Keywords: Analysis/Processing, Spinal Cord, MRI, Spine, Spine Labelling, Foundation Model, ML/AI

Motivation: Spine labelling is a step crucial for several important tasks such as MRI scan planning or associating image regions with mentions in clinical reports and others. Automating it can lead to significant benefits but developing automated solutions requires extensive annotations of vertebra labels.

Goal(s): To automate spine labelling without extensively training a DL model with manual annotations.

Approach: We adapted a vision foundational model-based approach that combines spine domain knowledge to predict spine labels.

Results: Our spine labelling method gives an average accuracy of 79% and 86% for cervical and lumbar high resolution T1 images, respectively.

Impact: Leveraging spatially relevant landmarks (disc) and vision foundation deep learning model, spine labels are predicted using one-shot localization. The proposed method doesn’t require any prior data for model training.

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence

Motivation: Morphometric assessment of cartilage(e.g.,thickness), through MRI yields accurate measurements on the progression of Osteoarthritis(OA). Such quantitative measurements require image segmentation techniques. Recent developments in Visual Foundational Models(VFM) bring opportunities to increasing generality and robustness.

Goal(s): What improvements can VFM-based approaches bring to automatic segmentation of knee 3DMRIs, and how it compares to traditional convolution networks(CNNs)?

Approach: Trained 2DVFM, 3DCNN, and a modified 3DVFM on 500MRI volumes. Evaluated qualitative and quantitatively on external datasets.

Results: The proposed 3D-VFM, demonstrates a slight advantage on quantitative morphological assessment, but strongly outperforms others when qualitatively assessed by radiologists, presenting a promising direction and better generalization.

Impact: By leveraging Visual Foundational Models (VFM) in the morphometric assessment of cartilage through 3D MRIs, our research demonstrates significant promise in enhancing the accuracy and generalization of knee segmentation to be applied to osteoarthritis progression measurements.

Keywords: Diagnosis/Prediction, fMRI (resting state), functional connectivity, phenotypic prediction, meta-learning, transfer learning

Motivation: Resting-state functional connectivity (RSFC) is widely used to predict phenotypes in individuals. However, predictive models may fail to generalize to new datasets due to differences in population, data collection, and processing across datasets.

Goal(s): To resolve the dataset difference issue, we aimed to generalize knowledge from multiple diverse source datasets and translate the model to new target data.

Approach: Here we proposed Multi-domain and Uni-domain Fusion (MUF) method that combines cross-domain learning and intra-domain learning, to capture both domain-general information and domain-specific information.

Results: The results show that our MUF outperformed 4 strong baseline methods on 6 target datasets. 

Impact: Our MUF method is adept at addressing the challenges introduced by different population profiles, fMRI processing pipelines, and prediction tasks. We offer a robust and universal learning strategy for domain-generalization in fMRI-based phenotypic prediction.

Keywords: AI/ML Software, Software Tools

Motivation: Magnetic Resonance Imaging (MRI) plays an important role in medical diagnosis, generating petabytes of image data annually in large hospitals. Local data processing demands substantial manpower and hardware investments. Data isolation across different healthcare institutions hinders cross-institutional collaboration.

Goal(s): Solve a series of problems existing in current hospitals.

Approach: Integrating cloud computing, 6G, edge computing, federated learning, and blockchain.

Results: Cloud-MRI transforms raw data to the Imaging Society for Magnetic Resonance in Medicine Raw Data (ISMRMRD) format and enables fast reconstruction, AI training, and analysis. Results are relayed to cloud radiologists.

Impact: This system safeguards data, promotes collaboration, and enhances diagnostic precision and efficiency.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, Federated Learning, Image-to-image Translation

Motivation: Applying machine learning (ML) in MRI necessitates the development of large and diverse datasets, which is a challenging process. Federated learning (FL) is a new frontier in ML that offers the possibility of multi-site data aggregation.

Goal(s): In our study, we examine a traditional deep convolutional neural network applied to multiple sources with that of the FL technique using different aggregation methods.

Approach: As a proof-of-concept, we employ four publicly available MRI datasets and carry out image-to-image translation between T1- and T2-weighted scans.

Results: Our findings suggest that the FL generalizes the model more effectively than using models trained at each site separately.

Impact: Our research demonstrated the crucial role of federated learning in medical imaging. It also emphasized the significance of selecting an appropriate aggregation algorithm considering the data type and degree of heterogeneity.

Keywords: Diagnosis/Prediction, Segmentation

Motivation: Site and scanner specific variations in prostate MRI impact performance of deep learning (DL) based models. Federated learning allows for privacy preserving training of DL models without the need for data sharing. 

Goal(s): In this study, we train DL models for prostate cancer segmentation on MRI using the Rhino Health federated computing platform. 

Approach: We adopt 3D UNet architecture to train the DL models on 2 publicly available datasets.

Results: DL models trained using a federated approach result in more generalizable models compared to those trained on single site data.

Impact: Successful development of deep learning based prostate cancer segmentation models on MRI using federated learning will result in reproducible and generalizable models. These can enhance clinical adoption and potentially improve downstream diagnostic and treatment workflows for prostate cancer.

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence, Diffusion MRI

Motivation: While anatomical brain parcellation has long been performed using anatomical MRI and atlas-based approaches, deep learning methods together with diffusion MRI techniques can improve parcellation performance and interpretation of prediction uncertainty.

Goal(s): Our goal is to design an uncertainty-aware deep learning network to utilize multiple diffusion MRI parameters for accurate brain parcellation while enabling voxel-level uncertainty estimation.

Approach: We include five evidential deep learning subnetworks and perform an evidence-based ensemble for parcellation prediction and uncertainty estimation.

Results: The results demonstrate our method’s superior parcellation performance over several state-of-the-art methods, its promising results in unseen patient scans, and potential applications in brain abnormality detection.

Impact: The proposed approach enables improved accuracy in brain parcellation from diffusion MRI, facilitating the understanding of the human brain in health and disease. It may also serve as an effective tool for brain abnormality detection, fostering inquiries into uncertainty-quantified diagnostics.

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence, saliency maps; biomarkers

Motivation: It is significant to understand if saliency maps can be considered as potential biomarkers by providing reliable anatomical information in 3D medical imaging classification. 

Goal(s): We found saliency maps can provide different (even mutually exclusive) information with randomised models. It is necessary to estimate the robustness of saliency maps under the stochastic training process.

Approach: We introduced a novel method by re-organising the saliency scores in the saliency maps and quantify the inter-map difference for estimating the robustness of saliency maps.

Results: All selected explanation methods were not able to exhibit strong performance in the estimation of robustness of saliency maps.

Impact: Our estimation provides evidence that saliency maps are not competent to maintain the robustness under the stochastic training process. Researchers should be critically careful when utilising saliency maps as biomarkers for interpretation. 

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence, fMRI, Working memory

Motivation: fMRI is allows studying human brain activity in vivo, but standard analyzing fMRI cannot capture nonlinear relationships between activity and variables. Utlizing deep learning (DL) models may capture such relationships, providing new insight into mechanisms underlying human health and disease.

Goal(s): To evaluate our interpretable DL pipeline in fMRI analysis using three large cohorts to demonstrate its generalizability and reproducibility.

Approach: We built a VGG-like network to predict task performance and generate saliency maps that can show brain regions important for task performance using three independent datasets.

Results: The DL generated saliency maps are consistent between each dataset.

Impact: We demonstrated that interpretable deep learning can be used as a reliable and generalizable tool to gain insight into brain regions whose activation impacts task performance.

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence

Motivation: Diseases related progress simulated through latent space image manipulation is difficult to interpret.

Goal(s): The goal was to develop an approach allowing for an improved interpretation of latent space image manipulation.

Approach: A StyleGAN model trained on MRI data from MS patients and healthy controls was used for image manipulation. The direction in latent space for generating images mimicking diseases progression towards MS was determined. The spatial changes were analyzed through eigenvalue decomposition.

Results: The decomposition approach revealed a pattern resembling a polynomial series, suggesting a parameterized data manipulation, with the second component being the most informative for illustrating disease related image changes.

Impact: The analysis method for disentanglement complex image changes through latent space manipulation offers improved predictive accuracy and enhances our understanding of disease progression in neuroimaging research by isolating disease-related image features with a parameter-free approach.

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence

Motivation: Macromolecular Proton Fraction quantification based on spin-lock MRI (MPF-SL) is a new technique for non-invasive imaging and characterization of macromolecule environment in tissues.

Goal(s): This study aims to develop an automated method for MPF quantification in the liver.

Approach: We present a deep learning framework for automated liver MPF quantification, incorporating an uncertainty-guided strategy for reliable region-of-interest (ROI) selection.

Results: Evaluation was conducted using clinical MPF data from 44 patients, demonstrating minimal error in MPF quantification and consistent and robust ROI selection. Our method shows promise in automated MPF measurement of the liver, offering both qualitative and quantitative evidence of its efficacy.

Impact: MPF-SL has been recently developed to measure macromolecule levels, showing potential in the non-invasive diagnosis of hepatic fibrosis.
This work automates MPF quantification using deep learning, showing the potential to decrease the cost of MPF-SL post-processing.

Keywords: Analysis/Processing, Quantitative Imaging, phase-cycled bSSFP, multi-parametric mapping, uncertainty quantification

Motivation: When using black-box regression models in diagnostic imaging, it is critical to quantify the uncertainty of such model predictions.

Goal(s): Implementing and understanding uncertainty quantification for multi-parametric quantitative MRI. Do prediction intervals reflect model uncertainty?

Approach: Train conditional quantile regression deep neural networks with subsequent conformalization steps for multi-parametric quantitative mapping without making distributional assumptions about the data.

Results: Conformalized relaxometry and magnetic field prediction intervals reflect model uncertainty. Conformalized quantile regression was successfully implemented and provides supportive information about intrinsic model uncertainty which is mandatory for clinical decision making.

Impact: A novel method for quantifying uncertainty of supervised machine learning models for multi-parametric quantitative MRI was successfully tested in silico and in vivo. Conformalized quantile regression allows prediction of confidence intervals without making assumptions about the training data distribution.

Keywords: Analysis/Processing, Brain

Motivation: Existing techniques for multi-parametric MR imaging-based brain tissue segmentation typically employ a generic feature combination strategy without incorporating task-specific guidance, making it challenging to ensure effective fusion.

Goal(s): In this work, we aim to develop a task-specific multi-parametric MR image fusion framework to enhance the brain tissue segmentation and quantification accuracy.

Approach: During preliminary experiments, we have identified a close correlation between prediction uncertainties and prediction errors. Therefore, we propose an uncertainty-guided task-specific multi-parametric MR image fusion framework to enhance fusion efficiency and decrease prediction uncertainty.

Results: Experiments on the iSeg-2019 dataset demonstrate that the proposed method achieves better results than existing techniques.

Impact: The outcome of this work has the potential to be utilized in clinical practice to help physicians better monitor brain development and diagnose brain diseases. Meanwhile, the framework can be extended to diverse fields where multi-modal image fusion is required.

Keywords: AI/ML Image Reconstruction, Elastography

Motivation: In Magnetic Resonance Elastography (MRE), accurate reconstruction of stiffness maps is essential for medical diagnosis. Traditional inversion techniques are limited by noise, discretization and/or low wavenumbers.

Goal(s): We aim to overcome these limitations using a neural network-based wave inversion (ElastoNet) with integrated uncertainty quantification ensuring reliable predictions with high detail resolution.

Approach: We trained ElastoNet on simulated wave patches. For inference, we combined all 3 motion encoding directions as input and used evidential deep learning as an uncertainty quantification method.

Results: ElastoNet achieves a substantial improvement in detail resolution compared to current neural network approaches and shows promising results in the low-frequency domain.

Impact: Our MR elastography neural network-based wave inversion is a promising method for enhanced accuracy and reliability in tissue property characterization. It effectively addresses challenges in reconstruction of stiffness maps, expanding the potential of MR elastography for medical diagnosis.

Keywords: Other AI/ML, Machine Learning/Artificial Intelligence, Explainability, interpretability

Motivation: The use of AI in clinical routine is often jeopardized by its lack of transparency. Explainable methods would help both clinicians and developers to identify model bias and interpret the automatic outputs.

Goal(s): We propose an explainable method providing insights into the decision process of an MS lesion segmentation network.

Approach: We adapt SmoothGrad to perform instance-level explanations and apply it to a U-Net, whose inputs are FLAIR and MPRAGE from 10 MS patients.

Results: Our saliency maps provide local-level information on the network's decisions. Predictions of the U-Net rely predominantly on lesions' voxel intensities in FLAIR and the amount of perilesional volume.

Impact: These results cast some light on the decision mechanisms of deep learning networks performing semantic segmentation. The acquired new knowledge can be an important step to facilitate AI integration into clinical practice.

Keywords: Analysis/Processing, Velocity & Flow, Plane reformatting, Deep reinforcement learning

Motivation: The standard approach for plane reformatting in 4D flow MRI is manual, leading to time-consuming and user-dependent results.

Goal(s): Our goal was to enhance plane reformatting in 4D flow MRI and overcome limitations associated with existing automated methods.

Approach: We introduce a novel approach that employs deep reinforcement learning (DRL) with a flexible coordinate system for precise and adaptable plane reformatting.

Results: Results demonstrate superior performance compared to baseline DRL and similar outcomes compared to those of landmark-based techniques, showing its potential for use in complex medical imaging scenarios beyond 4D flow MRI.

Impact: The proposed framework allows for automated, precise, and adaptive plane reformatting, facilitating the use of 4D flow MRI in clinical routines. It was trained with data sets from different vendors, making this approach widely applicable.