Deeply 3D-T1-TFE hypointense voxels are characteristic of phase-rim lesions in multiple sclerosis

In this study, we found that PRLs tended to contain deeply hypointense voxels on 3DT1TFE and that these markedly hypointense voxels were practically absent in nPR-WMLs. We also demonstrated that these differences can be explored on a lesion-wise basis by analyzing the lower percentile range of the normalized T1WI intensity of each lesion. Lesion classification based on the lesion-wise p5 intensity was very specific for PRL detection (specificity: 0.959).

The advent of disease-modifying therapies for progressive MS underscores the need for biomarkers for the early and accurate identification of treatment candidates [2, 3]. PRLs have previously been identified as putative biomarkers. For example, Absinta and colleagues demonstrated that PRLs were associated with less lesion-volume shrinkage than nPR-WMLs and became progressively more T1-hypointense over time. Autopsy reports then showed that PRLs corresponded to iron-laden inflammatory myeloid cells at the edges of lesions [8]. In a subsequent study, they prospectively analyzed 2019 patients with 3 T or 7 T MRI, 56% of which had PRLs. They found that patients with 4 or more PRLs reached disability status sooner than those with 3 or fewer; that PRLs tended to expand over time, unlike nPR-WMLs which tended to shrink; and that PRLs were more hypointense on T1WI, suggesting more profound tissue damage [9]. More recently, Treaba and colleagues showed that PRL count and volume on 7 T MRI were predictors of neurological disability progression in MS patients along with normalized subarachnoid cerebrospinal fluid volume, leukocortical lesion volume, and normalized white-matter volume, among other features [7]. These results are consistent with the hypothesis that PRLs and SELs represent smoldering or chronic inflammation, which would be more prevalent in progressive MS than in RRMS [6, 7, 10, 11]. Despite the promising results of this previous work, however, there are several reasons why these putative biomarkers are unlikely to be introduced into clinical workflows in the near future. First, evaluation of these biomarkers requires SWI sequences, and international guidelines do not include this imaging modality in standard MS MRI follow-up protocols [14]. Second, these biomarkers have been described chiefly in higher-magnetic field 7 T magnets and to a lesser degree on 3 T and 1.5 T scanners which are more widely available [10, 16‐18]. Finally, and most importantly, PRL identification is not straightforward and requires meticulous image examination. Thus, it is challenging to routinely quantify PRLs in clinical practice.

Nevertheless, our results herein demonstrate that 3DT1TFE MRI may be a novel screening method for PRLs. The identification of a profoundly hypointense lesion on this sequence may signal the possibility of a PRL and direct the radiologist to search for rim lesions on SWI. Moreover, in specific clinical scenarios, in the absence of SWI availability, profoundly hypointense lesions on 3DT1TFE could be considered a surrogate of PRLs. T1WI, either with or without contrast, is much more readily available than SWI and is routinely included in MRI brain scans irrespective of magnetic field type. Moreover, deeply T1WI hypointense foci are visible to the radiologist’s eye and are much more straightforward to objectively quantify by intensity histogram analysis than the presence of PRLs on SWIp.

The T1WI hypointensity of PRLs has been previously described in the literature [5, 8, 12, 13]. For example, a recent paper by Kee Kwong and colleagues demonstrated that 100% of PRLs (n = 25) identified in 45 MS patients corresponded to T1-hypointense lesions [13]. However, our study is the first to thoroughly and quantitatively analyze the normalized 3DT1TFE hypointensity of these lesions and propose its feasibility as a surrogate marker of smoldering inflammation. Furthermore, these findings suggest that PRLs are the sites of profound ongoing tissue damage.

The present study has several limitations. First is the use of non-quantitative imaging. Specifically, although our study is centered on intensity quantification, non-quantitative T1 imaging was performed. Nevertheless, these image sequences are the most widely available because they are recommended in clinical MS imaging guidelines [14, 15]. Therefore, to compensate for the shortcomings of this approach, we implemented an intensity normalization method that is commonly used in the literature [26]. Nevertheless, suboptimal normalization can introduce random noise to the data, thereby hindering analyses. The second limitation is that PRLs have often been described in the literature in the context of very high-field 7 T MRI scanners. Thus, our analysis of 3 T MRI scans may limit the number of PRLs that we can detect. However, the optimal MRI scanners available at most clinical centers worldwide are 3 T devices [14]. Third, our study is limited by its small patient sample size (20 subjects). However, this sample is homogeneous and of high quality; all images were performed on a 3 T MRI scanner following an identical protocol, in line with international guidelines. Furthermore, we performed voxel-wise and lesion-wise analyses rather than patient-wise analyses. Therefore, the smallest group analyzed contained 39 PRLs. In summary, we consider this a proof-of-concept study. Future research should consist of a multicentric analysis including a more varied array of image sequence parameters. Finally, we want to clarify that we refrain from making assumptions about patient-specific results in this paper, as the primary objective is not to investigate individual patient outcomes, for which a larger patient sample would be necessary.

In conclusion, this study demonstrated that deep T1WI hypointensity lesions on 3DT1TFE MRI are relatively specific for PRLs. These preliminary results suggest that these lesions could serve as surrogate markers for PRLs, paving the way for improved study of smoldering lesions and progressive MS activity. Moreover, the availability of T1WI, together with the straightforward evaluation and quantification of lesion intensity, renders it more amenable to clinical implementation than SWI-based PRL detection. Future studies might aim at confirming the reproducibility of these findings in larger subject samples and with different MRI acquisition parameter configurations, as well as exploring other methods of quantification and standardization of deep T1WI hypointensity and evaluating its prognostic value.