Keywords: Analysis/Processing, Safety, SAR, MRSaiFE

Motivation: MRI poses safety risks due to tissue damage via SAR hotspots. We have previously developed MRSaiFE, an AI-based SAR prediction tool.

Goal(s): This study expands MRSaiFE with experimental, in vivo, training data.

Approach: Images from a subject are segmented into a numerical model that is simulated to obtain SAR. The MRSaiFE input is the scanned image, and the predicted output SAR is obtained from training on the simulated SAR.

Results: Good agreement (0.4% MSE, 6% RMSE, and 81% SSIM) demonstrates feasibility of using 1) experimental training data and 2) scanned input images, enabling future prediction from in vivo localizers.

Impact: By replacing conservative SAR margins with patient-specific values, MRSaiFE offers potential for enhanced sensitivity, resolution, or reduced scan time. Additionally, it could notably enhance safety in patients with medical implants, hyperthermia treatments, and in MRI procedures at ultra-high fields.

Keywords: Analysis/Processing, Segmentation, Spine stations; Automatic prescription

Motivation: In spine scanning with MRI, multiple localizer scans are acquired to manually set the stations. 3D surface coil sensitivity maps, with low-resolution but large FOV, which are acquired as part of the prescan can potentially be used to automatically determine the station boundaries.  

Goal(s): Utilize the existing information in the MRI scanner to automatically predict the location of spine stations and thereby accelerate the workflow. 

Approach: Use a deep learning framework to automatically identify the stations of the spine anatomy from the coil sensitivity maps.

Results: The deep learning model shows good localization of spine stations with mean centroid errors less than 15mm.

Impact: Spine stations can be identified from large FOV, low-resolution surface coil sensitivity maps in MRIs using our deep learning framework, which can be used for fast and automatic spine anatomical planning and imaging. 

Keywords: Analysis/Processing, Breast, Nipple, Detection, Landmark

Motivation: As the nipple position knowledge becomes part of standardized report, the automatic detection can ease clinician’s workflow.

Goal(s): The aim of our work is to accurately detect the position of the nipples in a dynamic contrast-enhanced (DCE) MR image.

Approach: A reinforcement learning approach combined with a multi-constructor and multi-centric database enabled to initiate the development of a versatile tool in line with clinical real life. The detection problem was addressed using a Deep Q-Network trained with 248 breast DCE MR images.

Results: The nipple positioning error is less than 10 millimeters in most of the breasts tested, i.e. 95/102 breasts.

Impact: Nipple detection is a tedious task for clinicians and an arduous one for algorithms. Lesion to nipple distance is valuable information when planning surgery. This study explores the landmark detection domain to automate nipple detection using a reinforcement learning approach.

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence, thyroid segmentation, MRI, hypothyroidism, hyperthyroidism

Motivation: Assess the feasibility of calculating thyroid volumes using T2-weighted MR imaging.

Goal(s): Quantify thyroid volume using deep learning methods in healthy, hypo and hyperthyroid patients.

Approach: We assessed MRIs of 469 healthy, 606 hypothyroid, and 203 hyperthyroid individuals, matched for age, weight, and height.

Results: Findings indicated altered thyroid volumes in healthy (11.02 ml), hyperthyroid (9.42 ml), and hypothyroid (8.35 ml), BMI-normalized volumes also differed: healthy (0.445), hyperthyroid (0.387), and hypothyroid (0.337). There is a moderate association between thyroid volume and weight (0.41, p=3.2e-25) and a weaker link with height (0.17, p=3.9e-05).

Impact: MRI-based analysis of thyroid volumes in healthy, hyper and hypothyroid patients using deep learning, revealing varied absolute and BMI-based normalized volumes.

Keywords: Analysis/Processing, Breast

Motivation: Breast cancer has become the leading cancer worldwide. Hemodynamic features obtained from breast DCE-MRI perfusion maps can accurately quantify tumor pathophysiology. However, traditional estimation of perfusion parameter maps requires significant computational resources and time.

Goal(s): To investigate whether deep learning techniques can synthesize K^trans perfusion parameter maps from contrast-enhanced MRI. 

Approach: A pix2pix-based cGAN architecture was proposed to generate breast K^trans perfusion maps.

Results: The K^trans values of the tumor regions in the synthetic and real K^trans maps show a strong correlation. Two experienced radiologists could not distinguish between real and synthetic K^trans maps.

Impact: This study presents a novel feasible approach for synthesizing Ktrans perfusion maps, which enables rapid generation of high-quality and low-noise perfusion maps, thereby facilitating more effective application of these maps in clinical practice by physicians.

Keywords: Analysis/Processing, Elastography

Motivation: The assessment of liver MRE exams is time-consuming, as is the reconstruction process performed by the scanner.

Goal(s): Our objective was to automate the reconstruction and evaluation of stiffness maps, allowing for the calculation of liver stiffness based solely on MRE data, all accomplished within a matter of seconds.

Approach: To achieve this, we developed a U-Net-based model combination that takes both magnitude and phase MRE images as input. This model generates stiffness maps and corresponding ROIs while also estimating stiffness within the ROI.

Results: The proposed model successfully and accurately estimated liver stiffness, reducing the entire process to a few seconds.

Impact: The proposed model can effectively assess liver stiffness using MRE data, substantially decreasing image reconstruction and analysis time to just a few seconds - a crucial advancement for clinical applications.

Keywords: Analysis/Processing, Segmentation

Motivation: Paraspinal muscle mass estimation for liver transplant candidacy is practically limited by tedious segmentation.

Goal(s): Develop an automatic segmentation algorithm using a convolutional neural network (CNN) for segmentation of abdominal paraspinal muscles to calculate skeletal muscle index in cirrhotic patients.

Approach: A U-Net CNN was trained on spin echo images and evaluated with Dice coefficient. Skeletal muscle index of original and predicted masks was compared with independent t-test, ANOVA and a Bland-Altman plot.

Results: Dice coefficient was >0.88, with a mean bias of <1% between CNN SMI and manual SMI, while not being statistically significant. SMI and liver frailty were not directly associated.

Impact: Faster and precise segmentation of abdominal paraspinal muscles to calculate muscle mass in cirrhotic patients would reduce the time burden, thereby increasing practicality for MRI skeletal muscle index estimation.

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence

Motivation: A previously developed MRI method for quantitative myocardial oxygen extraction mapping (mOEF) showed promising results, but image quality suffered from distortion and inhomogeneity artifacts. 

Goal(s): The objective of this study is to evaluate a new CMR method for in vivo measurement of mOEF utilizing on a deep-learning quantification approach in healthy controls. 

Approach:  A new pulse sequence and a novel deep learning-based analysis method were created and evaluated on a group of healthy subjects.

Results: This investigation yielded dramatically improved image quality, which allowed reliable evaluation of reproducibility and distribution of mOEF within the heart.

Impact: Our study, involving 20 healthy volunteers, showcased outstanding reproducibility in the measurements, suggesting its potential for translation into imaging studies for patients with myocardial metabolic dysfunction.

Keywords: Analysis/Processing, Thermometry

Motivation: Noise is a challenge for real-time MR-thermometry. The Extended Kalman Filter (EKF) has been shown to reduce noise successfully in temperature maps.

Goal(s): We modified KalmanNet, a recurrent neural network emulating an EKF, using a proper orthogonal decomposition (POD) to reduce computational cost.

Approach: POD-KalmanNet was applied to microwave ablations of 14 bioprotein phantoms. Mean squared errors (MSE) and Sørensen-Dice-Coefficient (DSC) were compared between noisy and filtered data.

Results: Results show a highly significant reduction of MSE (p < 0.001) and a significant increase of DSC (p < 0.05) for filtered images compared to noisy images. POD-KalmanNet could enable real-time filtering of MR-Thermometry.

Impact: KalmanNet was modified using proper orthogonal decomposition. This modification allows lower memory usage and inference times. It makes the application of KalmanNet to 3D temperature maps practically feasible. Reducing noise in 3D temperature maps can improve outcomes of thermoablation procedures.

Keywords: Analysis/Processing, Segmentation, Quantitative analysis

Motivation: Despite the significance of regional myocardial analysis in clinical practice it's performed manually, which is a  time-consuming task. Therefore, automation of myocardium regional analysis is a relevant task.

Goal(s): The goal of this work is to develop a tool for myocardium regional quantitative analysis automation  

Approach: A trained neural network segment myocardium and fibrosis. The segmented myocardium undergoes additional segmentation into 17 segments using mathematical algorithm. The fibrosis volume in each segment is measured.

Results: U-Net achieved median DSC 0.75 for fibrosis and 0.85 myocardium. The fibrosis regional detection accuracy of our algorithm 0.71 according to F-score.  Our algorithm speed is about 30s/patient.

Impact: Our tool allows to speed up and improve the accuracy of myocardium regional analysis.

Keywords: Analysis/Processing, Perfusion, Machine Learning/Artificial Intelligence, Cancer

Motivation: Despite extensive research and promising initial results, quantitative dynamic contrast-enhanced (DCE) MRI is marginal in clinical practice, due to lack of automation and low reproducibility.

Goal(s): Introduce an end-to-end deep learning approach for an automated and more reproducible DCE-MRI pipeline.

Approach: Two networks, one reconstructing undersampled k-t data via Movienet and the other estimating perfusion and MR parameters, were merged into a unified, automated pipeline. The approach was tested on a volunteer and a patient with cervical cancer.

Results: Automated processing yielded images in under 2 seconds, comparable in quality to GRASP and providing multiparametric mapping of perfusion and MR from one acquisition.

Impact: The proposed fast automated data processing pipeline including deep learning reconstruction and quantification can be an important clinical tool to exploit the information from DCE-MRI to improve tumor diagnosis and treatment response evaluation.

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence

Motivation: While multiphasic contrast-enhanced MRI has propelled noninvasive pCR prediction in breast cancer, its limited temporal resolution restricts quantitative analysis, affecting generalizability and interpretability.

Goal(s): To enhance pCR prediction, we integrated retrospective pharmacokinetic quantification by addressing the temporal resolution limit using deep learning.

Approach: We incorporated a novel retrospective pharmacokinetic quantification approach into our pCR prediction model to better capture the tumor microenvironment's pharmacokinetic indicators.

Results: Our approach improved predictive accuracy in external test datasets, demonstrating the method's superior performance and broader applicability.

Impact: Deep-learning pharmacokinetic quantification enhances the accuracy and applicability of pCR prediction using multiphasic DCE-MRI, offering the potential for precise pre-treatment evaluation that could streamline NAC targeting and minimize initiation delays for breast cancer patients unlikely to respond to standard treatments.

Keywords: Analysis/Processing, Quantitative Imaging

Motivation: Mapping MRI signal into tissue parameters aims to identify robust physiologic-phenotypic associations. However, conventional methods are computationally expensive, limiting their applicability in research or clinical practice.

Goal(s): To develop fast and robust techniques for estimating quantitative tissue parameters under a probabilistic framework using MRI signals.

Approach: Train and evaluate the performance of variational autoencoders and compare its capabilities with state-of-the-art deep learning methods on both synthesized and real MRI data.

Results: Compared to existing autoencoder-based methods, both synthetic and real data experiments show enhanced performance of VAEs on tissue parameter estimation. Parameter maps produced from real data show higher similarity to gold-standard maps.

Impact: We show that variational autoencoders can be trained for fast inference of quantitative parameter estimation MRI data quantification in qMRI.

Keywords: Analysis/Processing, Vessels, Thoracic Aorta, Segmentation, Landmark Detection, Causal Discovery

Motivation: The thoracic aorta is often affected by life-threatening, undetected morphological changes. Prior works primarily focused on factors correlating with aortic aneurysms but lack the investigation of causal influences related to morphological changes.

Goal(s): Our goal is to perform automatic aortic shape analysis inline on the scanner. We investigate causal dependencies between metadata and thoracic aortic diameter in approx. 30,000 non-contrast-enhanced MRA.

Approach: We apply a deep learning framework for shape analysis and Peter-Clark-algorithm to investigate causal influences on the thoracic aorta.

Results: We found that sex, age, height, BMI, hypertension, and vascular-stiffness causally impact the aorta’s diameter, whereas diabetes lacks a causal relationship.

Impact: This study reveals causal influences on morphological changes of the thoracic aorta using a large epidemiological dataset (~30,000 non-contrast-enhanced-MRA). A deep-learning-based framework supports the identification of causal factors impacting the aortic diameter and thereby, enabling early detection of life-threatening risks.

Keywords: Analysis/Processing, DSC & DCE Perfusion, DEEP LEARNING, BRAIN TUMOR, PERFUSION PARAMETERS

Motivation: Quantitative analysis of dynamic-contrast-enhanced MRI(DCE-MRI) is valuable approach for mapping tumor physiology; however, traditional non-linear-least squares(NLLS) methods are slow and provide noisy maps. Deep-learning(DL) approach offers solutions, yet reported models rely on signal-intensity-time-curves(SIC) which are MRI-acquisition protocol dependent.

Goal(s): To develop DL network(CNN_CON) that uses concentration-time-curves(CTCs) to estimate perfusion-parameters(GTKM) and compare with SIC-based DL network(CNN_SIGNAL).

Approach: Two CNN networks were developed using CTC and SIC data(simulations and in-vivo). Performance of models was evaluated on simulated data with different protocols and experimental data.

Results: The CNN_CONC outperforms NLLS & CNN_SIGNAL in terms of speed, accuracy and smoothness of maps.

Impact: The proposed DL framework improves DCE-MRI analysis by providing more accurate and robust results in less time. It eliminates protocol dependence and holds the potential for routine clinical use in the diagnosis and treatment planning of brain tumor patients.