A5078167377- Radiomics and Machine Learning in Medical Imaging
- MRI in cancer diagnosis
- Ovarian cancer diagnosis and treatment
- Advanced X-ray and CT Imaging
- Medical Imaging Techniques and Applications
- Machine Learning and Data Classification
- Machine Learning and Algorithms
- Advanced MRI Techniques and Applications
- Brain Tumor Detection and Classification
- Adversarial Robustness in Machine Learning
- Computational Drug Discovery Methods
- Acute Ischemic Stroke Management
- AI in cancer detection
- Medical Image Segmentation Techniques
- Glioma Diagnosis and Treatment
- Advanced Neuroimaging Techniques and Applications
- Optical Imaging and Spectroscopy Techniques
- Cardiovascular Function and Risk Factors
- Cell Image Analysis Techniques
- Multiple Sclerosis Research Studies
- Intraperitoneal and Appendiceal Malignancies
- Neurological disorders and treatments
- Medical Imaging and Analysis
- Advanced Computing and Algorithms
- Functional Brain Connectivity Studies
University Medical Center Hamburg-Eppendorf
2023-2025
Universität Hamburg
2023-2025
Jungheinrich (Germany)
2024
University of Cambridge
2022-2024
Uncertainty quantification in automated image analysis is highly desired many applications. Typically, machine learning models classification or segmentation are only developed to provide binary answers; however, quantifying the uncertainty of can play a critical role for example active human interaction. especially difficult when using deep learning-based models, which state-of-the-art imaging The current approaches do not scale well high-dimensional real-world problems. Scalable solutions...
This study evaluated the potential to reduce scan duration in dopamine transporter (DAT) SPECT when using a second-generation multiple-pinhole (MPH) collimator designed for brain with improved count sensitivity and spatial resolution compared parallel-hole fanbeam collimators. <b>Methods:</b> The retrospective included 640 consecutive clinical DAT studies that had been acquired list mode triple-head system MPH collimators 30-min net after injection of 181 ± 10 MBq [<sup>123</sup>I]FP-CIT....
We propose strongly unrealistic data augmentation to improve the robustness of convolutional neural networks (CNNs) for automatic classification dopamine transporter SPECT against variability between sites and cameras.
Background Pathological response to neoadjuvant treatment for patients with high-grade serous ovarian carcinoma (HGSOC) is assessed using the chemotherapy score (CRS) omental tumor deposits. The main limitation of CRS that it requires surgical sampling after initial (NACT) treatment. Earlier and non-invasive predictors could improve patient stratification. We developed computed tomography (CT) radiomic measures predict before NACT as a gold standard. Methods Omental CT-based radiomics...
Governments and medical associations across the world, including US Food Drug Administration, UK Medicines Healthcare products Regulatory Agency, Royal College of Radiologists, European Society Radiology, believe advent health technologies associated with artificial intelligence (AI) will be most radical change in how care is delivered our lifetime.1Royal RadiologistsRCR position statement on intelligence.http://www.rcr.ac.uk/posts/rcr-position-statement-artificial-intelligenceDate: 2018Date...
Abstract Purpose To determine if pelvic/ovarian and omental lesions of ovarian cancer can be reliably segmented on computed tomography (CT) using fully automated deep learning-based methods. Methods A learning model for the two most common disease sites high-grade serous (pelvis/ovaries omentum) was developed compared against well-established “no-new-Net” framework unrevised trainee radiologist segmentations. total 451 CT scans collected from four different institutions were used training (...
Abstract Purpose MRI-derived brain volume loss (BVL) is widely used as neurodegeneration marker. SIENA state-of-the-art for BVL measurement, but limited by long computation time. Here we propose “BrainLossNet”, a convolutional neural network (CNN)-based method BVL-estimation. Methods BrainLossNet uses CNN-based non-linear registration of baseline(BL)/follow-up(FU) 3D-T1w-MRI pairs. computed parenchyma masks segmented in the BL/FU scans. The estimate corrected image distortions using apparent...
Background: Evaluating AI-based segmentation models primarily relies on quantitative metrics, but it remains unclear if this approach leads to practical, clinically-applicable tools. Purpose: To create a systematic framework for evaluating the performance of using clinically relevant criteria. Materials and Methods: We developed AUGMENT (Assessing Utility seGMENtation Tools), based structured classification main categories error in tasks. evaluate we assembled team 20 clinicians covering...
Abstract Purpose Deep convolutional neural networks (CNN) hold promise for assisting the interpretation of dopamine transporter (DAT)-SPECT. For improved communication uncertainty to user it is crucial reliably discriminate certain from inconclusive cases that might be misclassified by strict application a predefined decision threshold on CNN output. This study tested two methods incorporate existing label during training improve utility sigmoid output this task. Methods Three datasets were...
Abstract Purpose To determine if pelvic/ovarian and omental lesions of ovarian cancer can be reliably segmented on computed tomography (CT) using fully automated deep learning-based methods. Materials Methods A learning model for the two most common disease sites high grade serous (pelvis/ovaries omentum) was developed compared against well-established “no-new-Net” (nnU-Net) framework unrevised trainee radiologist segmentations. total 451 pre-treatment post neoadjuvant chemotherapy (NACT) CT...
Artificial intelligence (AI) methods applied to healthcare problems have shown enormous potential alleviate the burden of health services worldwide and improve accuracy reproducibility predictions. In particular, developments in computer vision are creating a paradigm shift analysis radiological images, where AI tools already capable automatically detecting precisely delineating tumours. However, such generally developed technical departments that continue be siloed from real benefit would...
Clinical validation of "BrainLossNet", a deep learning-based method for fast and robust estimation brain volume loss (BVL) from longitudinal T1-weighted MRI, the detection accelerated BVL in multiple sclerosis (MS) discrimination between MS patients with versus without disability progression.
Medical image analysis tasks often focus on regions or structures located in a particular location within the patient's body. Often large parts of may not be interest for task. When using deep-learning based approaches, this causes an unnecessary increases computational burden during inference and raises chance errors. In paper, we introduce CTARR, novel generic method CT Anatomical Region Recognition. The serves as pre-processing step any deep learning-based pipeline by automatically...
Objective: High-grade serous ovarian cancer (HGSOC) is the most lethal gynaecological with patients routinely diagnosed at advanced stages widespread disease. Evidence from screening trials indicates that early diagnosis may not reduce cancer-related deaths, possibly due to an underestimation of true extent disease screening. We aim characterise growth kinetics HGSOC understand why detection has failed so far and under what conditions it might prove fruitful. Methods Analysis: analysed a...
<title>Abstract</title> Objective To provide fully automatic scanner-independent 5-level categorization of the [<sup>123</sup>I]FP-CIT uptake in striatal subregions dopamine transporter SPECT. Methods A total 3,500 SPECT scans from two house (n = 1,740, n 640) and external 645, 475) datasets were used for this study. single convolutional neural network (CNN) was trained unilateral caudate putamen both hemispheres according to 5 levels: normal, borderline, moderate reduction, strong almost...
This study introduces an automated pipeline for renal cancer (RC) detection in non-contrast computed tomography (NCCT). In the development of our pipeline, we test three detections models: a shape model, 2D-, and 3D axial-sample model. Training (n=1348) testing (n=64) data were gathered from open sources (KiTS23, Abdomen1k, CT-ORG) Cambridge University Hospital (CUH). Results cross-validation revealed that 2D axial sample model had highest small ($\leq$40mm diameter) RC area under curve...
Uncertainty quantification in automated image analysis is highly desired many applications. Typically, machine learning models classification or segmentation are only developed to provide binary answers; however, quantifying the uncertainty of can play a critical role for example active human interaction. especially difficult when using deep learning-based models, which state-of-the-art imaging The current approaches do not scale well high-dimensional real-world problems. Scalable solutions...