A5102828453- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Digital Radiography and Breast Imaging
- MRI in cancer diagnosis
- Global Cancer Incidence and Screening
- Gamma-ray bursts and supernovae
- Advanced MRI Techniques and Applications
- Pulsars and Gravitational Waves Research
- Astrophysics and Cosmic Phenomena
- Advanced X-ray and CT Imaging
- Medical Imaging Techniques and Applications
University Hospital of Zurich
2020-2024
University of Zurich
2020-2022
Abstract Purpose The aim of this study was to develop and test a post-processing technique for detection classification lesions according the BI-RADS atlas in automated breast ultrasound (ABUS) based on deep convolutional neural networks (dCNNs). Methods materials In retrospective study, 645 ABUS datasets from 113 patients were included; 55 had classified as high malignancy probability. Lesions categorized 2 (no suspicion malignancy), 3 (probability < 3%), 4/5 > 3%). A network trained...
High breast density is a well-known risk factor for cancer. This study aimed to develop and adapt two (MLO, CC) deep convolutional neural networks (DCNN) automatic classification on synthetic 2D tomosynthesis reconstructions.In total, 4605 images (1665 patients, age: 57 ± 37 years) were labeled according the ACR (American College of Radiology) (A-D). Two DCNNs with 11 layers 3 fully connected each, trained 70% data, whereas 20% was used validation. The remaining 10% as separate test dataset...
Marked enhancement of the fibroglandular tissue on contrast-enhanced breast magnetic resonance imaging (MRI) may affect lesion detection and classification is suggested to be associated with higher risk developing cancer. The background parenchymal (BPE) qualitatively classified according BI-RADS atlas into categories "minimal," "mild," "moderate," "marked." purpose this study was train a deep convolutional neural network (dCNN) for standardized automatic BPE categories.This IRB-approved...
The aim of this study was to investigate the potential a machine learning algorithm classify breast cancer solely by presence soft tissue opacities in mammograms, independent other morphological features, using deep convolutional neural network (dCNN). Soft were classified based on their radiological appearance ACR BI-RADS atlas. We included 1744 mammograms from 438 patients create 7242 icons manual labeling. sorted into three categories: "no opacities" (BI-RADS 1), "probably benign 2/3) and...
Abstract Objectives Development of automated segmentation models enabling standardized volumetric quantification fibroglandular tissue (FGT) from native volumes and background parenchymal enhancement (BPE) subtraction dynamic contrast-enhanced breast MRI. Subsequent assessment the developed in context FGT BPE Breast Imaging Reporting Data System (BI-RADS)-compliant classification. Methods For training validation attention U-Net models, data coming a single 3.0-T scanner was used. testing,...
In DCE-MRI, the degree of contrast uptake in normal fibroglandular tissue, i.e., background parenchymal enhancement (BPE), is a crucial biomarker linked to breast cancer risk and treatment outcome. accordance with Breast Imaging Reporting & Data System (BI-RADS), it should be visually classified into four classes. The susceptibility such an assessment inter-reader variability highlights urgent need for standardized classification algorithm. this retrospective study, first post-contrast...
Abstract Background We investigated whether features derived from texture analysis (TA) can distinguish breast density (BD) in spiral photon-counting computed tomography (PC-BCT). Methods In this retrospective single-centre study, we analysed 10,000 images 400 PC-BCT examinations of 200 patients. Images were categorised into four-level scale ( a – d ) using Breast Imaging Reporting and Data System (BI-RADS)-like criteria. After manual definition representative regions interest, 19 (TFs)...
Abstract Objectives The aim of this study was to develop and validate a commercially available AI platform for the automatic determination image quality in mammography tomosynthesis considering standardized set features. Materials methods In retrospective study, 11,733 mammograms synthetic 2D reconstructions from 4200 patients two institutions were analyzed by assessing presence seven features which impact regard breast positioning. Deep learning applied train five dCNN models on detecting...
The aim of this study was to investigate the potential a machine learning algorithm accurately classify parenchymal density in spiral breast-CT (BCT), using deep convolutional neural network (dCNN). In retrospectively designed study, 634 examinations 317 patients were included. After image selection and preparation, 5589 images from different BCT sorted by four-level scale, ranging A D, ACR BI-RADS-like criteria. Subsequently four dCNN models (differences optimizer spatial resolution)...
The purpose of this study was to determine the feasibility a deep convolutional neural network (dCNN) accurately detect abnormal axillary lymph nodes on mammograms. In retrospective study, 107 mammographic images in mediolateral oblique projection from 74 patients were labeled three classes: (1) "breast tissue", (2) "benign nodes", and (3) "suspicious nodes". Following data preprocessing, dCNN model trained validated with 5385 images. Subsequently, tested "real-world" dataset performance...
Background: After breast conserving surgery (BCS), surgical clips indicate the tumor bed and, thereby, most probable area for relapse. The aim of this study was to investigate whether a U-Net-based deep convolutional neural network (dCNN) may be used detect in follow-up mammograms after BCS. Methods: 884 and 517 tomosynthetic images depicting calcifications were manually segmented classified. A segmentation trained with 922 validated 394 images. An external test dataset consisting 39...
Abstract Background The presence of a blurred area, depending on its localization, in mammogram can limit diagnostic accuracy. goal this study was to develop model for automatic detection blur diagnostically relevant locations digital mammography. Methods A retrospective dataset consisting 152 examinations acquired with mammography machines from three different vendors utilized. areas were contoured by expert breast radiologists. Normalized Wiener spectra (nWS) extracted sliding window...