Bringing MRI to low- and middle-income countries: Directions, challenges and potential solutions

Predicting Rheumatoid Arthritis Development Using Hand Ultrasound and Machine Learning-A Two-Year Follow-Up Cohort Study

BACKGROUND

The early diagnosis and treatment of rheumatoid arthritis (RA) are essential to prevent joint damage and enhance patient outcomes. Diagnosing RA in its early stages is challenging due to the nonspecific and variable clinical signs and symptoms. Our study aimed to identify the most predictive features of hand ultrasound (US) for RA development and assess the performance of machine learning models in diagnosing preclinical RA.

METHODS

We conducted a prospective cohort study with 326 adults who had experienced hand joint pain for less than 12 months and no clinical arthritis. We assessed the participants clinically and via hand US at baseline and followed them for 24 months. Clinical progression to RA was defined according to the ACR/EULAR criteria. Regression modeling and machine learning approaches were used to analyze the predictive US features.

RESULTS

Of the 326 participants (45.10 ± 11.37 years/83% female), 123 (37.7%) developed clinical RA during follow-up. At baseline, 84.6% of the progressors had US synovitis, whereas 16.3% of the non-progressors did (p < 0.0001). Only 5.7% of the progressors had positive PD. Multivariate analysis revealed that the radiocarpal synovial thickness (OR = 39.8), PIP/MCP synovitis (OR = 68 and 39), and wrist effusion (OR = 12.56) on US significantly increased the odds of developing RA. ML confirmed these US features, along with the RF and anti-CCP levels, as the most important predictors of RA.

CONCLUSIONS

Hand US can identify preclinical synovitis and determine the RA risk. The radiocarpal synovial thickness, PIP/MCP synovitis, wrist effusion, and RF and anti-CCP levels are associated with RA development.

Clinical application of bladder MRI and the Vesical Imaging-Reporting and Data System

Diagnostic work-up and risk stratification in patients with bladder cancer before and after treatment must be refined to optimize management and improve outcomes. MRI has been suggested as a non-invasive technique for bladder cancer staging and assessment of response to systemic therapy. The Vesical Imaging-Reporting And Data System (VI-RADS) was developed to standardize bladder MRI image acquisition, interpretation and reporting and enables accurate prediction of muscle-wall invasion of bladder cancer. MRI is available in many centres but is not yet recommended as a first-line test for bladder cancer owing to a lack of high-quality evidence. Consensus-based evidence on the use of MRI-VI-RADS for bladder cancer care is needed to serve as a benchmark for formulating guidelines and research agendas until further evidence from randomized trials becomes available.

Comparison of image quality and diagnostic efficacy of routine clinical lumbar spine imaging at 0.55T and 1.5/3T

PURPOSE

To compare image quality, assess inter-reader variability, and evaluate the diagnostic efficacy of routine clinical lumbar spine sequences at 0.55T compared with those collected at 1.5/3T to assess common spine pathology.

METHODS

665 image series across 70 studies, collected at 0.55T and 1.5/3T, were assessed by two neuroradiology fellows for overall imaging quality (OIQ), artifacts, and accurate visualization of anatomical features (intervertebral discs, neural foramina, spinal cord, bone marrow, and conus / cauda equina nerve roots) using a 4-point Likert scale (1 = non-diagnostic to 4 = excellent). For the 0.55T scans, the most appropriate diagnosis(es) from a picklist of common spine pathologies was selected. The mean ± SD of all scores for all features for each sequence and reader at 0.55T and 1.5/3T were calculated. Paired t-tests (p ≤ 0.05) were used to compare ratings between field strengths. The inter-reader agreement was calculated using linear-weighted Cohen's Kappa coefficient (p ≤ 0.05). Unpaired VCG analysis for OIQ was additionally employed to represent differences between 0.55T and 1.5/3T (95 % CI).

RESULTS

All sequences at 0.55T were rated as acceptable (≥2) for diagnostic use by both readers despite significantly lower scores for some compared to those at 1.5/3T. While there was low inter-reader agreement on individual scores, the agreement on the diagnosis was high, demonstrating the potential of this system for detecting routine spine pathology.

CONCLUSIONS

Clinical lumbar spine imaging at 0.55T produces diagnostic-quality images demonstrating the feasibility of its use in diagnosing spinal pathology, including osteomyelitis/discitis, post-surgical changes with complications, and metastatic disease.

Deep-learning-based reconstruction of undersampled MRI to reduce scan times: a multicentre, retrospective, cohort study

BACKGROUND

The extended acquisition times required for MRI limit its availability in resource-constrained settings. Consequently, accelerating MRI by undersampling k-space data, which is necessary to reconstruct an image, has been a long-standing but important challenge. We aimed to develop a deep convolutional neural network (dCNN) optimisation method for MRI reconstruction and to reduce scan times and evaluate its effect on image quality and accuracy of oncological imaging biomarkers.

METHODS

In this multicentre, retrospective, cohort study, MRI data from patients with glioblastoma treated at Heidelberg University Hospital (775 patients and 775 examinations) and from the phase 2 CORE trial (260 patients, 1083 examinations, and 58 institutions) and the phase 3 CENTRIC trial (505 patients, 3147 examinations, and 139 institutions) were used to develop, train, and test dCNN for reconstructing MRI from highly undersampled single-coil k-space data with various acceleration rates (R=2, 4, 6, 8, 10, and 15). Independent testing was performed with MRIs from the phase 2/3 EORTC-26101 trial (528 patients with glioblastoma, 1974 examinations, and 32 institutions). The similarity between undersampled dCNN-reconstructed and original MRIs was quantified with various image quality metrics, including structural similarity index measure (SSIM) and the accuracy of undersampled dCNN-reconstructed MRI on downstream radiological assessment of imaging biomarkers in oncology (automated artificial intelligence-based quantification of tumour burden and treatment response) was performed in the EORTC-26101 test dataset. The public NYU Langone Health fastMRI brain test dataset (558 patients and 558 examinations) was used to validate the generalisability and robustness of the dCNN for reconstructing MRIs from available multi-coil (parallel imaging) k-space data.

FINDINGS

In the EORTC-26101 test dataset, the median SSIM of undersampled dCNN-reconstructed MRI ranged from 0·88 to 0·99 across different acceleration rates, with 0·92 (95% CI 0·92-0·93) for 10-times acceleration (R=10). The 10-times undersampled dCNN-reconstructed MRI yielded excellent agreement with original MRI when assessing volumes of contrast-enhancing tumour (median DICE for spatial agreement of 0·89 [95% CI 0·88 to 0·89]; median volume difference of 0·01 cm3 [95% CI 0·00 to 0·03] equalling 0·21%; p=0·0036 for equivalence) or non-enhancing tumour or oedema (median DICE of 0·94 [95% CI 0·94 to 0·95]; median volume difference of -0·79 cm3 [95% CI -0·87 to -0·72] equalling -1·77%; p=0·023 for equivalence) in the EORTC-26101 test dataset. Automated volumetric tumour response assessment in the EORTC-26101 test dataset yielded an identical median time to progression of 4·27 months (95% CI 4·14 to 4·57) when using 10-times-undersampled dCNN-reconstructed or original MRI (log-rank p=0·80) and agreement in the time to progression in 374 (95·2%) of 393 patients with data. The dCNN generalised well to the fastMRI brain dataset, with significant improvements in the median SSIM when using multi-coil compared with single-coil k-space data (p<0·0001).

INTERPRETATION

Deep-learning-based reconstruction of undersampled MRI allows for a substantial reduction of scan times, with a 10-times acceleration demonstrating excellent image quality while preserving the accuracy of derived imaging biomarkers for the assessment of oncological treatment response. Our developments are available as open source software and hold considerable promise for increasing the accessibility to MRI, pending further prospective validation.

FUNDING

Deutsche Forschungsgemeinschaft (German Research Foundation) and an Else Kröner Clinician Scientist Endowed Professorship by the Else Kröner Fresenius Foundation.

Setting up a biomodeling, virtual planning, and three-dimensional printing service in Uruguay

Virtual surgical planning and three-dimensional (D) printing are rapidly becoming essential for challenging and complex surgeries around the world. An Ibero-American survey reported a lack of awareness of technology benefits and scarce financial resources as the two main barriers to widespread adoption of 3-D technologies. The Pereira Rossell Hospital Center is a publicly funded maternal and pediatric academic clinical center in Uruguay, a low-resource Latin American country, that successfully created and has been running a 3-D unit for 4 years. The present work is a step-by-step review of the 3-D technology implementation process in a hospital with minimal financial investment. References to training, software, hardware, and the management of human resources are included. Difficulties throughout the process and future challenges are also discussed.

Multi-contrast high-field quality image synthesis for portable low-field MRI using generative adversarial networks and paired data

Introduction

Portable low-field strength (64mT) MRI scanners promise to increase access to neuroimaging for clinical and research purposes, however these devices produce lower quality images compared to high-field scanners. In this study, we developed and evaluated a deep learning architecture to generate high-field quality brain images from low-field inputs using a paired dataset of multiple sclerosis (MS) patients scanned at 64mT and 3T.

Methods

A total of 49 MS patients were scanned on portable 64mT and standard 3T scanners at Penn (n=25) or the National Institutes of Health (NIH, n=24) with T1-weighted, T2-weighted and FLAIR acquisitions. Using this paired data, we developed a generative adversarial network (GAN) architecture for low- to high-field image translation (LowGAN). We then evaluated synthesized images with respect to image quality, brain morphometry, and white matter lesions.

Results

Synthetic high-field images demonstrated visually superior quality compared to low-field inputs and significantly higher normalized cross-correlation (NCC) to actual high-field images for T1 (p=0.001) and FLAIR (p<0.001) contrasts. LowGAN generally outperformed the current state-of-the-art for low-field volumetrics. For example, thalamic, lateral ventricle, and total cortical volumes in LowGAN outputs did not differ significantly from 3T measurements. Synthetic outputs preserved MS lesions and captured a known inverse relationship between total lesion volume and thalamic volume.

Conclusions

LowGAN generates synthetic high-field images with comparable visual and quantitative quality to actual high-field scans. Enhancing portable MRI image quality could add value and boost clinician confidence, enabling wider adoption of this technology.

Hypertrophic Cardiomyopathy Diagnosis and Treatment in High- and Low-Income Countries: A Narrative Review

Hypertrophic cardiomyopathy (HCM) is a hereditary cardiac condition characterized by unexplained left ventricular hypertrophy without a hemodynamic cause. This condition is prevalent in the United States, resulting in various clinical manifestations, including diastolic dysfunction, left ventricular outflow obstruction, cardiac ischemia, and atrial fibrillation. HCM is associated with several genetic mutations, with sarcomeric mutations being the most common and contributing to a more complex disease course. Early diagnosis of HCM is essential for effective management, as late diagnosis often requires invasive treatments and creates a substantial financial burden. Disparities in HCM diagnosis and treatment exist between high-income and low-income countries. High-income countries have more resources to investigate and implement advanced diagnostic and treatment modalities. In contrast, low-income countries face challenges in accessing diagnostic equipment, trained personnel, and affordable medications, leading to a lower quality of life and life expectancy for affected individuals. Diagnostic tools for HCM include imaging studies such as 2D echocardiography, cardiovascular magnetic resonance (CMR), and electrocardiograms (ECGs). CMR is considered the gold standard but remains inaccessible to a significant portion of the world's population, especially in low-income countries. Genetics plays a crucial role in HCM, with numerous mutations identified in various genes. Genetic counseling is essential but often limited in low-income countries due to resource constraints. Disparities in healthcare access and adherence to treatment recommendations exist between high-income and low-income countries, leading to differences in patient outcomes. Addressing these disparities is essential to improve the overall management of HCM on a global scale. In conclusion, this review highlights the complex nature of HCM, emphasizing the importance of early diagnosis, genetic counseling, and access to appropriate diagnostic and therapeutic interventions. Addressing healthcare disparities is crucial to ensure that all individuals with HCM receive timely and effective care, regardless of their geographic location or socioeconomic status.