem in radiomic studies is the high dimensionality of datasets, which stems from small sample sizes and many generic features extracted volume interest. Therefore, feature selection methods are used, aim to remove redundant as well irrelevant features. Because there algorithms, it key understand their performance context radiomics. Materials Methods total 29 algorithms 10 classifiers were evaluated on publicly available datasets. Feature compared for training times,...
Radiomics is a noninvasive method using machine learning to support personalised medicine. Preprocessing filters such as wavelet and Laplacian-of-Gaussian are commonly used being thought increase predictive performance. However, the use of preprocessing increases number features by up an order magnitude can produce many correlated features. Both substantially dataset complexity, which in turn makes modeling with techniques more challenging, possibly leading poorer We investigated impact...
Abstract Radiomic datasets can be class-imbalanced, for instance, when the prevalence of diseases varies notably, meaning that number positive samples is much smaller than negative samples. In these cases, majority class may dominate model's training and thus negatively affect predictive performance, leading to bias. Therefore, resampling methods are often utilized class-balance data. However, several exist, neither their relative performance nor impact on feature selection has been...
Background Recently, radiomics has emerged as a non-invasive, imaging-based tissue characterization method in multiple cancer types. One limitation for robust and reproducible analysis lies the inter-reader variability of tumor annotations, which can potentially cause differences extracted feature sets results. In this study, diagnostic potential rapid clinically feasible VOI (Volume Interest)-based approach to is investigated assess MR-derived parameters predicting molecular subtype,...
In radiomic studies, several models are often trained with different combinations of feature selection methods and classifiers. The features the best model usually considered relevant to problem, they represent potential biomarkers. Features selected from statistically similarly performing generally not studied. To understand degree which these similar differ, 14 publicly available datasets, 8 methods, classifiers were used in this retrospective study. For each combination classifier, a was...
To improve organ protection with the frozen elephant trunk (FET) procedure, a so-called four-sites perfusion in combination proximalization for distal aortic anastomosis was performed. The impact of these techniques on patient outcome is reported.Between February 2005 and April 2020, total 357 patients underwent FET procedure acute (54%) or chronic (22%) dissection aneurysmal disease (24%). level defined according to arch zones 0-3. Patients were divided into 3 groups intraoperative...
Purpose The aim of this study is to assess the impact bridging intravenous thrombolysis (IVT), infarct core growth rate (ICGR) and their interaction on neurological outcomes in patients undergoing endovascular thrombectomy (EVT) acute ischemic stroke (AIS) with anterior large vessel occlusion (LVO). Methods Patients EVT due LVO (ICA M2 branches) between 2018 2022 a tertiary care center were included. Patient's baseline characteristics, peri-procedural factors retrospectively analyzed. ICGR...
Abstract Patients with cardioembolic stroke often undergo CT of the left atrial appendage (LAA), for example, to determine whether thrombi are present in LAA. To guide imaging process, technologists first perform a localizer scan, which is preliminary image used identify region interest. However, lack well-defined landmarks makes accurate delimitation LAA localizers difficult and requires whole-heart scans, increasing radiation exposure cancer risk. This study aims automate using deep...
Abstract Class imbalance is often unavoidable for radiomic data collected from clinical routine. It can create problems during classifier training since the majority class could dominate minority class. Consequently, resampling methods like oversampling or undersampling are applied to class-balance data. However, must not be upfront all because it would lead leakage and, therefore, erroneous results. This study aims measure extent of this bias. Five-fold cross-validation with 30 repeats was...
Background Detection of ossification areas hand bones in X-ray images is an important task, e.g. as a preprocessing step automated bone age estimation. Deep neural networks have emerged recently de facto standard detection methods, but their drawback the need large annotated datasets. Finetuning pre-trained viable alternative, it not clear priori if training with small datasets will be successful, depends on problem at hand. In this paper, we show that can utilized to produce effective...
Our objective was to define an <sup>18</sup>F-FDG PET/MR enterography index as a hybrid surrogate marker for active ileocolonic inflammation in Crohn’s disease (CD) and assess its diagnostic performance comparison validated MR indices (MR of activity [MaRIA], Clermont score). <b>Methods:</b> Fifty-two CD patients with recurrent symptoms underwent ileocolonoscopy enterography. Three hundred three segments were assessed using MaRIA the score well newly defined index. On basis tobit regression,...
Our purpose was to assess the diagnostic potential of simultaneously acquired 18F-FDG PET and MRI data sets for therapy response assessment isolated limb perfusion (ILP) in patients with soft-tissue sarcomas (STS). Methods: In total, 45 histopathologically verified STS were prospectively enrolled an integrated PET/MRI examination before after ILP. Therapy assessed based on different MRI- PET-derived morphologic (RECIST MR-adapted Choi criteria) metabolic (PERCIST) criteria. addition, a...
To develop and evaluate fully automatic scan range delimitation for chest CT by using deep learning.For this retrospective study, ranges were annotated two expert radiologists in consensus 1149 (mean age, 65 years ± 16 [standard deviation]; 595 male patients) topograms acquired between March 2002 February 2019 (350 with pleural effusion, 376 atelectasis, 409 neither, 14 both). A conditional generative adversarial neural network was trained on 1000 randomly selected to generate virtual...
Short tau inversion recovery (STIR) sequences are frequently used in magnetic resonance imaging (MRI) of the spine. However, STIR require a significant amount scanning time. The purpose present study was to generate virtual (vSTIR) images from non-contrast, non-fat-suppressed T1- and T2-weighted using conditional generative adversarial network (cGAN). training dataset comprised 612 studies 514 patients, validation 141 133 patients. For validation, 100 original respective vSTIR series were...
We aimed to investigate treatment effect of endovascular thrombectomy (EVT) on the change National Institutes Health Stroke Scale (NIHSS) scores in acute ischemic stroke (AIS) patients with anterior large vessel occlusion (LVO). Predictors early neurological improvement (ENI) were assessed those successful reperfusion. Data from January 2018 December 2020 retrospectively analyzed. Anterior LVO was defined as internal carotid artery and/or M1/M2 branch middle cerebral artery. A reduction at...
Limited treatment options in patients with intrahepatic cholangiocarcinoma (iCCA) demand the introduction of new, catheter-based options. Especially, <sup>90</sup>Y radioembolization may expand therapeutic abilities beyond surgery or chemotherapy. Therefore, purpose this study was to identify factors associated an improved median overall survival (mOS) iCCA receiving a retrospective at 5 major tertiary-care centers. <b>Methods:</b> In total, 138 radioembolizations 128 (female, 47.7%; male,...
Abstract Objectives Cardiac computed tomography (CT) is essential in diagnosing coronary heart disease. However, a disadvantage the associated radiation exposure to patient which depends part on scan range. This study aimed develop deep neural network optimize delimitation of ranges CT localizers reduce dose. Methods On retrospective training cohort 1507 randomly selected from calcium scoring and angiography scans acquired between 2010 2017, optimized were delimited by two radiologists...