A5071573075- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Medical Imaging Techniques and Applications
- Advanced X-ray and CT Imaging
- Lung Cancer Diagnosis and Treatment
- Digital Imaging for Blood Diseases
- Brain Tumor Detection and Classification
- Prostate Cancer Treatment and Research
- Gastric Cancer Management and Outcomes
- Artificial Intelligence in Healthcare and Education
- Glioma Diagnosis and Treatment
- Image Processing and 3D Reconstruction
- Domain Adaptation and Few-Shot Learning
- Generative Adversarial Networks and Image Synthesis
Medical University of Vienna
2020-2025
Georgia Institute of Technology
2022-2023
Abstract Purpose Risk classification of primary prostate cancer in clinical routine is mainly based on prostate-specific antigen (PSA) levels, Gleason scores from biopsy samples, and tumor-nodes-metastasis (TNM) staging. This study aimed to investigate the diagnostic performance positron emission tomography/magnetic resonance imaging (PET/MRI) vivo models for predicting low-vs-high lesion risk (LH) as well biochemical recurrence (BCR) overall patient (OPR) with machine learning. Methods...
Background: This study investigated the performance of ensemble learning holomic models for detection breast cancer, receptor status, proliferation rate, and molecular subtypes from [18F]FDG-PET/CT images with without incorporating data pre-processing algorithms. Additionally, machine (ML) were compared conventional analysis using standard uptake value lesion classification. Methods: A cohort 170 patients 173 cancer tumors (132 malignant, 38 benign) was examined [18F]FDG-PET/CT. Breast...
Artificial Intelligence (AI) approaches in clinical science require extensive data preprocessing (DP) steps prior to building AI models. Establishing DP pipelines is a non-trivial task, mainly driven by purely mathematical rules and done scientists. Nevertheless, clinician presence shall be paramount at this step. The study proposes approach domain knowledge, where input, form of explicit non-explicit rules, directly impacts the algorithms' decision-making processes, thus, making planning...
Abstract Background Hybrid imaging became an instrumental part of medical imaging, particularly cancer processes in clinical routine. To date, several radiomic and machine learning studies investigated the feasibility vivo tumor characterization with variable outcomes. This study aims to investigate effect recently proposed fuzzy radiomics compare its predictive performance conventional cohorts. In addition, lesion vs. lesion+surrounding analysis was conducted. Methods Previously published...
Introduction Amino-acid positron emission tomography (PET) is a validated metabolic imaging approach for the diagnostic work-up of gliomas. This study aimed to evaluate sex-specific radiomic characteristics L-[S-methyl- 11 Cmethionine (MET)-PET images glioma patients in consideration prognostically relevant biomarker isocitrate dehydrogenase (IDH) mutation status. Methods MET-PET 35 astrocytic gliomas (13 females, mean age 41 ± 13 yrs. and 22 males, 46 17 yrs.) known IDH status were...
Background This study proposes machine learning-driven data preparation (MLDP) for optimal (DP) prior to building prediction models cancer cohorts. Methods A collection of well-established DP methods were incorporated the pipelines various clinical cohorts learning. Evolutionary algorithm principles combined with hyperparameter optimization employed iteratively select best fitting subset algorithms given dataset. The proposed method was validated glioma and prostate single center by 100-fold...
Recent model merging methods demonstrate that the parameters of fully-finetuned models specializing in distinct tasks can be combined into one capable solving all without retraining. Yet, this success does not transfer well when LoRA finetuned models. We study phenomenon and observe weights showcase a lower degree alignment compared to their counterparts. hypothesize improving is key obtaining better merges, propose KnOTS address problem. uses SVD jointly transform different an aligned...
Abstract Background Hybrid imaging became an instrumental part of medical imaging, particularly cancer processes in clinical routine. To date, several radiomic and machine learning studies investigated the feasibility vivo tumor characterization with variable outcomes. This study aims to investigate effect recently proposed fuzzy radiomics compare its predictive performance conventional cohorts. In addition, lesion vs. + surrounding analysis was conducted. Methods Previously published 11C...
Abstract Modern artificial intelligence (AI) approaches mainly rely on neural network (NN) or deep NN methodologies. However, these require large amounts of data to train, given, that the number their trainable parameters has a polynomial relationship neuron counts. This property renders not applicable in fields operating with small, albeit representative datasets such as healthcare. In this paper, we propose novel architecture which trains spatial positions soma and axon pairs, where...
Modern artificial intelligence (AI) approaches mainly rely on neural network (NN) or deep NN methodologies. However, these require large amounts of data to train, given, that the number their trainable parameters has a polynomial relationship neuron counts. This property renders NNs challenging apply in fields operating with small, albeit representative datasets such as healthcare. In this paper, we propose novel architecture which trains spatial positions soma and axon pairs, where weights...
Synthetic data (SIM) drawn from simulators have emerged as a popular alternative for training models where acquiring annotated real-world images is difficult. However, transferring trained on synthetic to applications can be challenging due appearance disparities. A commonly employed solution counter this SIM2REAL gap unsupervised domain adaptation, are using labeled SIM and unlabeled REAL data. Mispredictions made by such adapted often associated with miscalibration - stemming overconfident...
Abstract Purpose Risk stratification in patients with lung adenocarcinoma (LUAD) is mandatory for treatment guiding and outcome prediction. Amongst clinical parameters including histological analyses, imaging procedures provide important information. The present study aimed to investigate the ability of machine learning models trained on 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([ 18 F]FDG) positron emission tomography/computed tomography (PET/CT) derived radiomic data predict overall survival (OS),...