Improved cerebral microbleeds detection using their magnetic signature on T2*-phase-contrast: A comparison study in a clinical setting

Cerebral small vessel disease modifies outcomes after minimally invasive surgery for intracerebral haemorrhage

BACKGROUND

Minimally invasive surgery (MIS) for spontaneous supratentorial intracerebral haemorrhage (ICH) is controversial but may be beneficial if end-of-treatment (EOT) haematoma volume is reduced to ≤15 mL. We explored whether MRI findings of cerebral small vessel disease (CSVD) modify the effect of MIS on long-term outcomes.

METHODS

Prespecified blinded subgroup analysis of 288 subjects with qualified imaging sequences from the phase 3 Minimally Invasive Surgery Plus Alteplase for Intracerebral Haemorrhage Evacuation (MISTIE) trial. We tested for heterogeneity in the effects of MIS and MIS+EOT volume ≤15 mL on the trial's primary outcome of good versus poor function at 1 year by the presence of single CSVD features and CSVD scores using multivariable models.

RESULTS

Of 499 patients enrolled in MISTIE III, 288 patients had MRI, 149 (51.7%) randomised to MIS and 139 (48.3%) to standard medical care (SMC). Median (IQR) ICH volume was 42 (30-53) mL. In the full MRI cohort, there was no statistically significant heterogeneity in the effects of MIS versus SMC on 1-year outcomes by any specific CSVD feature or by CSVD scores (all Pinteraction >0.05). In 94 MIS patients with EOT ICH volume ≤15 mL, significant reduction in odds of poor outcome was found with cerebral amyloid angiopathy score <2 (OR, 0.14 (0.05-0.42); Pinteraction=0.006), absence of lacunes (OR, 0.37 (0.18-0.80); Pinteraction=0.02) and absence of severe white matter hyperintensities (WMHs) (OR, 0.22 (0.08-0.58); Pinteraction=0.03).

CONCLUSIONS

Following successful haematoma reduction by MIS, we found significantly lower odds of poor functional outcome with lower total burden of CSVD in addition to absence of lacunes and severe WMHs. CSVD features may have utility for prognostication and patient selection in clinical trials of MIS.

Segmented 3D Echo Planar Acquisition for Rapid Susceptibility-Weighted Imaging: Application to Microhemorrhage Detection in Traumatic Brain Injury

BACKGROUND

Susceptibility-weighted imaging (SWI) provides superior image contrast of cerebral microhemorrhages (CMBs). It is based on a three-dimensional (3D) gradient echo (GRE) sequence with a relatively long imaging time.

PURPOSE

To evaluate whether an accelerated 3D segmented echo planar imaging SWI is comparable to GRE SWI in detecting CMBs in traumatic brain injury (TBI).

STUDY TYPE

Prospective.

SUBJECTS

Four healthy volunteers and 46 consecutive subjects (38.0 ± 14.4 years, 16 females; 12 mild, 13 moderate, and 7 severe TBI).

FIELD STRENGTH/SEQUENCE

A 3 T scanner/3D gradient echo and 3D segmented echo planar imaging (segEPI).

ASSESSMENT

Brain images were acquired using GRE and segEPI in a single session (imaging time = 9 minutes 47 seconds and 1 minute 30 seconds, respectively). The signal-to-noise ratio (SNR) calculated from healthy volunteer thalamus and centrum semiovale were compared. CMBs were counted by three raters blinded to diagnostic information.

STATISTICAL TESTS

A t-test was used to assess SNR difference. Pearson correlation and Wilcoxon signed-rank test were performed using CMB counts. The intermethod agreement was evaluated using Bland-Altman method. Intermethod and interrater reliabilities of image-based diffuse axonal injury (DAI) diagnoses were evaluated using Cohen's kappa and percent agreement. P ≤ 0.05 was considered statistically significant.

RESULTS

Thalamus SNRs were 16.9 ± 2.2 and 16.5 ± 3 for GRE and segEPI (P = 0.84), respectively. Centrum semiovale SNRs were 25.8 ± 4.6 and 21.1 ± 2.7 (P = 0.13). The correlation coefficient of CMBs was 0.93, and differences were not significant (P = 0.56-0.85). For DAI diagnoses, Cohen's kappa was 0.62-0.84 and percent agreement was 85%-94%.

DATA CONCLUSION

CMB counts on segEPI and GRE were highly correlated, and DAI diagnosis was made equally effectively. segEPI SWI can potentially replace GRE SWI in detecting TBI CMBs, especially when time constraints are critical.

EVIDENCE LEVEL

1 TECHNICAL EFFICACY: Stage 2.

New Remote Cerebral Microbleeds on T2*-Weighted Echo Planar MRI After Intravenous Thrombolysis for Acute Ischemic Stroke

Background

The main and well-defined complication of intravenous administration of recombinant tissue plasminogen activator (tPA) in patients with acute ischemic stroke (AIS) is symptomatic intracranial hemorrhage (sICH). However, rtPA might also be connected with the formation of cerebral microbleeds (CMBs), located remotely from the ischemic lesions, that may remain clinically silent. This association might be important because the load of CMBs has been associated with cognitive impairment. We investigated whether administration of rtPA in AIS results in the appearance of new CMBs and if the initial load of CMBs is associated with hemorrhagic transformation.

Methods

A total of fifty-nine consecutive patients with AIS treated with rtPA underwent MRI including T2^*-weighted Echo Planar Imaging (T2^*-EPI) shortly before and 7–9 days after rtPA administration. We calculated the load of new CMBs located outside the MR diffusion restriction area in the follow-up imaging and assessed hemorrhagic transformation with ECASS-II scoring.

Results

A total of forty-nine patients were included for the final analysis. On initial T2^*-EPI-GRE, 37 baseline microbleeds (CMBs) were observed in 14 patients (28.6%). On follow-up T2^*-EPI-GRE amount of CMBs increased to a total number of 103. New CMBs were found in 5 (14.3%) of 35 patients without and in 9 (64.3%) of 14 with any baseline CMBs. Multiple logistic regression analysis indicated that presence of baseline CMBs (risk ratio [RR] 5.95, 95% CI 2.69–13.20, p < 0.001) and lower platelets level (risk ratio [RR] 0.992, 95% CI 0.986–0.998, p = 0.007) were independently associated with new CMBs. The baseline load of CMBs was not associated with the risk of hemorrhagic transformation.

Conclusion

In this study, new CMBs were found in nearly 30% of patients with AIS on the 7–9 days after rtPA treatment. Baseline CMBs correlated with a higher risk of new CMBs appearing after the rtPA treatment, independently of other factors. At the same time, in our sample, baseline CMBs did not correlate with an increased risk of hemorrhagic transformation. Since the associations between the CMBs load and cognitive impairment have already been proved, further studies are warranted to investigate possible associations between the thrombolytic treatment of patients with AIS, mainly those with baseline CMBs, and the risk of earlier cognitive decline.

Cerebral Microbleed Detection via Convolutional Neural Network and Extreme Learning Machine

Aim: Cerebral microbleeds (CMBs) are small round dots distributed over the brain which contribute to stroke, dementia, and death. The early diagnosis is significant for the treatment. Method: In this paper, a new CMB detection approach was put forward for brain magnetic resonance images. We leveraged a sliding window to obtain training and testing samples from input brain images. Then, a 13-layer convolutional neural network (CNN) was designed and trained. Finally, we proposed to utilize an extreme learning machine (ELM) to substitute the last several layers in the CNN for detection. We carried out an experiment to decide the optimal number of layers to be substituted. The parameters in ELM were optimized by a heuristic algorithm named bat algorithm. The evaluation of our approach was based on hold-out validation, and the final predictions were generated by averaging the performance of five runs. Results: Through the experiments, we found replacing the last five layers with ELM can get the optimal results. Conclusion: We offered a comparison with state-of-the-art algorithms, and it can be revealed that our method was accurate in CMB detection.

Clinical Integration of Quantitative Susceptibility Mapping Magnetic Resonance Imaging into Neurosurgical Practice

OBJECTIVE

To introduce quantitative susceptibility mapping (QSM), a novel magnetic resonance imaging sequence, to the field of neurosurgery.

METHODS

QSM is introduced both in its historical context and by providing a brief overview of the physics behind the technique tailored to a neurosurgical audience. Its application to clinical neurosurgery is then highlighted using case examples.

RESULTS

QSM offers a quantitative assessment of susceptibility (previously considered as an artifact) via a single, straightforward gradient echo acquisition. QSM differs from standard susceptibility weighted imaging in its ability to both quantify and precisely localize susceptibility effects. Clinical applications of QSM are wide reaching and include precise localization of the deep nuclei for deep brain stimulation electrode placement, differentiation between blood products and calcification within brain lesions, and enhanced sensitivity of cerebral micrometastasis identification.

CONCLUSIONS

We present this diverse range of QSM's clinical applications to neurosurgical care via case examples. QSM can be obtained in all patients able to undergo magnetic resonance imaging and is easily integratable into busy neuroradiology programs because of its short acquisition time and straightforward, automated offline postprocessing workflow. Clinical integration of QSM may help clinicians better identify and characterize neurosurgical lesions, thereby improving patient care.