At our academic tertiary care referral center and designated stroke center, all adult brain MRI studies include both a standard DWI series (three-orthogonal diffusion gradient) and a 15-direction DTI with parallel acquisition technique, which appears at the radiologist's workstation as an isotropic DWI series (15-direction isotropic DWI). All brain MRI studies from a single MRI scanner over a 1-year period (November 1, 2010, to October 31, 2011) were included in this study (2656 imaging studies).

This study was approved by our institutional committee on human research. All brain MRI studies included a DTI sequence, with the exception of pediatric patients under anesthesia, patients being scanned for evaluation of deep brain stimulator placement, and patients unable to cooperate for a full examination. Of a total of 2656 patients, 2537 studies included DTI.

All 2537 studies performed with DTI during the 1-year interval at a single university medical center MRI scanner were reviewed for assessment of potential for clinical impact of 15-direction iso-tropic DWI over routine DWI. Cases were initially reviewed by single observers to identify those that were positive for cerebral infarction by standard DWI or 15-direction isotropic DWI. Cases of restricted diffusion secondary to surgery, trauma, multiple sclerosis, or abscess were not included. The total pool of infarction-positive imaging studies were reviewed by individual observers and divided into two groups: infarcts readily identified on both types of diffusion studies and those in which there appeared to be significant differences between the two studies, meriting a more detailed review. This latter group of studies (n = 102) were reviewed on a PACS by the two neuroradiologists, both of whom hold a certificate of added qualification, to stratify cases into categories by consensus to reflect the impact on clinical management (Table 1).

All subjects were scanned on a single 1.5-T scanner (Signa, GE Healthcare) equipped with gradients with a maximum slew rate of 120 mT/m/ms and a maximum strength of 33 mT/m. A standard receive-only multielement (eight-element) surface head coil was used. Essentially the same single-shot echo-planar imaging technique was used for both routine (three-direction) DWI and the 15-direction acquisition, with a diffusion sensitivity of b = 1000 s/mm². The diffusion data obtained along each of three orthogonal axes (x, y, and z) were automatically processed to yield a standard isotropic DWI, in which 5-mm axial slices (26 slices) were acquired with a 1-mm interslice gap using an FOV of 26 cm. Similarly, the diffusion data from 15 different directions were automatically processed to yield an isotropic DWI, expressed as DWI_iso = (DW₁ through DW₁₅)^1/15. The 15-direction acquisition was performed at a 5-mm thickness without gap for 29 slices. Each study type used a 128 × 128 acquisition matrix. The parallel imaging sequences were performed using the generalized autocalibrating partially parallel acquisition (GRAPPA) technique, with a reduction factor of 2. For DWI, the TR/TE was 8000/80, and for the 15-direction study TR/TE was 8000/96.7. Our routine DWI protocol entails two signals acquired. The 15-direction acquisition with one signal acquired was performed in addition to this routine protocol. Total acquisition time for standard DWI was 1 minute 40 seconds and for the 15-direction DTI study was 2 minutes 24 seconds.

The diffusion-weighted images of 22 consecutive negative (normal) studies with indications of low probability for acute findings were reviewed on the PACS. Two neuroradiologists with the certificate of added qualification in neuroradiology from the American Board of Radiology, each with more than 4 years of experience in neuroradiology image interpretation, evaluated image quality. The radiologists were blinded to imaging technique and patient identity. Readers were asked to rate, in consensus, the image quality with a 5-point scale, similar to one previously used in the evaluation of sensitivity encoding [9]. Reviewers were able to change windowing as necessary to assess image quality. Images were evaluated for presence of artifacts (susceptibility-induced image distortions, delineation of anatomic detail, and image sharpness). A score of 1 was assigned to images that were of nondiagnostic quality. A score of 2 (poor) was assigned to cases with distortions in some but not all parts of the images, such that image quality could interfere with accurate diagnosis. A score of 3 (satisfactory) was assigned when distortions were clearly present but not thought to interfere with diagnosis, and all anatomic regions were well shown. A score of 4 (good) was assigned in cases with minor artifacts. A score of 5 (excellent) was assigned in cases in which there were no artifacts.

Quantitative image analysis was performed in 46 consecutive studies and 51 lesions. These were the first 46 consecutive patients of the study group (November 1, 2010, through December 31, 2010). This subset of lesions was taken to be a representative but manageable sampling of the total lesion pool. Image evaluation was based on relative tissue signal intensity (SI) measurements. Regions of interest (ROIs) were placed at the radiology reading monitor on PACS-archived images by one author. ROIs were placed on axial diffusion-weighted images. ROI size was variable and dependent on the lesion size. Standard DWI and 15-direction isotropic DWI in each case were evaluated concurrently. Every effort was made to evaluate paired DWI and 15-direction isotropic DWI at identical levels. Identically sized ROIs were placed in identical positions on image pairs. The adjacent ROI was the same size as the infarct ROI; if the infarct ROI occurred in white matter, the adjacent ROI was placed in white matter, and if the infarct occurred in gray matter, the adjacent ROI was placed in gray matter.

Calculation of lesion contrast—We calculated the change in SI of DWI hyperintense lesions (infarcts) according to the following equation: ΔSI = SI_lesion − SI_{adj tiss}, where SI_lesion is the SI of the lesion visible on the diffusion-weighted image and SI_{adj tiss} is the SI of the normal brain tissue adjacent to the lesion, as has been described [9]. To normalize this value, the relative SI (r-SI) is expressed as a percentage of the baseline SI for the image (r-SI = ΔSI / SI_{adj tiss}).

Calculation of percentage difference in lesion contrast with 15-direction isotropic DWI versus routine DWI—The r-SI of the lesion on 15-direction iso-tropic DWI was then subtracted from the r-SI of the lesion on routine DWI, and the difference was expressed as a percentage of the r-SI on routine DWI, (r-SI_{15-direction-isoDWI} − r-SI_DWI) / r-SI_DWI).

Estimation of lesion size—Lesion sizes were estimated on the single planar image on which the lesion was largest and were intended as only rough estimations because infarcts can be multifocal and infarct margins can be indistinct and irregular in shape.

Statistical analyses including scatterplots with linear regression as well as Pearson correlation (Fig. 1) were performed using SPSS statistical software, version 20. For image quality assessment (Fig. 2) a one-way analysis of variance test was performed.

Image quality scores are shown in Figure 2. The images were reviewed on a PACS. The presence of various artifacts contributed to lower image quality scores for standard DWI over 15-direction isotropic DWI. Because routine DWI is the clinical standard, these images were considered to be diagnostic in 18 of 20 patients, with most images falling into the satisfactory (seven of 20) and good (11 of 20) categories. In every case, iso-tropic DTI with parallel imaging was judged to produce higher image quality. Using single-factor analysis of variance, image quality scores were higher for 15-direction isotropic DWI (F = 4.07, p < 0.001). Average image quality scores for standard DWI were 3.45 (variation, 0.45) and 4.68 (variation, 0.42) for 15-direction isotropic DWI.

The percentage difference in contrast of 15-direction isotropic DWI relative to conventional DWI was also evaluated as a function of lesion size (Fig. 1A). Fifteen-direction isotropic DWI shows improved detection of small lesions (less than 5 mm). A regression line curve-fitting program identified the best fit as logarithmic (R² = 0.363, F statistic, 26.7), consistent with improved lesion detection of the 15-direction isotropic DWI study over routine DWI, with greater improvement for smaller lesions.

To objectively detail the improved ability to detect small infarcts with isotropic DW images, differences in relative SI of infarcts were explored. Evaluation of 51 lesions in 46 consecutive infarct-positive patients showed that relative SI of infarcts was inversely correlated with the absolute SI of the lesion (Fig. 1B). More simply phrased, lesions of lesser signal intensity were better seen with 15-direction isotropic DWI than with routine DWI, with a Pearson correlation of −0.634 and significance < 0.05.

Images were reviewed to learn the impact of improved image quality of 15-direction isotropic DWI on clinical decision making. Clinical impact was reflected in generally increased confidence of image interpretation, cases of multiple infarcts in which increased image quality showed additional lesions in a second vascular distribution that could affect clinical decision making regarding the source of the emboli, and lesions that were seen only on the 15-direction series (Table 1). Of 365 cases that showed infarction, 102 studies were deemed to benefit by the use of 15-direction isotropic DWI over conventional DWI. Twenty-nine cases of single or multiple infarcts were more conspicuous on 15-direction isotropic DWI. In 49 cases, additional small infarcts could be seen in the same vascular territory as seen on routine DWI. In 12 cases of multiple infarcts, additional lesions could be seen in a different vascular territory than was identified on routine DWI. In 12 cases an infarct that could not be seen on routine DWI could only be identified retrospectively on routine DWI after being identified on 15-direction isotropic DWI. These numbers suggest a clinical impact in terms of increased confidence of interpretation in up to 28% of infarct cases and with a more quantifiable and direct clinical impact (categories 2c, 3cii, and 3d as detailed in Table 1) in 6.6% of positive cases (24 of the total 365 infarct-positive cases).

Lesions in the midbrain and brainstem were better seen with 15-direction isotropic DWI. Figure 3 illustrates improved detection of a lesion in the left inferior olivary nucleus that was not prospectively seen on conventional DWI and also not seen on the routine apparent diffusion coefficient (ADC) map. The ADC map corresponding to the 15-direction isotropic DWI clearly shows the lesion. Figure 4 shows improved detection of supratentorial lesions with scattered small embolic infarcts, many of which are much better seen on the 15-direction study. An isolated single focus of restricted diffusion in the medial right frontal lobe could not be seen prospectively with routine DWI but is much more apparent with 15-direction isotropic DWI (Fig. 5).

Improved diffusion-image quality is of clinical significance because there is increasing awareness of the importance of identifying small infarcts and ischemic lesions in patients with transient symptoms. Transient ischemic attack (TIA) can correlate with focal ischemic abnormalities in up to 68% of reported cases, typically with a diameter less than 15 mm and volume of 1–3 mL [11]. Accurate diagnosis of transient symptoms with an ischemic lesion is important because the short-term risk of stroke is high [11, 12]. Clinical findings of TIA can be nonspecific or can be explained by another diagnosis [11, 13]. One study found that approximately 50% of lesions of TIA are cortical [14], and it has been claimed that failure to recognize small cortical infarcts is “partly due to inability of conventional techniques to identify acute small lesions” [13]. It has also been noted that conventional DWI is relatively insensitive to small lesions in the brainstem and medullary region [13, 15–17], an area heavy affected by geometric distortion with the conventional spin-echo single-shot echo-planar DWI technique [18]. Higher-quality DWI studies have the potential to increase detection of small infarcts, which can be of high clinical significance.

A review of 365 consecutive infarct-positive studies showed that the 15-direction iso-tropic DWI parallel acquisition technique offered increased confidence of interpretation in virtually all cases involving tiny infarcts or infarcts of decreased SI. Beyond increased confidence of interpretation, which was deemed significant in up to 28% of total infarcts, lesions were identified in a second vascular distribution in 12 cases (3.3% of total infarct cases) and 13.6% of cases with multiple small (< 5 mm) infarcts. In 12 cases, infarcts were only seen initially on the 15-direction study (3.3% of total infarct-positive cases). Improved lesion conspicuity was seen to inversely correlate with lesion SI; lesions with less brightness were better seen with DTI (15-direction isotropic DWI). Similarly, improved lesion conspicuity was seen to inversely correlate with lesion size; smaller lesions were better seen with 15-direction isotropic DWI. These findings serve to objectify the general impression of interpreting neuroradiologists at our institution that 15-direction isotropic DWI is better for visualizing infarcts at the limits of DWI capability—infarcts that are small and those that are less hyperintense on routine images. Because it has been shown that TIA may correlate with tiny lesions, some less than 5 mm [19], and that DWI or ADC signal changes in TIA were less pronounced and smaller in volume than lesions in stroke [20], these improvements in image quality offer greater sensitivity for detection of small infarcts, such as those that might be expected to correlate with TIA.

Improved detection of small infarcts with 15-direction isotropic DWI is due to four factors: one, the use of parallel acquisition technique in DTI acquisition; two, increased numbers of diffusion gradient directions; three, increased signal-to-noise ratio (SNR) from an overall increased scanning time relative to the routine three-direction study (2 minutes 24 seconds vs 1 minute 40 seconds); and four, absence of an interslice gap. Parallel acquisition technique has received considerable attention as a tool to aid in the reduction of geometric distortion at higher field strengths; few reports discuss the use of parallel imaging together with routine DWI at 1.5 T. A recent study in a small clinical cohort reported improved quality of standard DWI at 1.5 T with parallel acquisition technique showing reduced geometric distortion and improved diagnostic image quality with significance for infarcts in the posterior fossa [21].

Standard DWI typically includes an inter-slice gap, and exclusion of the gap may require special software and result in significantly increased scanning time. A 1-mm gap with a 5-mm slice width has the potential to contribute to decreased SI for a significant percentage of small lesions. Although we have not quantified this effect, the absence of a gap in the DTI series is likely to improve the sensitivity for small infarctions.

Factors two and three—increased numbers of gradient directions and increased SNR due to increased scanning time—are related and not easily separated experimentally. A study could be designed to investigate the role of increased SNR relative to increased numbers of gradient directions. For example, the number of signals acquired for the three-direction study could be increased to five so that scanning times and SNR become comparable between the DWI and DTI series. DWI with parallel acquisition technique and without a slice gap could then be compared with the DTI trace series to rigorously address the contribution of increased gradient directions alone; however, such a DWI series would not represent a clinical standard.

Our study has some limitations. The TEs were defined automatically by the pulse sequence and were different for scans with and without parallel imaging as has been noted by others [22]. Shorter TE values of the magnitude reported here are expected to improve SNR but not strongly influence the measured diffusion properties [22]. Although the two diffusion sequences are acquired within minutes of each other during the same scanning session, patients may move slightly between scanning. Because our scanning thickness is 5 mm in both sequences and our interest is in very small lesions, one might contend that differences in slice selection contribute to differences in the image appearance and in the conspicuity of some lesions. Such differences in slice selection occur because the routine DWI includes a 1-mm interslice gap, whereas the DTI series is without an interslice gap. Direct comparison of the two series, however, consistently favors 15-direction isotropic DWI over routine DWI, indicating that simple random differences in slice selection do not explain the observed increased quality of the 15-direction isotropic DWI series.

Small lesions can be difficult to confirm using the traditional methods of comparison with ADC maps and T2-weighted images to identify T2 shine-through. The standard ADC map generated from three gradient directions will suffer from the same lack of contrast and image distortion as the diffusion-weighted images and thus cannot be used to validate a lesion seen on isotropic DWI and not seen on routine DWI. Many lesions can be confirmed by comparison with the 15-direction ADC map, which is generated with the isotropic diffusion-weighted images. The possibility of T2 shine-through mimicking real restricted diffusion is expected to occur with isotropic DWI just as with conventional DWI, necessitating review of T2-weighted images and ADC maps, and we have no evidence to suggest that this issue changes with the diffusion technique. As with most clinical radiology, our best evidence that these lesions are real relies on the clinical setting and the experience of the interpreting neuroradiologist. To assess the value of the iso-tropic DWI series, we attempted to minimize other variables by analyzing the output of a single MRI scanner at our institution. The relative improvements of 15-direction isotropic DWI relative to routine clinical DWI will likely vary among vendors and the details of the routine DWI series protocol.