3328

Diagnostic performance of a new multicontrast one minute full brain exam (EPIMix) in neuroradiology
Anna Falk Delgado1, Annika Kits2, Jessica Bystam2, Magnus Kaijser2, Mikael Skorpil2, Tim Sprenger2,3, and Stefan Skare2

1Clinical Neuroscience, Karolinska, Stockholm, Sweden, 2Karolinska, Stockholm, Sweden, 3GE Healthcare, Stockholm, Sweden

Synopsis

This study assess if a new one minute multi contrast MRI method has comparable diagnostic performance as conventional MRI in brain imaging. Prospectively consecutively included patients (n=101) underwent a conventional clinical brain MRI in addition to a new 80 seconds multi contrast MRI sequence named EPIMix. The diagnostic performance to categorize a clinical brain MRI scan with EPIMix or conventional MRI as abnormal (symptom-causing lesion) was comparable between EPIMix (AUC 0.97 (95%CI 0.94-1.00) and 0.98 (95% CI 0.96-1.00)) and conventional MRI (AUC 1.00 (95% CI 1.00-1.00)), (n=96-101).

INTRODUCTION:

Various multi-contrast MRI techniques have been proposed recently for brain MRI1–3, aiming for reduced exam time by acquiring some or all of the image contrasts necessary for the particular patient. Recently, we proposed an improved version4 of our EPI-based multi-contrast sequence, "EPIMix", which can produce T1-FLAIR, T2-FLAIR, T2*w, T2w, DWI and ADC images in about 70-80 s with full brain coverage (34-36 slices). In this study, we have added EPIMix to our clinical brain MRI protocols to be able to compare the diagnostic performance by reading only the EPIMix data vs reading only the conventional MRI data in the same exam.

METHODS:

101 consecutive patients eligible for brain MRI were included after informed consent (January 2018-May 2018). The patients were enrolled without sub-grouping into different disease types. All included patients were first imaged with the relevant clinical protocol given their disease type, followed by an EPIMix scan. To assess the diagnostic confidence in determining an exam as normal or abnormal, two neuroradiologists were reading each exam twice. First, only using the six MR contrast from the EPIMix data, and then 2-3 weeks later, the conventional MR images were read. For each read, a form was filled in with presence of symptom-causing lesions, disease type (neoplastic, postoperative, ischemic, hemorrhagic, neuroinflammatory, vascular, or other), artifact levels etc. No patient history or clinical information were available to the neuroradiologists. The conventional MRI was used as gold standard. The diagnostic performance to categorize a clinical brain MRI scan with EPIMix or conventional MRI as abnormal (symptom-causing lesion) or normal was assessed via receiver operating characteristics (ROC) and the area under curve (AUC). The differences between ROC curves were estimated with DeLong’s test and comparison of confidence intervals (CI). All exams were performed on a 3T Discovery 750w and a Optima 450 1.5T MR system (GE Healthcare, WI, USA).

RESULTS:

The diagnostic performance to categorize a full brain MRI investigation as abnormal or normal was comparable between EPIMix (AUC 0.97 (95%CI 0.94-1.00) and 0.98 (95% CI 0.96-1.00)) and conventional MRI , (n=96-101). The sensitivity was 93% (95% CI 85-99, reader 1) and 90% (95% CI 82-96, reader 2), and specificity 100% (95% CI 76-100, reader 1) and 100% (95% CI 82-100, reader 2). Fig. 1 summarizes the diagnostic performance of EPIMix in a pivot table. Disease categorization into pre-specified subgroups were congruent between EPIMix and conventional MRI in 91% (92 of 101) of the cases for reader 1 and in 88% (84 of 96) of the cases for reader 2. The disease categories based on the reference standard conventional MRI for both readers are detailed in Fig. 2. Among the false negative or false disease categorization cases, the pituitary region pathology was the most common. In Fig. 3-4, two patients are presented. Fig. 3 illustrates high correspondence between EPIMix and conventional MRI for all contrasts in visualizing common features of a glioblastoma. Fig. 4 illustrates EPIMix's ability to detect small lesions. In Fig. 5, the prevalences of artifacts reported while reading the EPIMix data are summarized in three categories: not present, present but not degrading diagnostic confidence, or present and degrading diagnostic confidence.


DISCUSSION:

In general, there was a good diagnostic capability of the new technique to classify a scan as normal or abnormal. Moreover, in most cases the same diagnosis could be made with EPIMix vs conventional MRI data. Patients with pituitary region pathology could not be classified with sufficient clinical confidence, which is expected as this is located at the tissue-air interface of the brain subject to large field inhomogeneities resulting in strong geometric distortions in the EPIMix data.

CONCLUSION:

EPIMix has comparable diagnostic performance as conventional MRI in categorizing a clinical brain MRI as abnormal or normal.

Acknowledgements

Fredrik Granath

Yords Österman

Mitra Pedersen

Lars Blomberg

Maria Sandell

Yvonne Millstam

Amir Hossein Montazeri Najafabadi

Birgitta Löwegren

Yennie Gysell

References

1. Jeong J, Nam Y, Lee J. Penta-contrast imaging: a Novel Pulse Sequence for Simultaneous Acquisition of Proton Density, T1, T2, T2* and FLAIR images. In: Proceedings of the 25th Annual Meeting of the ISMRM, Honolulu, Hawaii. ; p. 3881.

2. Breutigam N-J, Frost R, Eickel K, Porter D. Simultaneous Multi-Contrast Imaging with Readout-Segmented EPI. In: Proceedings of the 25th Annual Meeting of the ISMRM, Honolulu, Hawaii. ; p. 520.

3. Skare S, Sprenger T, Norbeck O, et al. A 1-minute full brain MR exam using a multicontrast EPI sequence. Magn. Reson. Med. 2017 doi: 10.1002/mrm.26974.

4. Sprenger T, Engström M, Norbeck O, Rydén H, Avventi E, Berglund J, Skare S. Multi-contrast EPI - Towards clinical application. In: Proceedings of the Joint Annual Meeting ISMRM-ESMRMB 2018, Paris, France.


Figures

Figure 1. Pivot table (5x5) of diagnostic reading for reader 1. Scans of conventional MRI read by a specialist in neuroradiology and classified as definitely normal, probably normal, indeterminate, probably abnormal or definitely abnormal (5-grade Likert scale). After a memory wash-out period EPIMix was read by the same reader and classified again.

Figure 2. Diagnostic reading classifying the scans of the reference standard into pre-specified subgroups of diseases by reader 1 (figure 2 a) and reader 2 (figure 2 b). Categories were normal, neoplastic, postoperative, ischemic, hemorrhagic, neuroinflammation, vascular, other.

Figure 3. Splenic glioblastoma demonstrating high correspondence of EPIMix (left a-g) and conventional MR (right h-n): (b, i) T1 FLAIR before and (a,h) after contrast shows enhancing necrotic splenic tumor (arrow) with peritrigonal satellite on left side (dashed arrow). (c, j) T2-weighted and (d, k) T2 FLAIR high signal corresponding to non-enhancing tumor (arrow) and surrounding vasogenic oedema (dashed arrow) in right occipital lobe. (e,l) DWI high signal and (f,m) ADC low signal demonstrate slightly restricted diffusion suggesting regions of higher tumor cellularity (arrows). (g,n) T2*w low signal related to blood products (arrows) in necrotic component.

Figure 4. Multiple acute embolic infarcts in posterior circulation demonstrating that even small lesions are detected on EPIMix. Axial EPImix (left column a-f) and conventional MR (right column g-l) images. One pontine and two cerebellar infarctions (arrows) with low signal on (a,g) T1 FLAIR, and high signal both on (b, h) T2w and (c,i) T2-FLAIR. (d,e) DWI high signal with corresponding (j,k) ADC low signal confirms restricted diffusion consistent with acute infarctions. No signs of hemorrhage on (f,l) T2*w images.

Figure 5. Artifacts in EPIMix at review scored by reader 1 as "Not present", "Present but not degrading diagnostic confidence” or "Present but degrading diagnostic confidence". The numbers in the figure indicate counts.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
3328