A5101665089- Medical Imaging Techniques and Applications
- Radiomics and Machine Learning in Medical Imaging
- Advanced X-ray and CT Imaging
- Radiation Detection and Scintillator Technologies
- COVID-19 diagnosis using AI
- Advanced MRI Techniques and Applications
- Advanced Radiotherapy Techniques
- Radiation Therapy and Dosimetry
- Lung Cancer Diagnosis and Treatment
- AI in cancer detection
- Cell Image Analysis Techniques
- Radiation Dose and Imaging
- Digital Radiography and Breast Imaging
- Medical Image Segmentation Techniques
- Domain Adaptation and Few-Shot Learning
- Image and Signal Denoising Methods
- Nuclear Physics and Applications
- Advanced Neural Network Applications
- Machine Learning and Algorithms
- Brain Tumor Detection and Classification
- Dementia and Cognitive Impairment Research
- Acute Ischemic Stroke Management
- Lanthanide and Transition Metal Complexes
- Machine Learning in Materials Science
- MRI in cancer diagnosis
University Hospital of Geneva
2019-2025
University of Geneva
2022-2023
Geneva College
2021-2022
Karolinska Institutet
2021
Amirkabir University of Technology
2019
Tehran University of Medical Sciences
2019
Tendency is to moderate the injected activity and/or reduce acquisition time in PET examinations minimize potential radiation hazards and increase patient comfort. This work aims assess performance of regular full-dose (FD) synthesis from fast/low-dose (LD) whole-body (WB) images using deep learning techniques.
Our purpose was to assess the performance of full-dose (FD) PET image synthesis in both and sinogram space from low-dose (LD) images sinograms without sacrificing diagnostic quality using deep learning techniques. Methods: Clinical brain PET/CT studies 140 patients were retrospectively used for LD-to-FD conversion. Five percent events randomly selected FD list-mode data simulate a realistic LD acquisition. A modified 3-dimensional U-Net model implemented predict projection (PSS) (PIS) their...
Abstract Objectives The current study aimed to design an ultra-low-dose CT examination protocol using a deep learning approach suitable for clinical diagnosis of COVID-19 patients. Methods In this study, 800, 170, and 171 pairs full-dose images were used as input/output training, test, external validation set, respectively, implement the prediction technique. A residual convolutional neural network was applied generate from images. quality predicted assessed root mean square error (RMSE),...
We aimed to analyze the prognostic power of CT-based radiomics models using data 14,339 COVID-19 patients.
Purpose The generalizability and trustworthiness of deep learning (DL)–based algorithms depend on the size heterogeneity training datasets. However, because patient privacy concerns ethical legal issues, sharing medical images between different centers is restricted. Our objective to build a federated DL-based framework for PET image segmentation utilizing multicentric dataset compare its performance with centralized DL approach. Methods from 405 head neck cancer patients 9 formed basis this...
Attenuation correction and scatter compensation (AC/SC) are two main steps toward quantitative PET imaging, which remain challenging in PET-only PET/MRI systems. These can be effectively tackled via deep learning (DL) methods. However, trustworthy, generalizable DL models commonly require well-curated, heterogeneous, large datasets from multiple clinical centers. At the same time, owing to legal/ethical issues privacy concerns, forming a collective, centralized dataset poses significant...
Purpose The availability of automated, accurate, and robust gross tumor volume (GTV) segmentation algorithms is critical for the management head neck cancer (HNC) patients. In this work, we evaluated 3 state-of-the-art deep learning combined with 8 different loss functions PET image using a comprehensive training set its performance on an external validation HNC Patients Methods 18 F-FDG PET/CT images 470 patients presenting which manually defined GTVs serving as standard reference were used...
Abstract Background Despite the prevalence of chest CT in clinic, concerns about unoptimized protocols delivering high radiation doses to patients still remain. This study aimed assess additional dose associated with overscanning and develop an automated deep learning-assisted scan range selection technique reduce patients. Results A significant (31 ± 24) mm was observed clinical setting for over 95% cases. The average Dice coefficient lung segmentation 0.96 0.97 anterior–posterior (AP)...
We aimed to construct a prediction model based on computed tomography (CT) radiomics features classify COVID-19 patients into severe-, moderate-, mild-, and non-pneumonic. A total of 1110 were studied from publicly available dataset with 4-class severity scoring performed by radiologist (based CT images clinical features). The entire lungs segmented followed resizing, bin discretization radiomic extraction. utilized two feature selection algorithms, namely bagging random forest (BRF)...
Contrast-enhanced computed tomography (CECT) provides much more information compared to non-enhanced CT images, especially for the differentiation of malignancies, such as liver carcinomas. Contrast media injection phase is usually missing on public datasets and not standardized in clinic even same region language. This a barrier effective use available CECT images clinical research.
The scintillation light distribution produced by photodetectors in positron emission tomography (PET) provides the depth of interaction (DOI) information required for high-resolution imaging. goal positioning techniques is to reverse photodetector signal’s pattern map coordinates incident photon energy position. By considering DOI information, monolithic crystals offer good spatial, energy, and timing resolution along with high sensitivity. In this work, a supervised deep neural network was...
We present a deep learning (DL)-based automated whole lung and COVID-19 pneumonia infectious lesions (COLI-Net) detection segmentation from chest computed tomography (CT) images. This multicenter/multiscanner study involved 2368 (347'259 2D slices) 190 (17 341 volumetric CT exams along with their corresponding manual of lungs lesions, respectively. All images were cropped, resized, the intensity values clipped normalized. A residual network non-square Dice loss function built upon TensorFlow...
Reducing the injected activity and/or scanning time is a desirable goal to minimize radiation exposure and maximize patients' comfort. To achieve this goal, we developed deep neural network (DNN) model for synthesizing full-dose (FD) time-of-flight (TOF) bin sinograms from their corresponding fast/low-dose (LD) TOF sinograms. Clinical brain PET/CT raw data of 140 normal abnormal patients were employed create LD FD The created through 5% undersampling list-mode PET data. split into seven bins...
Abstract We aim to synthesize brain time‐of‐flight (TOF) PET images/sinograms from their corresponding non‐TOF information in the image space (IS) and sinogram (SS) increase signal‐to‐noise ratio (SNR) contrast of abnormalities, decrease bias tracer uptake quantification. One hundred forty clinical 18 F‐FDG PET/CT scans were collected generate TOF sinograms. The sinograms split into seven time bins (0, ±1, ±2, ±3). predicted was reconstructed performance both models (IS SS) compared with...
Image artefacts continue to pose challenges in clinical molecular imaging, resulting misdiagnoses, additional radiation doses patients and financial costs. Mismatch halo occur frequently gallium-68 (
Abstract Introduction Providing tools for comprehensively evaluating scintigraphy images could enhance transthyretin amyloid cardiomyopathy (ATTR-CM) diagnosis. This study aims to automatically detect and score ATTR-CM in total body using deep learning on multi-tracer, multi-scanner, multi-center datasets. Methods In the current study, we employed six datasets (from 12 cameras) various tasks purposes. Dataset #1 (93 patients, 99m Tc-MDP) was used develop 2D-planar segmentation localization...
Abstract Purpose Low-dose CT protocols are widely used for emergency imaging, follow-ups, and attenuation correction in hybrid PET/CT SPECT/CT imaging. However, low-dose images often suffer from reduced quality depending on acquisition patient parameters. Deep learning (DL)-based organ segmentation models typically trained high-quality images, with limited dedicated noisy images. This study aimed to develop a DL pipeline ultra-low-dose Materials methods 274 raw datasets were reconstructed...
This study focuses on assessing the performance of active learning techniques to train a brain MRI glioma segmentation model.The publicly available training dataset provided for 2021 RSNA-ASNR-MICCAI Brain Tumor Segmentation (BraTS) Challenge was used in this study, consisting 1251 multi-institutional, multi-parametric MR images. Post-contrast T1, T2, and T2 FLAIR images as well ground truth manual were input model. The data split into set 1151 cases testing 100 cases, with remaining...
Abstract Purpose To derive and validate an effective radiomics-based model for differentiation of COVID-19 pneumonia from other lung diseases using a very large cohort patients. Methods We collected 19 private 5 public datasets, accumulating to 26,307 individual patient images (15,148 COVID-19; 9,657 with e.g. non-COVID-19 pneumonia, cancer, pulmonary embolism; 1502 normal cases). Images were automatically segmented validated deep learning (DL) the results carefully reviewed. first cropped...
We assess the performance of a recurrent frame generation algorithm for prediction late frames from initial in dynamic brain PET imaging.Clinical 18 F-DOPA PET/CT studies 46 subjects with ten folds cross-validation were retrospectively employed. A novel stochastic adversarial video model was implemented to predict last 13 (25-90 minutes) (0-25 minutes). The quantitative analysis predicted performed test and validation dataset using established metrics.The images demonstrated that is capable...
Partial volume effect (PVE) is a consequence of the limited spatial resolution PET scanners. PVE can cause intensity values particular voxel to be underestimated or overestimated due surrounding tracer uptake. We propose novel partial correction (PVC) technique overcome adverse effects on images.Two hundred and twelve clinical brain scans, including 50 18F-Fluorodeoxyglucose (18F-FDG), 18F-Flortaucipir, 36 18F-Flutemetamol, 76 18F-FluoroDOPA, their corresponding T1-weighted MR images were...
Objective. Cerebral CT perfusion (CTP) imaging is most commonly used to diagnose acute ischaemic stroke and support treatment decisions. Shortening CTP scan duration desirable reduce the accumulated radiation dose risk of patient head movement. In this study, we present a novel application stochastic adversarial video prediction approach acquisition time.Approach. A variational autoencoder generative network (VAE-GAN) were implemented in recurrent framework three scenarios: predict last 8...
We propose a small-animal PET scanner design combining two sets of monolithic crystals with different thicknesses.The detectors thinner serve for high resolution imaging while the thicker retain detection efficiency.Two models based on 10 and 12 detector blocks made LYSO were implemented in GEANT4 Monte Carlo toolkit.In each these models, half consisted crystal thickness mm whereas second had 2 mm.The scintillator coupled to SiPM arrays.For first model, arranged full-ring polygonal geometry...