Keywords: Acquisition Methods, Machine Learning/Artificial Intelligence, Learned Sampling Patterns

Motivation: Joint learning of sampling pattern (SP) and deep-learning (DL) reconstruction can have their loss functions in the k-space or image domain. It is not clear which approach is better.

Goal(s): Investigate this question by comparing the results of both loss functions, exploiting the flexibility of k-space domain loss functions for joint learning.

Approach: We modify the training of DL reconstructions to compare image or k-space domain losses. We tested on two DL networks and two different datasets, always using raw k-space as input.

Results: The differences in image quality are very small, but there are visual differences in the learned SPs.

Impact: This investigation shows that image loss is also a good option, but k-space loss is more flexible to control the shape of the SP. Interestingly, different learned SPs, with slightly different distributions of k-space samples, led to similar quality results.

Keywords: Acquisition Methods, Signal Representations, Workflow, Reproducibility

Motivation: Adapt the MR imaging for patient specific anatomical coverage.

Goal(s): Adapt MRI protocol while maintaining contrast, SNR and scan time in patient specific MR imaging.

Approach: The Intelligent slice placement estimated the brain segmentation and key landmarks for determining imaging volume center, orientation and coverage. The MRI protocol is then adjusted to demonstrate that contrast, SNR and scan time can be maintained within limits.

Results: The MRI images with patient adaptive MR imaging have similar view and coverage of anatomy while contrast, SNR and scan time is maintained. 

Impact: MRI protocol adaption (while maintaining contrast, SNR and scan time within limits) demonstrated in this abstract along with anatomical coverage for patient adaptive scanning is essential for consistent high quality MRI imaging within and across clinical sites. 

Keywords: Acquisition Methods, RF Pulse Design & Fields, simultaneous multi-slice

Motivation: Designing radiofrequency pulses can be a challenging, time consuming, iterative process.

Goal(s): Simplify the radiofrequency pulse design process using deep learning. 

Approach: First, a complex model was trained on a dataset of ~34k RF pulses. Second, a simpler model was trained on a dataset of 1.2M RF pulses. Both models output the characteristics needed to generate a fully sampled radiofrequency waveform.

Results: Model 2 performed better than Model 1, however, the root mean squared error in expected vs. generated slice profiles on a subset of the test data was still high at 36.5%. Future work will implement an optimization loop. 

Impact: A fully functioning deep learning model could serve as a tool for researchers designing power independent of number of slices pulses to improve slice profiles for SMS imaging as well as novel applications such as in ex-nuclei imaging.

Keywords: Acquisition Methods, Machine Learning/Artificial Intelligence, Automated plane prescription, workflow

Motivation: Automated plane prescription tools can make cardiac MRI more accessible, but their accuracy needs to be validated. 

Goal(s): We aim to determine the accuracy requirements for an AI-driven automated prescription tool.

Approach: To determine the accuracy requirements for an AI-driven automated prescription tool we compare landmarks set by the tools to those set manually by operators with different levels of experience. 

Results: The prototype can match the average performance of an operator group, outperforming the less experienced individuals.

Impact: To be reliable the performance of automated prescription tools needs to be established. Here an AI-driven automated prescription tool for cardiac MRI planes could match the average performance of an operator group, outperforming the less experienced individuals.

Keywords: Acquisition Methods, Machine Learning/Artificial Intelligence, Spectroscopy, Brain, Artifacts, Convolutional Neural Network

Motivation: Deep learning is a promising new tool for post-processing MRS data. Neural network-based MRS acquisition methods do not yet exist, but should lead to higher-quality data.

Goal(s): The goal of this work was to create an “intelligent MRS scan” by integrating a Convolutional Neural Network (CNN) directly into a MRS acquisition protocol.

Approach: Here, a CNN-powered pre-scan collects a single-transient from 48 different gradient geometries, and updates future scans, without human intervention, to avoid out-of-voxel (OOV) artifacts.

Results: The AI-informed scan produced high-quality data for all participants while the control parameters failed half of the time in the artifact-rich mPFC region.

Impact: The work demonstrates the first AI-integrated MRS scan protocol in which an intelligent pre-scan modifies scan parameters to improve data quality, here reducing out-of-voxel artifacts.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, mute, acoustic reduction, Deep learning

Motivation: MRI mute technique reduces scanning noise by lowering the gradient slew rate which increase echo spacing, resulting in image blurring and longer scan time. 

Goal(s): To design a scanning method that simultaneously reduces scan time and scanning noise without compromising image SNR and clarity. 

Approach: Develop a DL-Recon post-processing model to enhance SNR and clarity of images which are acquired from optimized knee scanning protocol with mute technique.

Results: The proposed scanning method improves SNR about 38% and reduces scan time and noise sound pressure separately about 44.8% and 86%.

Impact: Combining DL-Recon with mute sequences help doctors to diagnose more patients. It also enhances the success of clinical scans by decreasing scan noise to improve patient comfort. This is a successful application of DL-Recon with mute scanning technique.

Keywords: Acquisition Methods, High-Field MRI

Motivation: The introduction of Machine Learning-based Image Reconstruction ("Deep Learning") offers a fresh opportunity to explore MR parameter space without the restrictive requirement to maximise MR signal.

Goal(s): Demonstrate an application of the Low Stochastic Regime approach using low flip angle refocusing with good SNR and strong tissue contrast.

Approach: Fast Spin Echo (FSE) images were acquired using reduced flip angle refocusing and extended echo train lengths, in combination with Deep Learning-Reconstruction (“DL-Recon”). 

Results: Using DL-Recon to effectively weaken the conventional constraint of maximising MR signal, the redesigned sequence produced images with lower RF power deposition but similar contrast to the product CPMG sequence.

Impact: Clinical applications where T2-weighted imaging is SAR-limited.

Keywords: AI/ML Image Reconstruction, Data Processing, SMS,VarNet

Motivation: To improve reconstruction quality of Simultaneous Multi-Slice (SMS) imaging

Goal(s): To develop a deep learning reconstruction method without sacrificing image detail and fidelity, even in data-scarce scenarios.

Approach: Our approach involves a novel integration of subject-specific RAKI and a Variational Network (VarNet) within an unrolled iteration framework, testing three different guidance strategies to improve reconstruction quality.

Results: The RAKI-VarNet In-Iteration Parallel method yielded the most promising results, showing a reduction in noise and artifacts while maintaining robustness on both seen and unseen data sets, including challenging EPI data.

Impact: This technique presents a robust solution for high-fidelity SMS MRI reconstruction with improved generalizability and detail preservation.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, Transfer learning, Generalization

Motivation: Long MR procedure times often result in a shortage of patient data for specific cases, affecting the performance of data-dependent deep networks. Transfer learning offers a remedy, enabling pretrained models to adapt to new domains with limited data availability.

Goal(s): Our goal is to create a network capable of producing clinically acceptable reconstructions with limited data.

Approach: We leverage representation learning to refine low-resolution data and enhance final reconstructions in data-limited scenarios.

Results: Successful transfers with 100 and 40 training sample sets were achieved. Both networks achieve comparable results to the large dataset (240 samples) trained network.

Impact: Our approach has broader clinical uses beyond acquisition protocols, extending to vendor differences and scenarios with limited access to disease scans due to privacy concerns.  It presents an opportunity to tackle limited data generalization challenges without adding architectural complexity.

Keywords: AI/ML Image Reconstruction, Quantitative Imaging, T1 mapping, Higher resolution, Deep learning-based Image reconstruction

Motivation: The MOLLI acquisition scheme is frequently used for T1 mapping of the heart. MOLLI restricts the spatial resolution of the resulting T1 maps due to acquiring the inversion recovery images in single-shot fashion.

Goal(s): To allow the acquisition of higher spatial resolution T1 maps.

Approach: Single-shot acquisitions are accelerated and image sets are reconstructed using a neural network. The deep learning-based reconstruction is integrated into an existing T1 mapping sequence.

Results: The proposed method produces higher spatial resolution T1 maps. The corresponding T1 values do not differ significantly from T1 values computed by the vendor sequence.

Impact: The acquisition of higher spatial resolution T1 maps is achieved. The proposed method may improve the detection of small focal lesions without increasing the required scan time or breath hold duration.

Keywords: AI/ML Image Reconstruction, Quantitative Imaging, Fat-water seperation, PDFF, Deep Learning, Fat Quantification, Physics Informed Deep Learning

Motivation: The novel Deep Learning (DL)-based Ad-Hoc Reconstruction (AHR) method for fat-water separation in Multi Echo-Magnetic Resonance Imaging (ME-MRI) has absolute generalizability. It can perform fat-water separation with the ME-MRIs from any anatomical region and views with varied numbers of echoes.

Goal(s):  This research investigates the fat-water separation performance of spatial smoothing incorporated DL-based AHR method in ME-MRIs with and without noise.

Approach:  The fat-water separation biophysical model based loss in AHR is added with spatial smoothing constraints.

Results: Results demonstrate that incorporating spatial smoothing in AHR improves the fat-water separation performance in ME-MRIs without noise, however, no performance improvements in ME-MRIs containing noise.

Impact: The PDFF maps obtained from fat-water separation in Multi Echo-MRI (ME-MRI) are of diagnostic and prognostic value in many diseases. This study investigates the performance of a novel Deep Learning-based Ad-Hoc Reconstruction method with spatial smoothing for fat-water separation.

Keywords: AI/ML Image Reconstruction, Tumor, Rectal adenocarcinoma, Imaging quality, Diagnostic performance

Motivation: Accelerated MRI is an imminent need clinically to satisfy the growing disease burden of rectal cancer.

Goal(s): This study aims to conduct clinical assessment of DLR-based accelerated rectal MRI, encompassing image acquisition, image quality, diagnostic performance, and reding efficiency.

Approach: Two sets of T2WI using standard fast spin-echo (FSE_standard) and DLR-based accelerated FSE (FSE_DL) were prospectively compared.

Results: FSE_DL showed superior image quality and reading efficiency than FSE_standard, with a 65% reduction in acquisition time. DLR could assist to enhance the accuracy of T-staging for junior radiologists, preserving equivalent diagnostic performance in N staging, EMVI, and MRF.

Impact: This study offered a comprehensive and viable perspective on the application of DLR in rectal MRI, which facilitated improved image quality and reading efficiency, while reducing acquisition time. Moreover, it enhanced the accuracy of T-staging for junior radiologists.

Keywords: AI/ML Image Reconstruction, Low-Field MRI, Brain

Motivation: High, isotropic resolution (e.g., 1mm) is desirable for lesion detection and biomarkers extraction for cognitive disorders. However, ultra-low-field (ULF) MRI severely suffers from low spatial resolution and signal-to-noise ratio.

Goal(s): To investigate the potential of 3D deep learning in generating <=1mm isotropic resolution results from 2D partial Fourier-sampled, low-resolution noisy brain images acquired from our custom-made 0.055T scanner.

Approach: We advanced 3D deep learning partial Fourier reconstruction and super-resolution method (PF-SR) to achieve 3x/4x super-resolution factors.

Results: Preliminary results indicate possibility of PF-SR with higher super-resolution factors on reconstructing experimental ULF T1w/T2w data to 1/0.75mm3 with reduced artefacts and noise.

Impact: Enhancing image resolution and fidelity for fast ultra-low-field brain imaging at 0.055T using data-driven 3D deep learning approach to <=1mm3 resolution potentially enables image-guided therapies and valuable neuroimaging analysis for assessing aging and cognitive conditions.

Keywords: AI/ML Image Reconstruction, Cancer

Motivation: MRI is a powerful diagnosis tool for renal tumors with a long acquisition time. How to improve image quality while shortening acquisition time is a major research focus.

Goal(s): To use a deep learning (DL) algorithm to reconstruct low-resolution T2-weighted turbo spin-echo (TSE) renal MRI scans and compare with standard-resolution T2-weighted TSE sequence.

Approach: A total of 14 patients with clinically suspected renal tumors who underwent renal low-resolution DL-reconstructed T2-weighted TSE sequence(T2_DL) and standard-resolution T2-weighted TSE sequence (T2_S) were included.

Results: T2_DL reduced acquisition time by 32% and improved overall image quality compared with T2_S.

Impact: A DL reconstruction method for low-resolution renal T2-weighted TSE sequence has the potential to reduce acquisition time and improve image quality compared with standard acquisition method, which may help detect renal lesions early and improve the survival rates of patients.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence

Motivation: Complex-valued deep learning framework has not been fully investigated in human normal-field and low-field MRI reconstruction and preprocessing.

Goal(s): We aim to replace conventional numerical methods with deep learning network, which reconstruct and preprocess the k-space data in parallel.

Approach: An artificial Fourier transform network (AFT-Net) is proposed to directly processes the complex-valued raw data in the sensor domain.

Results: An evaluation of accelerated reconstruction and denoised reconstruction shows that AFT-Net demonstrated the ability to reconstruct the data with significantly accelerate acquisition and random Gaussian noise. The proposed AFT-Net is an efficient and accurate approach for MRI reconstruction and preprocessing from raw data.

Impact: MRI reconstruction and preprocessing with AFT-Net should be able to determine the domain-manifold mapping and process k-space data directly, which shows superior performance and can be served as an efficient and accurate approach for human high-field and low-field MRI acquisition.

Keywords: AI/ML Image Reconstruction, Brain

Motivation: Theoretically, results of general unfolding network are not a convergence point of the ill-posed problem for MRI reconstruction.

Goal(s): Our goal was to design an accelerated convergence unfolding network that is easier to approach the convergence point.

Approach: Using accelerated gradient descent method as the framework, the proximal gradient descents of MRI high-frequency and low-frequency information are completed in a single iteration, which achieves faster convergence.

Results: The reconstructed MRI of our unrolled network performs better than others.

Impact: The convergence point can be effectively approximated by accelerating convergence rate, but it is still not guaranteed to be the optimal point, and further work should seek the optimal value.