Diagnostic accuracy of semiautomatic lesion detection plus quantitative susceptibility mapping in the identification of new and enhancing multiple sclerosis lesions

The changes of oxygen extraction fraction in different types of lesions in relapsing-remitting multiple sclerosis: A cross-sectional and longitudinal study

OBJECTIVES

To explore the oxygen metabolism level of different types of lesions in relapsing-remitting multiple sclerosis (RRMS) patients by oxygen extraction fraction (OEF) both cross-sectionally and longitudinally.

METHODS

Forty-six RRMS patients and forty-one healthy controls (HC) went MRI examination. The quantitative susceptibility mapping (QSM) and OEF map were reconstructed from a 3D multi-echo gradient echo sequence. MS lesions in white matter were classified as contrast-enhancing lesions (CELs) on post-gadolinium T1-weighted sequence, paramagnetic rim lesions (PRLs), hyperintense lesions and non-hyperintense lesions on QSM, respectively. The susceptibility and OEF of different types of lesions were compared. The susceptibility and OEF values were measured and compared among different types of lesions. Among these RRMS patients, seventeen had follow-up MRI and 232 lesions, and baseline to follow-up longitudinal changes in susceptibility and OEF were measured.

RESULTS

PRLs had higher susceptibility and lower OEF than CELs, hyperintense lesions, and non-hyperintense lesions. The hyperintense lesions had higher susceptibility and lower OEF than non-hyperintense lesions. In longitudinal changes, PRLs had susceptibility increased (P < 0.001) and OEF decreased (P < 0.001). The hyperintense lesions showed significant decreases in susceptibility (P = 0.020), and non-hyperintense lesions showed significant increases in OEF during follow-up (P = 0.005). Notably, hyperintense lesions may convert to PRLs or non-hyperintense lesions as time progresses, accompanied by changes of OEF and susceptibility in the lesions.

CONCLUSION

This study revealed tissue damage and oxygen metabolism level in different types of MS lesions. The OEF may contribute to further understanding the evolution of MS lesions.

Evaluation of the Statistical Detection of Change Algorithm for Screening Patients with MS with New Lesion Activity on Longitudinal Brain MRI

BACKGROUND AND PURPOSE

Identification of new MS lesions on longitudinal MR imaging by human readers is time-consuming and prone to error. Our objective was to evaluate the improvement in the performance of subject-level detection by readers when assisted by the automated statistical detection of change algorithm.

MATERIALS AND METHODS

A total of 200 patients with MS with a mean interscan interval of 13.2 (SD, 2.4) months were included. Statistical detection of change was applied to the baseline and follow-up FLAIR images to detect potential new lesions for confirmation by readers (Reader + statistical detection of change method). This method was compared with readers operating in the clinical workflow (Reader method) for a subject-level detection of new lesions.

RESULTS

Reader + statistical detection of change found 30 subjects (15.0%) with at least 1 new lesion, while Reader detected 16 subjects (8.0%). As a subject-level screening tool, statistical detection of change achieved a perfect sensitivity of 1.00 (95% CI, 0.88-1.00) and a moderate specificity of 0.67 (95% CI, 0.59-0.74). The agreement on a subject level was 0.91 (95% CI, 0.87-0.95) between Reader + statistical detection of change and Reader, and 0.72 (95% CI, 0.66-0.78) between Reader + statistical detection of change and statistical detection of change.

CONCLUSIONS

The statistical detection of change algorithm can serve as a time-saving screening tool to assist human readers in verifying 3D FLAIR images of patients with MS with suspected new lesions. Our promising results warrant further evaluation of statistical detection of change in prospective multireader clinical studies.

Overview of quantitative susceptibility mapping using deep learning: Current status, challenges and opportunities

Quantitative susceptibility mapping (QSM) has gained broad interest in the field by extracting bulk tissue magnetic susceptibility, predominantly determined by myelin, iron and calcium from magnetic resonance imaging (MRI) phase measurements in vivo. Thereby, QSM can reveal pathological changes of these key components in a variety of diseases. QSM requires multiple processing steps such as phase unwrapping, background field removal and field-to-source inversion. Current state-of-the-art techniques utilize iterative optimization procedures to solve the inversion and background field correction, which are computationally expensive and require a careful choice of regularization parameters. With the recent success of deep learning using convolutional neural networks for solving ill-posed reconstruction problems, the QSM community also adapted these techniques and demonstrated that the QSM processing steps can be solved by efficient feed forward multiplications not requiring either iterative optimization or the choice of regularization parameters. Here, we review the current status of deep learning-based approaches for processing QSM, highlighting limitations and potential pitfalls, and discuss the future directions the field may take to exploit the latest advances in deep learning for QSM.

Subsecond accurate myelin water fraction reconstruction from FAST-T2 data with 3D UNET

PURPOSE

To develop a 3D UNET convolutional neural network for rapid extraction of myelin water fraction (MWF) maps from six-echo fast acquisition with spiral trajectory and T₂ -prep data and to evaluate its accuracy in comparison with multilayer perceptron (MLP) network.

METHODS

The MWF maps were extracted from 138 patients with multiple sclerosis using an iterative three-pool nonlinear least-squares algorithm (NLLS) without and with spatial regularization (srNLLS), which were used as ground-truth labels to train, validate, and test UNET and MLP networks as a means to accelerate data fitting. Network testing was performed in 63 patients with multiple sclerosis and a numerically simulated brain phantom at SNR of 200, 100 and 50.

RESULTS

Simulations showed that UNET reduced the MWF mean absolute error by 30.1% to 56.4% and 16.8% to 53.6% over the whole brain and by 41.2% to 54.4% and 21.4% to 49.4% over the lesions for predicting srNLLS and NLLS MWF, respectively, compared to MLP, with better performance at lower SNRs. UNET also outperformed MLP for predicting srNLLS MWF in the in vivo multiple-sclerosis brain data, reducing mean absolute error over the whole brain by 61.9% and over the lesions by 67.5%. However, MLP yielded 41.1% and 51.7% lower mean absolute error for predicting in vivo NLLS MWF over the whole brain and the lesions, respectively, compared with UNET. The whole-brain MWF processing time using a GPU was 0.64 seconds for UNET and 0.74 seconds for MLP.

CONCLUSION

Subsecond whole-brain MWF extraction from fast acquisition with spiral trajectory and T₂ -prep data using UNET is feasible and provides better accuracy than MLP for predicting MWF output of srNLLS algorithm.

Identification of patients with relapsing multiple sclerosis eligible for high-efficacy therapies

Relapsing multiple sclerosis (RMS) presents a highly variable clinical evolution among patients, and its management should be personalized. Although there is no cure at present, effective disease-modifying therapies (DMTs) are available. Selection of the most appropriate DMT for each patient is influenced by several clinical, radiological and demographic aspects as well as personal preferences that, at times, are not covered in the regulatory criteria. This may be a source of difficulty, especially in certain situations where so-called 'high-efficacy DMTs' (usually considered second-line) could be of greater benefit to the patient. In this narrative review, we discuss evidence and experience, and propose a pragmatic guidance on decision-making with respect to the indication and management of high-efficacy DMT in adult patients with RMS based on expert opinion.

Heterogeneity of multiple sclerosis lesions in fast diffusional kurtosis imaging

Background

Mean kurtosis (MK), one of the parameters derived from diffusion kurtosis imaging (DKI), has shown increased sensitivity to tissue microstructure damage in several neurological disorders.

Methods

Thirty-seven patients with relapsing-remitting MS and eleven healthy controls (HC) received brain imaging on a 3T MR scanner, including a fast DKI sequence. MK and mean diffusivity (MD) were measured in the white matter of HC, normal-appearing white matter (NAWM) of MS patients, contrast-enhancing lesions (CE-L), FLAIR lesions (FLAIR-L) and black holes (BH).

Results

Overall 1529 lesions were analyzed, including 30 CE-L, 832 FLAIR-L and 667 BH. Highest MK values were obtained in the white matter of HC (0.814 ± 0.129), followed by NAWM (0.724 ± 0.137), CE-L (0.619 ± 0.096), FLAIR-L (0.565 ± 0.123) and BH (0.549 ± 0.12). Lowest MD values were obtained in the white matter of HC (0.747 ± 0.068 10⁻³mm²/sec), followed by NAWM (0.808 ± 0.163 10⁻³mm²/sec), CE-L (0.853 ± 0.211 10⁻³mm²/sec), BH (0.957 ± 0.304 10⁻³mm²/sec) and FLAIR-L (0.976 ± 0.35 10⁻³mm²/sec). While MK differed significantly between CE-L and non-enhancing lesions, MD did not.

Conclusion

MK adds predictive value to differentiate between MS lesions and might provide further information about diffuse white matter injury and lesion microstructure.

Quantitative susceptibility-weighted imaging in predicting disease activity in multiple sclerosis

PURPOSE

Repeated use of Gadolinium (Gd) contrast for multiple sclerosis (MS) imaging leads to Gd deposition in brain. We aimed to study the utility of phase values by susceptibility weighted imaging (SWI) to assess the iron content in MS lesions to differentiate active and inactive lesions.

METHODS

MS persons who underwent MRI were grouped into group 1 with active lesions and group 2 with inactive lesions based on the presence or absence of contrast enhancing lesions. Phase values of lesions (P_L) and contralateral normal white matter (P_N) were calculated using the SPIN software by drawing ROI. Subtracted phase values (P_S = P_L - P_N) and iron content (P_S/3) of the lesions were calculated in both groups.

RESULTS

We analyzed 69 enhancing lesions from 22 patients (group 1) and 84 non-enhancing lesions from 29 patients (group 2). Mean-subtracted phase values and iron content corrected for voxels in ROI were significantly lower in enhancing lesions compared to non-enhancing lesions (p < 0.001). A cut-off value 2.8 μg/g for iron content showed area under the curve of 0.909 with good sensitivity.

CONCLUSION

Quantification of iron content using SWI phase values holds promise as a biomarker to differentiate active from inactive lesions of MS.

The emerging role of artificial intelligence in multiple sclerosis imaging

BACKGROUND

Computer-aided diagnosis can facilitate the early detection and diagnosis of multiple sclerosis (MS) thus enabling earlier interventions and a reduction in long-term MS-related disability. Recent advancements in the field of artificial intelligence (AI) have led to the improvements in the classification, quantification and identification of diagnostic patterns in medical images for a range of diseases, in particular, for MS. Importantly, data generated using AI techniques are analyzed automatically, which compares favourably with labour-intensive and time-consuming manual methods.

OBJECTIVE

The aim of this review is to assist MS researchers to understand current and future developments in the AI-based diagnosis and prognosis of MS.

METHODS

We will investigate a variety of AI approaches and various classifiers and compare the current state-of-the-art techniques in relation to lesion segmentation/detection and prognosis of disease. After briefly describing the magnetic resonance imaging (MRI) techniques commonly used, we will describe AI techniques used for the detection of lesions and MS prognosis.

RESULTS

We then evaluate the clinical maturity of these AI techniques in relation to MS.

CONCLUSION

Finally, future research challenges are identified in a bid to encourage further improvements of the methods.

Estimation of Multiple Sclerosis lesion age on magnetic resonance imaging

We introduce the first-ever statistical framework for estimating the age of Multiple Sclerosis (MS) lesions from magnetic resonance imaging (MRI). Estimating lesion age is an important step when studying the longitudinal behavior of MS lesions and can be used in applications such as studying the temporal dynamics of chronic active MS lesions. Our lesion age estimation models use first order radiomic features over a lesion derived from conventional T1 (T1w) and T2 weighted (T2w) and fluid attenuated inversion recovery (FLAIR), T1w with gadolinium contrast (T1w+c), and Quantitative Susceptibility Mapping (QSM) MRI sequences as well as demographic information. For this analysis, we have a total of 32 patients with 53 new lesions observed at 244 time points. A one or two step random forest model for lesion age is fit on a training set using a lesion volume cutoff of 15 mm³ or 50 mm³ . We explore the performance of nine different modeling scenarios that included various combinations of the MRI sequences and demographic information and a one or two step random forest models, as well as simpler models that only uses the mean radiomic feature from each MRI sequence. The best performing model on a validation set is a model that uses a two-step random forest model on the radiomic features from all of the MRI sequences with demographic information using a lesion volume cutoff of 50 mm³ . This model has a mean absolute error of 7.23 months (95% CI: [6.98, 13.43]) and a median absolute error of 5.98 months (95% CI: [5.26, 13.25]) in the validation set. For this model, the predicted age and actual age have a statistically significant association (p-value <0.001) in the validation set.

Clinical Integration of Automated Processing for Brain Quantitative Susceptibility Mapping: Multi-Site Reproducibility and Single-Site Robustness

BACKGROUND AND PURPOSE

Quantitative susceptibility mapping (QSM) of the brain has become highly reproducible and has applications in an expanding array of diseases. To translate QSM from bench to bedside, it is important to automate its reconstruction immediately after data acquisition. In this work, a server system that automatically reconstructs QSM and exchange images with the scanner using the DICOM standard is demonstrated using a multi-site, multi-vendor reproducibility study and a large, single-site, multi-scanner image quality review study in a clinical environment.

METHODS

A single healthy subject was scanned with a 3D multi-echo gradient echo sequence at nine sites around the world using scanners from three manufacturers. A high-resolution (HiRes, .5 × .5 × 1 mm³ reconstructed) and standard-resolution (StdRes, .5 × .5 × 3 mm³ ) protocol was performed. ROI analysis of various white matter and gray matter regions was performed to investigate reproducibility across sites. At one institution, a retrospective multi-scanner image quality review was carried out of all clinical QSM images acquired consecutively in 1 month.

RESULTS

Reconstruction times using a GPU were 29 ± 22 seconds (StdRes) and 55 ± 39 seconds (HiRes). ROI standard deviation across sites was below 24 ppb (StdRes) and 17 ppb (HiRes). Correlations between ROI averages across sites were on average .92 (StdRes) and .96 (HiRes). Image quality review of 873 consecutive patients revealed diagnostic or excellent image quality in 96% of patients.

CONCLUSION

Online QSM reconstruction for a variety of sites and scanner platforms with low cross-site ROI standard deviation is demonstrated. Image quality review revealed diagnostic or excellent image quality in 96% of 873 patients.

Quantitative Susceptibility Mapping of Time-Dependent Susceptibility Changes in Multiple Sclerosis Lesions

BACKGROUND AND PURPOSE

MR imaging studies have demonstrated that magnetic susceptibility in multiple sclerosis lesions is dependent on lesion age. The objective of this study was to use quantitative susceptibility mapping to determine whether lesions with a hyperintense rim, indicative of iron-laden inflammatory cells (rim+), follow a unique time-dependent trajectory of susceptibility change compared with those without (rim-).

MATERIALS AND METHODS

We studied patients with MS with at least 1 new gadolinium-enhancing lesion and at least 3 longitudinal quantitative susceptibility mapping scans obtained between 1.1 and 6.1 years. Lesions were classified as rim+ if a hyperintense rim appeared on quantitative susceptibility mapping at any time. A multilevel growth curve model compared longitudinal susceptibility among rim+ and rim- lesions.

RESULTS

Thirty-two new gadolinium-enhancing lesions from 19 patients with MS were included, and 16 lesions (50%) were identified as rim+. Quantitative susceptibility mapping rim+ lesions were larger than rim- lesions with gadolinium enhancement (P < .001). Among all lesions, susceptibility increased sharply after enhancement to a peak between 1 and 2 years followed by a decrease. The overall susceptibility curve height for rim- lesions was 4.27 parts per billion lower than that for rim+ lesions (P = .01). Rim- lesions demonstrated a higher linear slope relative to rim+ lesions (P = .023) but faster cubic decay relative to rim+ lesions (P = .005). Rim- lesions started decaying approximately 2 years earlier compared with rim+ lesions.

CONCLUSIONS

There was a marked difference in the susceptibility temporal trajectory between rim+ and rim- lesions during the first 6 years of lesion formation. Most rim+ lesions retain iron for years after the initial lesion appearance.

Clinical feasibility of brain quantitative susceptibility mapping

PURPOSE

To evaluate the quality of brain quantitative susceptibility mapping (QSM) that is fully automatically reconstructed in clinical MRI of various neurological diseases.

METHODS

393 consecutive patients in one month were recruited for this evaluation study. QSM was reconstructed using Morphology Enabled Dipole Inversion without zero reference regularization (MEDI) and using MEDI with cerebrospinal fluid automatic zero-reference regularization to generate susceptibility values (MEDI+0). Two neuroradiologists independently assessed the image quality of MEDI+0 and MEDI and image concordance between them. Lesion susceptibility values were measured in 20 cases of glioma, 21 cases of ischemic stroke and 43 multiple sclerosis (MS) cases on both MEDI+0 and MEDI images.

RESULTS

The two neuroradiologists rated the MEDI+0 image qualities of the 393 cases as 351 (89.3%) and 362 (92.1%) excellent, 29 (7.4%) and 24 (6.1%) diagnostic, and 13 (3.3%) and 7 (1.8%) poor, and scored the concordances between MEDI+0 and MEDI as 364 (92.6%) and 351 (89.3%) excellent, 13 (3.3%) and 31 (7.9%) good, 14 (3.6%) and 9 (2.3%) intermediate, 2 (0.5%) and 2 (0.5%) poor, and 0 (0%) and 0 (0%) none. There was good correlation between MEDI+0 and MEDI in lesion susceptibility contrast of glioma, ischemic stroke, and MS cases (all p < 0.05). The MS lesion susceptibility time course from this patient cohort was found to be similar to the reported pattern: isointense initially for acute enhancing lesions, and hyperintense over the following years for active chronic lesions.

CONCLUSION

Brain QSM images of various neurological diseases have reliable diagnostic quality in clinical MRI, with MEDI+0 providing susceptibility values automatically referenced to CSF in longitudinal and cross-center studies.

Patents on Quantitative Susceptibility Mapping (QSM) of Tissue Magnetism

BACKGROUND

Quantitative susceptibility mapping (QSM) depicts biodistributions of tissue magnetic susceptibility sources, including endogenous iron and calcifications, as well as exogenous paramagnetic contrast agents and probes. When comparing QSM with simple susceptibility weighted MRI, QSM eliminates blooming artifacts and shows reproducible tissue susceptibility maps independent of field strength and scanner manufacturer over a broad range of image acquisition parameters. For patient care, QSM promises to inform diagnosis, guide surgery, gauge medication, and monitor drug delivery. The Bayesian framework using MRI phase data and structural prior knowledge has made QSM sufficiently robust and accurate for routine clinical practice.

OBJECTIVE

To address the lack of a summary of US patents that is valuable for QSM product development and dissemination into the MRI community.

METHOD

We searched the USPTO Full-Text and Image Database for patents relevant to QSM technology innovation. We analyzed the claims of each patent to characterize the main invented method and we investigated data on clinical utility.

RESULTS

We identified 17 QSM patents; 13 were implemented clinically, covering various aspects of QSM technology, including the Bayesian framework, background field removal, numerical optimization solver, zero filling, and zero-TE phase.

CONCLUSION

Our patent search identified patents that enable QSM technology for imaging the brain and other tissues. QSM can be applied to study a wide range of diseases including neurological diseases, liver iron disorders, tissue ischemia, and osteoporosis. MRI manufacturers can develop QSM products for more seamless integration into existing MRI scanners to improve medical care.