Introduction

Leukodystrophies and hereditary leukoencephalopathies constitute a clinically and pathologically heterogeneous group of genetic diseases affecting the white matter of the central nervous system (CNS) [1]. While correctly diagnosing them remains challenging, modern neuroimaging, genetic, and histopathologic techniques facilitate the differential diagnosis and allow for a more precise classification [1, 2].

Mutations in the colony-stimulating factor 1 receptor (CSF1R) gene can cause adult-onset primary CNS microgliopathy, affecting the axon-glia integrity and leading to progressive and eventually fatal leukoencephalopathy [3]. CSF1R gene mutations are known to cause both familial pigmentary orthochromatic leukodystrophy (POLD), first reported in a Belgian family in 1936 [4], and many cases of hereditary diffuse leukoencephalopathy with spheroids (HDLS), which was first described in a Swedish family in 1984 [5]. In 2011, CSF1R gene mutations were identified [3], and surprisingly some POLD and HDLS families demonstrated a shared genetic defect [6]. However, the initial Swedish HDLS family was recently found to have a different mutation in the alanyl-tRNA synthetase (AARS) gene. Therefore, the current nomenclature of these classes of disorders has changed to CSF1R-related leukoencephalopathy and AARS-related leukoencephalopathy. In addition, there are known cases that are phenotypically and pathologically indistinguishable but lack mutations in these genes. They are collectively classified as CSF1R/AARS-negative adult-onset leukoencephalopathy with axonal spheroids and pigmented glia (ALSP) [7, 8], (Z. K. Wszolek, personal communication, August 24, 2021).

Due to the varying terminology over time and advances in genetic analyses, there is value in integrating the literature on neuroimaging manifestations of the now genetically defined group of CSF1R-related leukoencephalopathy [9]. To date, 96 CSF1R mutations have been described in around 200 families [2]. The mutations lead to translation and transcription of a dysfunctional tyrosine kinase receptor protein, which normally potentiates proliferation, differentiation, and survival of microglia [10]. Although CSF1R mutations are inherited in an autosomal dominant pattern, many cases are sporadic, due to novel mutations or mosaicism, which may complicate and delay the diagnosis [2].

The CSF1R mutations result in cognitive impairment, behavioral disturbances, and motor dysfunction with a typical onset in the fourth or fifth decade of life. At the onset, the clinical presentation may mimic a wide range of neurological disorders, such as multiple sclerosis, frontotemporal dementia, Parkinson's disease, or atypical parkinsonism. Sometimes, the initial symptom is a single seizure or severe migraine [11], and other times it presents insidiously with a slow decline in health [12]. Due to its heterogeneity, rarity, and mimicry of more common neurological disorders, CSF1R-related leukoencephalopathy is frequently misdiagnosed in its initial stage [13]. The disease is, however, fatal. Correct and timely diagnosis is therefore paramount for correct prognostication and care [14]. More recently, allogeneic stem cell transplantation and microglia replacement therapy have shown promising therapeutic results [15-17].

Given the complexities of clinically differentiating CSF1R-related leukoencephalopathy from other disorders, neuroimaging plays an important role in discovering and suggesting the diagnosis, which may lead to diagnostic genetic testing [2]. Previous literature on the subject indicates that the illness may present with radiological findings that are rather consistent, thereby potentially aiding in diagnosis [2]. However, given the limited number of described families/cases, evolving nomenclature, and substantial overlap among all three major disorders—CSF1R-related leukoencephalopathy, AARS-related leukoencephalopathy, and CSF1R/AARS-negative ALSP—the complete picture of the current body of knowledge in terms of imaging is lacking.

We, therefore, aimed to systematically summarize the available evidence on CSF1R-related leukoencephalopathy neuroimaging phenotypes and to identify gaps of knowledge. Based on the findings, we present pictorial examples that can aid clinicians in identifying cases and we furthermore provide neuroimaging recommendations. The findings may provide a foundation for future diagnostic imaging criteria and highlight possible imaging biomarkers for clinical trials of emerging therapies.

Materials and methods

The study was registered in the international prospective register of systematic reviews and performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines [18]. PubMed, Web of Science, and Ovid EMBASE were searched on 25 August 2021 for original studies and case reports/series with neuroimaging findings in CSF1R-related leukoencephalopathy. Methodological details are provided in Table S1. The search covered previous terminologies such as HDLS, POLD, and ALSP, but only cases with confirmed CSF1R gene mutations were included. The detailed search strategy is described in Table S2. All studies were quality assessed according to the Joanna Briggs Institute Critical Appraisal Checklist for Case Reports 2017, provided in Table S3 [19]. Results were qualitatively compared and summarized in a narrative fashion. A meta-analysis was performed for the age of onset, age at imaging, and the diagnostic delay between disease onset and neuroimaging.

Ethical considerations

In order to aid clinicians in recognizing neuroimaging phenotypes of CSF1R-related leukoencephalopathy, pictorial examples are provided in this article. The images come from previously reported cases but have not been published before. They have been obtained with informed consent and are published with approval from the Swedish Ethical Review Authority (reg. no. EPM 2019-05742).

Results Eligible publications

A total of 6912 publications were obtained in the systematic database searches. References from review articles provided nine additional publications. After deduplication and title and abstract screening, a total of 299 publications were eligible for full-text analysis (4%). Details of the reference selection are reported in Fig. 1. A total of 78 publications were included for the final analysis.

PRISMA flow chart. CSF1R, colony-stimulating factor 1 receptor; MRI, magnetic resonance imaging.

Risk of bias assessment

Overall, most studies were considered to have a moderate (36 publications, 45%) or high (32 publications, 42%) risk of bias, while the remaining had a low risk of bias (10 publications, 13%). However, all studies showed a low risk of bias for the Q4A and Q4B domains, that is, whether the diagnostic tests and their results were clearly described. The complete risk of bias assessment is provided in Table S4.

General study and case characteristics

Of 78 publications, all provided brain magnetic resonance imaging (MRI) data (195 cases, 88 women [54%], 74 men [46%]; the sex was not reported in 33 cases) and four additionally provided spinal cord MRI data (seven cases). Furthermore, publications reported on findings on brain computed tomography (CT) (19 publications, 32 cases), brain positron emission tomography (PET) with glucose and amyloid tracers (11 publications, 12 cases), and brain single-photon emission computed tomography (SPECT) with perfusion tracers and DaTscan (10 publications, 10 cases). Altogether, these publications comprised a total of 195 cases with a median onset age of 41 years (interquartile range [IQR] 17 years). The median age of onset was 43 years for men (IQR 14 years) and 40 years for women (IQR 15 years). Women had a statistically significant earlier age of onset (p = 0.041 by Wilcoxon rank test). Mean delay between symptom onset and neuroimaging was 2.3 years (standard deviation 2.8 years) and was not different between sexes. Age of onset was not reported in 11 cases (6%). The most commonly reported ethnicities of eligible cases were Caucasian (40 cases, including three German, two Italian, one Croatian, one French, one Greek, and one Polish, 48% of cases with reported ethnicity), followed by Chinese (19 cases, including two cases from Taiwan, 23%), Japanese (18 cases, 21%), Korean (six cases, 7%), and African American (one case). Ethnicity was not reported in 111 cases (57%). Demographic data of all cases are summarized in Table 1 and the most common neuroimaging findings by modality are summarized in Table 2.

Table 1. Demographic characteristics of patients by neuroimaging modality Sex (N) Modality Cases (N) Male Female N/A Age of onset Age at imaging Ethnicity reported (N) MRI, brain 195 74 88 33 41, 17 44, 21 81 MRI, spinal cord 7 1 6 0 32, 14 34, 15 5 CT 32 7 18 7 38, 11 34, 16 19 PET 14 7 6 1 41, 21 29, 21 10 SPECT 10 4 6 0 48, 6.5 56, 10 5 Note. Age is given in years as median, interquartile range. Abbreviations: CT, computed tomography; MRI, magnetic resonance imaging; N, number; N/A, not applicable; PET, positron emission tomography; SPECT, single-photon emission computed tomography. Table 2. The most prevalent neuroimaging findings by imaging modality Imaging modality Number of cases Number of publications Commonly reported imaging findings CT 32 19 White matter hypodensities mainly in frontal, parietal, and periventricular areas White matter calcifications (spotty, stepping-stone, serpiginous patterns) mainly in frontal and periventricular areas MRI 195 78 White matter lesions mainly in frontal, parietal, and periventricular areas Cortical atrophy (especially frontoparietal) and central atrophy (especially corpus callosum) Diffusion restriction in the periphery of white matter lesions that is long-lasting Spinal cord lesions and diffusion restriction in the spinal cord may occur Rarely, contrast-enhancement in white matter lesions PET 14 13 FDG-PET shows cortical hypometabolism in prefrontal, frontal, parietal, and orbitofrontal regions Amyloid PET data are limited but suggest that amyloid deposition may occur SPECT 10 10 Marked hypoperfusion in frontal and parietal lobes Heterogeneity exists both within and between patients DaTscan data are limited but have been normal or show reduced uptake in the putamina Abbreviations: CT, computed tomography; FDG, fluorodeoxyglucose; MRI, magnetic resonance imaging; PET, positron emission tomography; SPECT, single-photon emission computed tomography. MRI findings

White matter pathology. The most commonly observed MRI features were white matter hyperintensities on T2-weighted and fluid-attenuated inversion recovery imaging (157 cases, 81%), and white matter hypointensities on T1-weighted images. The most commonly reported white matter lesion pattern was either patchy or confluent, often with a symmetrical distribution. Less frequently, an ovoid shape of individual lesions was reported. In one case, a tumor-like lesion appearance was reported [20]. White matter lesions were mostly located in frontoparietal or periventricular areas (87 cases, 45%). Less frequently, they were also observed in the deep white matter (45 cases, 23%), occipital lobes (13 cases, 7%), corticospinal tracts (eight cases, 4%), cerebellum (seven cases, 4%), and juxtacortically (six cases, 3%). Of 195 cases, no gadolinium-enhancement was specifically reported in 13 cases, while one case showed gadolinium-enhancement [21]. In addition, corpus callosum pathology has been reported by several publications with T2 hyperintense lesions in the corpus callosum described in 35 cases (18%).

Brain atrophy. Atrophy has been reported in frontal/frontoparietal lobes (55 cases, 28%), parietooccipital lobes (four cases, 2%), and the cerebellum (three cases, 2%). Cortical atrophy was reported in 25 cases (13%) and central atrophy in 13 cases (7%). Callosal atrophy was reported in 56 cases (29%).

Diffusion-weighted imaging. Diffusion restriction was reported in 52 cases (27%), in a similar distribution as the white matter lesions (bifrontal/parietally, deep white matter, periventricular, and in the corpus callosum). During follow-up imaging, foci of diffusion restriction were persistent for several months to years.

Vascular pathology was not identified in three cases with magnetic resonance angiography.

Spinal cord pathology. Four cases with spinal cord MRI have been reported; in two cases, no abnormal findings were found [22, 23]. One case exhibited T2 hyperintensities along the spinal axis (including the corticospinal tract) [24], and one case had diffusion restriction along the corticospinal tracts [25].

Longitudinal brain MRI data. In 19 cases, follow-up brain MRI was performed after a median of 3 years after the baseline brain MRI, with a maximum follow-up of 11 years. In many cases, imaging findings such as white matter lesions and brain atrophy progressed during follow-up (nine cases, 47%).

Advanced MRI findings. Susceptibility-weighted imaging (SWI) has also been applied, demonstrating the absence of microbleeds [26, 27], as well as putaminal iron depositions [28]. Volumetric analysis has revealed both grey and white matter atrophy. The white matter lesion volume has been shown to be strongly correlated with mini-mental state examination scores [29]. More detailed volumetric analysis has shown thalamic and hippocampal atrophy [30]. Multiparametric quantitative MRI has been applied with machine learning approaches to differentiate CSF1R-leukoencephalopathy from multiple sclerosis. The myelin content in the cortex was a distinguishing feature with cortical demyelination being less prominent in CSF1R-leukoencephalopathy [31]. Magnetic resonance spectroscopy was applied in six publications (15 cases) [26, 32-36]. Several publications noted a decrease in N-acetyl aspartate (nine cases) as well as an increase in choline and inositol (seven cases) and lactate (three cases). In one publication, more advanced diffusion kurtosis imaging was performed, which revealed diffusion abnormalities in the corpus callosum, periventricularly, deep white matter, subcortical U-fibers, cortical midline structures, and deep grey matter structures [30]. Using resting-state functional MRI, the same study demonstrated reduced functional connectivity between the caudate nuclei and the contralateral hippocampi [30]. All brain and spinal cord MRI findings are presented in Tables S5 and S6, respectively.

CT findings

Calcifications. Brain parenchymal calcifications were reported in 24 out of 32 cases (75%), most commonly described as a “stepping-stone pattern” (six cases, 19%) or “spotty calcifications” (five cases, 16%). One case had asymmetric serpiginous calcifications [37]. These calcifications have been located in the frontal white matter (adjacent to the frontal horns in six cases, 19%), subcortical areas (three cases, 9%), periventricular white matter (three cases, 9%), pericallosal white matter (two cases, 6%), as well as the callosal truncus and splenium (one case). In contrast to calcifications in the frontal lobes, one case with a mutation in exon 19(p. Glu847Val.) showed a “confluent nature of the large calcifications extending posteriorly from the posterior horns of the lateral ventricles” [37]. In four cases, an absence of calcifications was reported (13%) [23, 38-40].

Hypodensities. White matter hypodensities have been described frontally and periventricularly (two cases, 6%).

Atrophy. Diffuse brain atrophy (two cases), enlargement of lateral ventricles (two cases), or generalized infratentorial atrophy (one case) have also been reported. All brain CT findings are presented in Table S7.

PET findings

Both brain fluorodeoxyglucose- (FDG) and amyloid-PET have been acquired in cases with CSF1R-related leukoencephalopathy.

FDG-PET. Glucose metabolism has been reported in 12 of 13 PET publications (92%) and 12 of 14 cases (86%). Hypometabolism has been observed in medial frontal areas, anterior pericallosal areas, bilateral parietal lobes, inferior temporal lobes, and occipital lobes. The most consistently reported regions with hypometabolism have been the frontal and parietal lobes, both reported in 8/14 cases (57%). Cortical hypometabolism was reported in 6/14 cases (43%). In one case, hypometabolism was disseminated in the entire cerebral cortex, except for the motor area [41]. Less frequently, mild hypometabolism has been observed in the basal ganglia, thalami, and putamen (2/14 cases, 14%).

Amyloid-PET. Two publications acquired amyloid-PET data (two cases). One publication using Pittsburgh Compound B reported no amyloid depositions in their case [42], and another publication using florbetapen noted positive amyloid-PET in temporal and occipital regions [43]. All brain PET findings are presented in Table S8.

SPECT findings

Of 10 publications acquiring brain SPECT (10 cases), abnormal metabolism was found in periventricular white matter and hypometabolism in frontal (2/5 cases) and parietal lobes (one case); generalized cortical hypoperfusion has also been described (one case) [44]. Imaging of the specific binding ratio of the dopamine transporter with 123Iodine-labelled N-(3-fluoropropyl)-2β-carbomethoxy-3β-(4-iodophenyl) nortropane (123I-FP-CIT) excluded tauopathy-associated neurodegenerative disease [45]. DaTscan has shown both normal findings (one case) [46], or reduced uptake in the putamina (one case) [40]. All brain SPECT findings are presented in Table S9.

Discussion

CSF1R-related leukoencephalopathy is a fatal disease, often with a delayed diagnosis in index cases. This study provides a systematic overview of neuroimaging findings in CSF1R-related leukoencephalopathy, comprising MRI, CT, PET, SPECT, and DaTscan results. This comprehensive compilation derives recommendations regarding the diagnostic workup of suspected cases. While a rare disease, our analyses comprise a total of 195 CSF1R-related leukoencephalopathy cases. Many different neuroimaging features have been reported, including cerebral atrophy and confluent white matter lesions with a predilection to the frontoparietal, callosal, and periventricular anatomical regions. Additionally, calcifications, mostly located frontally or symmetrically aligned with the upper edges of the lateral ventricles, have been described.

Findings in the context of existing evidence

In our summarized cohort, reported cases of CSF1R-related leukoencephalopathy with neuroimaging data consist of slightly more women (54%). In addition, women had a statistically significant earlier onset of disease (40 years vs 43 years in men), which is in line with findings from earlier studies [13]. Mutations in the CSF1R gene have been linked to primary microgliopathy disrupting the axon-glia integrity [3]. Based on this, it has been speculated that the different prevalence and disease onsets could be caused by sex differences in microglia function, as it has been observed in animal models functionally depleted of microglia cells [47]. A study in mice showed that microglia present a more inflammatory phenotype in female mice compared to male mice in both healthy and HDLS model disease. Also, female mice seem to express higher levels of P2X4 and P2Y4 (purinergic) receptors, which are associated with microglial activation and their transition from a quiescent to an inflammatory phenotype (reviewed in the study by Salter and Stevens [48]).

The cerebral white matter is a predilection site in CSF1R-related leukoencephalopathy, where microglial cells are abundant [48]. White matter lesions are also among the most typical neuroimaging findings in CSF1R-related leukoencephalopathy. These white matter lesions are typical without contrast-enhancement, and with a predilection to the frontal and parietal lobes and both the periventricular and deep white matter. They also seem to confluence and become more symmetric with ongoing disease progression [49]. Also, other strategic white matter structures can be affected in CSF1R-related leukoencephalopathy, such as the corpus callosum, presenting with overall thinning or calcifications [49]. Interestingly, a pathology study in seven autopsy brains identified a potential pathology correlate for the preferential callosal atrophy; the study found larger spheroids as well as more microglial alteration in the corpus callosum compared to the centrum semiovale [50]. However, the underlying cellular/molecular pathology that gives rise to the MRI white matter lesions in CSFR1-related leukoencephalopathy is still largely unknown due to the scarcity of studies correlating MRI features with corresponding pathology findings. One autopsy of a close relative of a CSF1R-related leukoencephalopathy case found significant tissue pathology within the deep cerebral white matter, the periventricular white matter, and the arcuate fibers: demyelination and vacuolization of myelin, axonal spheroids, gliosis, lipid-laden macrophages, as well as diffuse and widespread microglial activation. In contrast, the white matter tracts in the brainstem, including the medullary pyramids, were mostly unaffected [51].

Unlike some other adult-onset leukodystrophies [52], blood-brain-barrier disruption with contrast-enhancement has only rarely been reported in CSF1R-related leukoencephalopathy [21]. Meanwhile, diffusion restriction is a more common finding. The long-lasting hindered or restricted diffusion is a helpful imaging sign to raise the possibility of CSF1R-related leukoencephalopathy since this is seldomly seen in other neuroinflammatory or neurodegenerative disorders. It can be speculated that the signal may arise from fluid trapped between degenerating layers of the myelin sheets [26, 53]. This would be in line with histological findings in ALSP of axonal swelling and white matter degeneration [54].

In contrast to white matter involvement, gray matter involvement has less consistently been reported in CSF1R-related leukoencephalopathy cases. While microglia are also present in the gray matter [48], pathology studies in HDLS (in work performed before the CSF1R-gene mutations were discovered) suggest a paucity of microglial activation within the cortex [51, 55]. Yet, brain atrophy, predominantly located in the frontoparietal regions, as well as cortical volume loss has been described by several studies. It is likely that Wallerian degeneration from white matter axonal degeneration is at least partly responsible for the overt signs of brain volume loss [47, 51]. In addition, data from animal models with functional microglia ablation showed that healthy microglia support the survival of specific types of nerve cells, such as neurons in different cortical layers. Inhibition of microglia's activity showed decreased secretion of insulin-like growth factor 1 in mice. As a result, a reduced level of insulin-like growth factor 1 was related to apoptosis induction of layer V neurons and its death [47].

Importantly, white matter lesions and brain atrophy are highly unspecific MRI features. Thus, common mimics of CSF1R-related leukoencephalopathy include diseases with predominant white matter involvement, such as multiple sclerosis, cerebral vasculitis or cerebral autosomal dominant/recessive arteriopathy with subcortical infarcts and leukoencephalopathy [26, 56]. However, the concomitant presence of spotty, stepping-stone or serpiginous calcifications, mostly located in the frontal or periventricular white matter, could be levered since these imaging hallmarks are not typically observed in multiple sclerosis or vasculitis. The cause of these calcifications is insufficiently understood. Yet it has been shown that microglial ablation in mice by a CSF1R inhibitor does lead to aggravated vascular calcifications [57].

Nuclear medicine imaging modalities can also be harnessed in the diagnostic workup of suspected CSF1R-related leukoencephalopathy cases. FDG-PET may show glucose hypometabolism in the cerebral lobes, especially in the frontal lobes [58]. In brain SPECT scans, the most frequent and prominent finding was hypometabolism in the frontal and parietal lobes, but mild hypometabolism in basal ganglia, thalamus, and putamen was also identified.

Limitations

With the orphan disease status of CSF1R-related leukoencephalopathies, there is an extreme scarcity of cases. Hence, redundant reporting of cases among different publications is possible. Even though we screened respective publications for redundant cases, it is possible that cases synthesized by this review are listed twice as separate cases. However, with the relatively large number of eligible cases, overall conclusions on imaging findings are not likely to be affected.

Recommendations for neuroimaging

There are currently no guidelines or consensus for appropriate imaging of CSF1R-related leukoencephalopathy. Based on the summative data in this review, we primarily recommend brain MRI to be performed for differential diagnostics and follow-up of persons with CSF1R-related leukoencephalopathy. Ideally, the brain MRI protocol should entail high-spatial-resolution 3D anatomical imaging, which will allow for the visualization of both white matter lesions and atrophy with high sensitivity. Furthermore, when possible, administration of a gadolinium-based contrast agent is recommended to identify concurrent blood-brain-barrier disruption, which also helps to distinguish inflammatory white matter lesions from degenerative/microvascular white matter lesions. Diffusion-weighted imaging should also be included since long-lasting hindered/restricted diffusion in the leading edge of white matter lesions is relatively specific to leukoencephalopathies/leukodystrophies, differing them from inflammatory and other neurodegenerative lesions [52]. SWI is also recommended, ideally while preserving phase data. This will allow for the detection of parenchymal calcifications, which is also uncharacteristic of neuroinflammatory and other neurodegenerative disorders. In patients with suspected spinal involvement, there may be added diagnostic value in spinal cord imaging to characterize the spinal lesion burden.

Brain CT provides added value over MRI and is therefore also recommended, especially if SWI is not applied, due to its sensitivity to small white matter calcifications. Though calcium-containing compounds, even in small quantities, can often be detected on susceptibility-weighted MRI, this depends on the composition of the compound and is not always easy to interpret [59]. To visualize brain parenchymal calcifications, thin slices of ≤1mm are recommended as well as multiplanar reconstructions in all three planes.

The clinical value of nuclear medicine modalities is currently rather uncertain, which is why we cannot give a general recommendation to employ them. However, seeing that CSF1R-related leukoencephalopathy is a primary microgliopathy, microglia-related tracers in PET, such as 18-kDa translocator protein tracers [60], are likely useful, but it is yet an unexplored field of research and an unmet need.

To facilitate clinicians and radiologists to be able to identify the findings highlighted in this systematic review, we have provided pictorial examples of the neuroimaging phenotypes in Fig. 2. Recommendations for brain CT and MRI are provided in Table 3. For research purposes, we recommend using the brain MRI scoring system developed by Sundal et al. to quantify the lesion burden [61]. The scale considers imaging characteristics that are common in CSF1R-related leukoencephalopathy, such as white matter changes and atrophy with a topographical delineation and an overall grading scale score. Additional references from the literature search [62-107] can be found in the Supplementary materials.

Pictorial examples of neuroimaging findings in CSF1R-related leukoencephalopathy.

Examples of typical findings in CSF1R-leukoencephalopathy in a 52-year-old male (top row), 43-year-old female (middle row), and 41-year-old female (bottom row). A, Confluent symmetric frontoparietal white matter lesions with a sparing of the U-fibers on axial T2-weighted brain magnetic resonance imaging (MRI). Moderate symmetric frontoparietal atrophy is also seen; B, lesions extending from the frontoparietal white matter along the corticospinal tracts (arrows) to the pons on coronal T2-weighted brain fluid-attenuated inversion recovery (FLAIR) MRI; C, multiple foci of lasting diffusion restriction (arrows), confirmed on apparent diffusion coefficient map (not shown) on axial diffusion-weighted brain MRI; D, gadolinium contrast-enhancement (arrows) in the periphery of white matter lesions on axial T1-weighted brain FLAIR MRI. Of note, even though contrast-enhancement can be an imaging feature of CSF1R-leukoencephalopathy, it was only rarely reported in our systematic review; E, an infratentorial lesion on the right posterior aspects of the pons (arrow) on axial T2-weighted brain FLAIR MRI; F, short spinal cord lesion (arrow) in the dorsal columns at the C3 level on axial T2-weighted cervical MRI; G, calcification (arrow) in the left ffrontal white matter on axial brain computed tomography (CT); H, calcifications in a “stepping-stone” pattern (arrows) along with the corpus on sagittal brain CT; I, corpus callosal atrophy on sagittal T1-weighted brain MRI.

Table 3. Recommended brain imaging Imaging modality Imaging details Main findings in CSF1R-related leukoencephalopathy Brain CT Thin slices (≤1 mm) reconstructed in three planes Atrophy white matter lesions, calcifications Brain MRI 3D T1WI Atrophy, severe demyelination 3D T2-weighted FLAIR White matter lesions 2D/3D T2WI White matter lesions 3D SWI (with phase data) Calcifications 2D DWI Long-lasting diffusion restriction T1-weighted with GBCA Contrast-enhancement Abbreviations: CT, computed tomography; DWI, diffusion-weighted imaging; FLAIR, fluid-attenuated inversion recovery; GBCA, gadolinium-based contrast agent; MRI, magnetic resonance imaging; SWI, susceptibility-weighted imaging; WI, weighted imaging. Conclusions

This systematic review provides high-level evidence for neuroimaging findings in CSF1R-related leukoencephalopathy. These imaging features include MRI white matter lesions, mostly in the frontoparietal and periventricular regions, brain atrophy including callosal thinning, (persistent) diffusion restriction, as well as spotty white matter calcifications on CT. The synthesis of these findings as well as the proposed imaging guidelines herein can expedite diagnostic workup of suspected CSF1R-related leukoencephalopathy cases and also potentially foster the identification of additional imaging biomarkers.

Acknowledgments

We thank the search consultation group of Karolinska Institutet Library for performing the initial literature search. We would also like to acknowledge Dan and Stina Englesson for their assistance with graphical illustrations.

Funding

No specific funding was obtained to perform this systematic review.

Conflict of interests

Goda-Camille Mickeviciute was funded by the Erasmus plus program. Monika Valiuskyte reports no conflict of interests. Michael Plattén was funded by Karolinska Institutet, not directly related to the conductance of this study. Zbigniew K. Wszolek is partially supported by the NIH/NIA and NIH/NINDS (1U19AG063911, FAIN: U19AG063911), Mayo Clinic Center for Regenerative Medicine, Mayo Clinic in Florida Focused Research Team Program, the gifts from The Sol Goldman Charitable Trust, and the Donald G. and Jodi P. Heeringa Family, the Haworth Family Professorship in Neurodegenerative Diseases fund, and The Albertson Parkinson's Research Foundation. He serves as PI or Co-PI on Biohaven Pharmaceuticals, Inc. (BHV4157-206 and BHV3241-301), Neuraly, Inc. (NLY01-PD-1), and Vigil Neuroscience, Inc. (VGL101-01.001) grants. He serves as Co-PI of the Mayo Clinic APDA Center for Advanced Re