In this study, we evaluated a new, clinically feasible, enhanced CS algorithm combined with a TSE T2w sequence using the PROPELLER-technique (T2PROPELLER CS) against a conventional, CS-accelerated Cartesian TSE T2w sequence (T2Cartesian CS) for MR imaging of the brain in infants and children. In comparison to the conventional T2Cartesian CS sequence, the T2PROPELLER CS sequence enabled faster and motion-robust imaging significantly accelerating acquisition time by 31%, reducing (motion-)artifacts, increasing image sharpness, basal ganglia delineation, lesion conspicuity, and overall image quality. aCNR and aSNR were reduced without significantly affecting the diagnostic value.

In pediatric MRI, reducing the rate of non-diagnostic scans due to artifacts and shortening acquisition time are crucial not only for economic reasons but also to minimize sedation or general anesthesia. Despite strategies such as exam preparation with MRI simulation and in-scan entertainment, the MRI environment is often intimidating and requires sedation or general anesthesia for children under the age of 6 or those with serious illnesses and difficulties to remain still during lengthy exams [33]. While sedation is generally well tolerated by children, the potential risks of increased doses of anesthetic medications during prolonged MRI examinations and the economic as well as operational expenses associated with the use of anesthesia underscore the need to minimize the use of anesthesia and decrease scan durations [33].

Over the last decade, sparse reconstruction techniques, including the CS framework, have been extensively investigated for their potential to decrease scan durations in neuroimaging and pediatric MRI [8,9,10]. The need for speed concerns not only the scan duration itself, but also the robustness and the efficiency of the data acquisition [11]. Besides minimizing the effects of involuntary motion and physiological noise by reducing the acquisition time, motion artifacts due to in-plane rotation and translation can be further avoided by combining CS with radial sampling trajectories such as the PROPELLER-technique that oversample the center of k-space [34, 35]. Yet only recently a CS solution for the PROPELLER approach has become clinically available which allows a significant reduction in acquisition time [24, 25].

During the re-gridding process of the PROPELLER-technique, any k-space data that was compromised by motion is dispersed through the Cartesian-like matrix, reducing the impact of motion [3]. Even though the T2PROPELLER CS sequence in this study was acquired in non-sedated patients after the T2Cartesian CS sequence toward the end of the MRI exam when children tended to be more agitated, it provided images without impairment by motion artifacts that were of good to excellent quality and rated superior to the conventional images by three experienced radiologists in all evaluated categories [36]. Particularly, no significant differences in image quality were observed between sedated and non-sedated patients for the T2PROPELLER CS sequence underscoring its effectiveness in reducing artifacts. In contrast, the conventional T2Cartesian CS sequence was highly susceptible to motion artifacts that were significantly more pronounced in patients without sedation, resulting in images of non-diagnostic or poor quality that required repeat acquisition. Beyond motion artifacts, physiological noise was also significantly reduced in the T2PROPELLER CS sequence irrespective of sedation. The diagnostic superiority of the new sequence is further supported by the significantly improved lesion conspicuity, basal ganglia delineation, image sharpness, and overall image quality, regardless of whether the children were sedated or not. Analogous to our study, Vertinsky et al. were able to demonstrate the superior motion artifact reduction and diagnostic confidence of a T2w PROPELLER TSE against a conventional T2w TSE sequence and concluded that particularly young infants not undergoing anesthesia will benefit from the PROPELLER-technique [22]. Forbes et al. compared a PROPELLER T2w TSE sequence with enhanced reconstruction to a T2w single-shot TSE sequence in unsedated pediatric patients and found them to provide equal motion correction, with the PROPELLER approach enabling better assessment of the brain parenchyma [37]. Further accelerating the PROPELLER T2w TSE sequence using the new, clinically feasible, enhanced CS algorithm with improved motion correction and contrast weighting featured in this study holds the potential to further minimize or potentially even circumvent sedation time [6]. Moreover, shorter examinations allow for increased patient throughput, offering further financial incentives. Lastly, Andre et al. examined the prevalence, severity, and cost associated with motion artifacts and found that they are a common cause of MR image degradation and impose significant costs on radiology departments [38].

In this study, aCNR and aSNR of the T2PROPELLER CS sequence were reduced in comparison to the T2Cartesian CS sequence. While the PROPELLER-technique oversamples the k-space center, which typically improves the CNR and partially enhances the SNR, SNR still poses a significant challenge particularly when combined with acceleration techniques of any kind [3]. The increased bandwidth of the T2PROPELLER CS sequence in comparison to the T2Cartesian CS sequence may additionally contribute to the reduced aCNR and aSNR of the T2PROPELLER CS sequence in this study. Still, it is to note that the decreased aCNR and aSNR of the T2PROPELLER CS sequence did not affect the overall image quality as all radiologists preferred the T2PROPELLER CS over the T2Cartesian CS sequence. Metal artifacts were prominent in both sequences, though locally slightly more pronounced in the T2PROPELLER CS sequence. In the T2Cartesian CS sequence, we observed an additional Zipper artifact in phase-encoding direction at the level of the shunt valve that was not present in the T2PROPELLER CS sequence. Due to the rotating frequency encoding direction, the radial acquisition scheme of the T2PROPELLER CS sequence results in more pronounced local susceptibility artifacts. This is counteracted with a higher bandwidth, which in turn results in a slightly reduced aSNR. Zipper artifacts along the phase-encoding direction in the T2Cartesian CS sequence caused by imperfections of the CS reconstruction due to strong B0 inhomogeneities are avoided by the rotating phase encoding in the T2PROPELLER CS sequence.

Our study has several limitations. While we enrolled pediatric patients referred for brain MRI for a wide variety of indications, the relatively short study period of a few months and the inclusion of just more than 30 study patients may limit the generalizability of our findings. Future studies with larger patient populations will be valuable to further assess the performance of the T2PROPELLER CS sequence in routine clinical practice.

Furthermore, although we demonstrated good intra- and interrater reliability, involving more reviewers with varying levels of experience could further enhance the consistency and objectivity of image interpretation.

We utilized a high-end, clinical 3-T MRI system with advanced reconstruction hardware, resulting in short reconstruction times of around 30 s for the CS images. Yet, it is important to consider that image quality and acquisition as well as reconstruction times may vary depending on the specific MRI system employed [11, 34]. In particular, applicability of the T2PROPELLER CS sequence on a 1.5-T MRI system remains to be evaluated.

Lastly, this study is an initial evaluation of the T2PROPELLER CS sequence. Future studies will focus on further optimization by pushing the undersampling limits to further accelerate imaging and enhance diagnostic efficiency.