Investigating Dyslexia through Diffusion Tensor Imaging across Ages: A Systematic Review

Dyslexia is a neurodevelopmental disorder that presents a deficit in accuracy and/or fluency while reading or spelling that is not expected given the level of cognitive functioning. Research indicates brain structural changes mainly in the left hemisphere, comprising arcuate fasciculus (AF) and corona radiata (CR). The purpose of this systematic review is to better understand the possible methods for analyzing Diffusion Tensor Imaging (DTI) data while accounting for the characteristics of dyslexia in the last decade of the literature. Among 124 articles screened from PubMed and Scopus, 49 met inclusion criteria, focusing on dyslexia without neurological or psychiatric comorbidities. Article selection involved paired evaluation, with a third reviewer resolving discrepancies. The selected articles were analyzed using two topics: (1) a demographic and cognitive assessment of the sample and (2) DTI acquisition and analysis. Predominantly, studies centered on English-speaking children with reading difficulties, with preserved non-verbal intelligence, attention, and memory, and deficits in reading tests, rapid automatic naming, and phonological awareness. Structural differences were found mainly in the left AF in all ages and in the bilateral superior longitudinal fasciculus for readers-children and adults. A better understanding of structural brain changes of dyslexia and neuroadaptations can be a guide for future interventions.


Introduction
Dyslexia is a neurodevelopmental disorder that impairs fluency and/or speed of reading, as well as word recognition; it may also impact spelling.The difficulties should not be explained by other cognitive, health, or socio-economic factors [1].The lack of accurate and fluent reading is a product of poor recognition and decoding abilities, and despite that, verbal comprehension is not equally affected in dyslexia [2,3].
Developmental dyslexia is classified within the framework of the diagnosis-specific learning disorder of reading [4], and it manifests in a spectrum from mild, to moderate, to severe [5].It is present in over 80% of people with a learning disability in studied languages, having some variations regarding severity and type of difficulties according to the language structural characteristics [6].Dyslexia's mean prevalence is estimated at around 7% of the world population, with a predominance in male individuals [2,3].Still, the incidence and prevalence are unclear due to the heterogeneity of literacy and language cultures with wide variances in terms of definitions, diagnostic instruments, rules, guidelines, and protocols for assessing dyslexic children and adults [1,7].
Two main medical classifications (ICD-11 and DSM-5) define the diagnostic criteria for dyslexia, and the assessment focuses mainly on identifying the reading and spelling discrepancies compared to the general population performance together with the assessment of other aspects that could confirm or rule out the diagnosis, for example, intelligence, phonological awareness, word and pseudoword reading, verbal fluency, verbal working memory, reading, and naming under stress [8].This careful and time-consuming evaluation aims to support the clinical exclusion criteria, which contributes to reducing the prevalence by avoiding wrongful diagnoses of individuals struggling to read.Furthermore, there has been an effortful search in different fields for a better understanding of dyslexia's development and neuroimaging contribution in clarifying neuroanatomical forming behind reading and dyslexia.
Clinical neuroimaging has been transformed by Diffusion-Weighted Imaging (DWI) and Diffusion Tensor Imaging (DTI), making it possible to examine the brain's architecture and identify pathology earlier and more accurately than traditional magnetic resonance imaging sequences.Nowadays, diffusion is already used in clinical practice for stroke, trauma, tumors, demyelinating conditions, and neurosurgical planning.In psychiatric and neurological conditions, it is still used mainly in the research context [9].
The foundation of diffusion imaging lies in the behavior of water molecules, which move freely through space equally in all directions when unimpeded by structures.However, when encountering obstacles like cell membranes, water molecules tend to diffuse in alignment with the orientation of those barriers [10].Magnetic resonance imaging, facilitated by DWI sequences, enables the measurement of water displacement in various directions for a brief duration.This information can then be used to assess tissue integrity, particularly in white matter fiber pathways [9].
Recent advancements in DTI research have expanded the focus beyond the assessment of a straightforward diffusion scalar to emphasize the significance of the more intricate 3D diffusion pattern.Axial diffusivity (AD), radial diffusivity (RD), mean diffusivity (MD), apparent diffusion coefficient (ADC), and others are additional imaging metrics that are increasingly used [11].
A standard DTI metric utilized in evaluating a variety of neuropathologic processes, from traumatic brain injury to demyelinating illness, is fractional anisotropy (FA), which quantifies the degree of this directionality [11].Despite its promise for identifying subtle illnesses and alterations not discernible with conventional MRI sequences, the clinical applications of DTI have been questioned regarding the specificity of the findings [11].
In developmental neuroscience, DTI metrics have been shown to be a useful biomarker of white matter tract development and tissue injury, being used for treatment monitoring and potentially acting as outcome predictors.Children's normal and pathological brain maturation has been described by the ADC/FA scalars since it is well known that as brain myelination and maturation advance, ADC values fall and FA values rise [12].
The literature has shown that people with dyslexia or reading disability may show brain structural changes with lower FA values in the left frontal and temporoparietal regions that coincide with the majority of studies on the left arcuate fasciculus (AF) and corona radiata (CR).Few studies have suggested a role for the posterior part of the corpus callosum or more ventral tracts like the inferior longitudinal fasciculus (ILF) or the inferior frontooccipital fasciculus (IFOF) [13].And more recently, a meta-analysis found no differences between dyslexics and typical readers when observing studies that conducted Voxel-Based Analysis (VBA) of FA and compared it to reading ability [14].
The purpose of this systematic review is to search for a better understanding of the multitude of possible methods for analyzing DTI data while accounting for the characteristics of a clinical population such as developmental dyslexia in the literature in the last 10 years.

Inclusion Criteria
The review included only original articles written in English published within the last 10 years, and full text available about dyslexia in structural analyses by DTI.According to the patient/problem, intervention, comparison, and outcome (PICO) criterion, the problem was: the structural brain changes in dyslexia are unclear; the intervention was: structural analysis by DTI; the comparison was: differences between dyslexia and typical readers volunteers; and the outcome was: the structural brain pattern of dyslexia of development.

Exclusion Criteria
We excluded studies based on the following criteria: (i) reviews or meta-analyses; (ii) publications written in languages other than English; (iii) indexed articles published in more than one database (duplicates); (iv) articles that included dyslexia with other neurological or psychiatric comorbidities such as stroke, brain injury, epilepsy, autism, traumatic brain injury, aphasia, and mood disorders; (v) articles that performed any analysis other than DTI, such as machine learning, graph theory, and only methodological comparison; (vi) articles in which the diagnosis of dyslexia is unclear; (vii) articles without at least one outcome or method of analysis reporting DTI measures and/or correlation with demographic or neuropsychological measures; (viii) case reports; (ix) neonates; and (x) dyslexia with genetic alterations.

Data Compilation
In this review, seven of the authors (B.M., M.Y.B., E.M.D., L.F.C., A.S.C., K.L., and M.P.N.), in pairs, independently and randomly analyzed, reviewed, and assessed the eligibility of titles and abstracts according to the strategy of established search.The authors B.M., M.Y.B., E.M.D., L.F.C., A.S.C., K.L., and M.P.N. selected the final articles by evaluating the texts that met the selection criteria.The authors B.M., E.M.D., L.F.C., and A.S.C. were responsible for the search for the demographic, clinical, and neuropsychological charac-teristics of volunteers and dyslexia patients, being checked by the senior authors (K.L. and M.P.N.).The authors B.M., M.Y.B., L.F.C., and A.S.C. searched for the characteristics of structural brain analyses and their outcomes, and all data were checked by the senior authors (K.L. and M.P.N.).All of the authors contributed to writing the entire text of this review.

Data Extraction
The selected articles were analyzed using two topics, which were represented in tables that addressed the following characteristics: (1) the demographic characteristics of the population sample, their language, their nationality, and the neuropsychological tools used to characterize the reading disorder; and (2) the characteristics of DTI acquisition, the parameters used in the image analyses and corrections, the structural outcomes between groups, and the correlation with clinical data when reported.

Risk of Bias Assessment
The selection of articles was performed in pairs, and a third independent author decided if the articles should be included.The data selected in the tables were divided by the authors into the groups already described above, and the checking of the data was carried out by the following group.The final inclusion of studies into the systematic review was by agreement of all reviewers.

Data Analysis
The data from the articles included in the tables were analyzed descriptively using the percentage, mean, and standard deviation; the variation to characterize each factor attributed to the demographic and neuropsychological characteristics of the participants in each study; and the characteristics of the acquisition, analysis, and results of the structural assessment performed by DTI image acquisition.

Overview of the Screening Process of the Included Studies
Following the inclusion and exclusion criteria described above, we found 124 articles in the last ten years throughout the Scopus and PubMed databases, with 116 from Scopus and 8 from PubMed.Of the 116 articles found in Scopus, 64 were excluded after screening, two studies were with participants with alexia, one with aphasia, two with autism, one with mood disorder, two neonates, 20 without dyslexia diagnosis, two with genetic alterations, nine with brain injury (such as stroke, epilepsy, cortical lesion, and TBI), six were case reports, three were meta-analyses, seven were reviews, three were methodological studies, and six only featured morphometric analyses.The eligibility analysis excluded a further four articles.Three studies reported different methodologies of DTI analysis (machine learning, graph theory, and manual and automatic segmentation), and one did not report the DTI results, resulting in 48 studies included in this selection.Out of the eight articles identified in PubMed, five were excluded during screening: two lacked a dyslexia diagnosis, and three were duplicates from Scopus.After the eligibility assessment, an additional two studies were excluded.One study conducted DTI analysis using machine learning, while another did not report DTI results.Consequently, only one study from PubMed was included  in this systematic review, 48 were included from Scopus, and one was included from PubMed, as shown in Figure 1.

Demographic and Neuropsychological Characteristics of Studied Dyslexia Subjects
This systematic review aimed to provide an overview of dyslexia in various stages of life, from early childhood (preschool to kindergarten) with a familiar history of dyslexia to elementary school children, adolescents, and adults, following the structural brain changes involved in this neurobiological disorder.In the selected studies about dyslexia and structural brain analysis by tractography included in this systematic review, 15% of studies [16][17][18][19][20][21][22][23] included only children below 6 years old with the family risk of dyslexia before reading acquisition (classified as pre-readers); followed by 70% of the studies that analyzed older groups, at different stages of reading proficiency (classified as reading children) from 7 years old up to 19, males and females (classified as readers); and 15% of studied male and female adults aged from 20 to 33 years old (classified as reading adults) (Table 1).Another important aspect is that dyslexia can manifest similarly across languages, but the characteristics and challenges may vary based on the language's structure, writing system, and phonological rules.Because of this, we include the demographic characteristics of the country and the language participants spoke.The pre-reader studies were done mainly in the United States with English language speakers (50%) [16][17][18]22], followed by 40% [20, 21,23] in Belgium with Dutch speakers, and 10% [19] in Germany with German speakers, including young children under 6 years old male and females and balanced distribution (a total of 193 females to 215 males) and with a sample variation of 10 to 46 children in each comparison group.In studies with reading-stage children, 47% spoke Englishfrom the USA (45%) [25,27,[31][32][33]39,42,44,48,50,[52][53][54]56] and Canada (2%) [28,29]; 17% spoke French (the study was carried out in France [26,30,34,38,45]); 11% spoke Dutch-the studies were carried out in Belgium and Netherlands [37,41]; 8% were German; 6% spoke Mandarin-the studies were carried out in China [43] and Taiwan [36]; and 3% of studies were done with speakers of Arabic (Egypt) [35], Spanish (from Spain) [47], Portuguese (from Brazil) [49], or Italian (from Italy) [51].In adults, the language of studies was less varied: 38% spoke English, and the studies were carried out in New Zealand [59] and the USA [62,64]; 25% spoke German (Germany) [58,60], 25% Dutch (Belgium) [61,63], and 12% Finnish (Finland) [57] (Table 1).
The neuropsychological characterization of the studied subjects followed the specificity of dyslexia diagnostic criteria, that the neuropsychological tests that were used included assessment of intelligence quotient (IQ); word and non-word reading and spelling tests; reading comprehension tests; rapid automatized naming (RAN); phonological awareness; and language, attention, and executive functions (description in Table 1).The primary purpose of the neuropsychological evaluation was to compare the performance of the dyslexic groups with the control group.
Considering the neuropsychological outcome of reading children in the three language groups most represented in this revision (English, French, and Dutch), there was no difference in the cognitive skills found impaired in dyslexic volunteers compared to the controls.
The basic pulse sequence repetition time (TR) and echo time (TE) parameters ranged from 3000 to 14,000 ms and from 55 to 110 ms, respectively; the slice image ranged from 23 slices with 5 mm of thickness to 160 slices with around 1.7 mm to cover the entire brain, and the field of view (FOV) ranged from 208 to 282 mm.
Artifacts in DWI acquisitions lead to errors in tensor estimation, and Eddy Current (EC) distortions and Head Motion (HM) are the two primary intrinsic DTI acquisition abnormalities that may obliterate the voxel-wise correlation across all the DWIs.Most of the selected studies (73%) reported in the pre-processing step comprise the EC and HM (with a cutoff from 1.5 mm to 6 mm) corrections.Other studies (10%) reported corrections to the image (EPI) distortions, 4% for the Gibbs artifact (truncation or ringing artifact) or Marchenko-Pastur Principal Component Analysis (MP-PCA), and only 14% did not report any correction in the preprocessing image step.Some FSL tools were used for these corrections such as CATNAP (Coregistration, Adjustment, and Tensor-solving a Nicely Automated Program), which is a data processing pipeline for Philips PAR/REC Magnetic Resonance data files, performing motion correction for both diffusion and structural images using FSL FLIRT; it adjusts the diffusion gradient directions for scanner settings (i.e., slice angulation, slice orientation, etc.) and motion correction (i.e., the rotational component of the applied transformation) and computes tensor and derived quantities (FA, MD, colormaps, eigenvalues, etc).Also, the TOPUP tool of FSL is used to correct images of the susceptibility-induced distortions (fix EPI distortions), and QUAD (Quality Assessment for DMRI) for automatically performing image quality control (QC) at the single subject .
Different types of atlas were reported in the selected studies to segment the cortical and white matter region; 54% used one of FSL atlas such as Julich-Brain Cytoarchitectonic Atlas, MNI (Montreal Neurological Institute) Structural Atlas, JHU (Johns Hopkins University) ICBM-DTI-81 White-Matter Tractography Atlas, and Harvard-Oxford atlas; 19% used one of three FS's atlas (Desikan-Killiany, Destrieux, and Desikan-Killiany-Tourville cortical atlas); 9% used native space; 7% used the automated anatomical atlas (AAL), the template for SPM, AFQ, ExploreDTI, and PANDA software; 3% used Talairach atlas; and 9% did not report this information.
Regarding the results of the structural analysis of the brain, 89% of the studies described this finding, with the predominance of a decrease in FA in the left hemisphere, when comparing the dyslexia group with the control group, and 10% of the studies did not show a significant difference between the groups [19,25,46,59,64], occurring mainly in the group of adults (25%) [59,64].
The tractography analyses of the pre-reader group reported by the selected studies were based on the FA changes according to the risk of familial history of dyslexia (FHD); 13% reported higher FA of right SLF [16] or CC [17] compared to the children with positive FHD than TR negative FHD, and 50% showed lower FA of the left AF (as also long portion) [18,[21][22][23] and 13% in IFOF [20], as shown in Figure 2, highlighting the frequency in green color.Almost all of these regions also showed a positive correlation with age and some language tests and predicted decoding skills or reading impairment.In the reading children, 42% of these studies reported a significant change in FA values (27% lower [27,38,[42][43][44][45]47,49,55] and 15% higher values [40,41,48,52,54] in dyslexic children than the TR group), decreasing mainly in the left tracts such as AF, SLF, ILF, CR, and IFOF and increasing in the right SLF, aThR, cingulum, and CC.These studies also showed 6% of AD [43,50] decreased in dyslexic children in IFOF, SLF, UF, FMn, CG, and the right ThR [50], as well as in the left AF and ILF [43], and 12% of MD [28,32,39] changes between groups (6% higher [28,32] and 6% lower values in dyslexic children [32,39]).Only 3% of these studies reported higher values of ODI in the left ventral optical tract [33] of DYX and lower values of RD and HMOA (in UF of DYX males) [24], as shown in Figure 2, highlighting the frequency in red color.No group differences were reported in 6% of these studies [25,46], and 30% did not report brain changes [26,[29][30][31][34][35][36][37]51,56] In the selected studies with adult participants, 50% showed group differences, in which the DYX had low FA in lateral geniculate nucleus, TOC, and left-AF [60,63], and another DTI metric was the QA with high values for DYX group in left-UF, CS, CC, ThR, FMj, and parietal tracts and low in left SLF, CS, and OF in comparison to TR group [57], as shown in Figure 2, highlighting the frequency in purple color.In total, 25% of these studies did not find a group difference [59,64] or report this information [61,62].

Discussion
This systematic review provided an overview of the main structural brain changes findings by the DTI technique in developmental dyslexia in different age groups, including early at-risk children due to a family history of the disorder, as well as children and adults with reading disorders who have been diagnosed with dyslexia.By comparing the results of several studies, we describe converging evidence on structural abnormalities of the brain and the associations between imaging results and changes in this population's characteristics.This systematic review observed that the assessment of different cognitive functions by the neuropsychological instruments of the selected studies varied according to age, country, and language; however, the lack of standardized procedures among age and language groups drastically reduces the possibility of comparing the behavioral outcomes and its relation to the structural brain changes.While some aspects of reading, mainly word decoding and recognition, are the most assessed skills in all the age groups, others such as language, attention, and working memory are reported exceptionally.The intelligence measure showed an expected outcome in terms of experimental and control group matching, with more studies reporting lower performance of children with dyslexia on verbal but not on non-verbal tests of intelligence.The intelligence assessment has generated a long-time debate in the field of learning, and while in typically developing children intelligence performance generally correlates with the achievement level, in reading impairment the intellectual disability is inconsistent with a diagnosis of dyslexia [65].
The importance of matching groups on intelligence measures is also supported by a growing body of evidence linking the brain white matter organization to intelligence level and processing efficiency.A recent study showed stronger integration of white matter structures within a local community (ex.frontal region) and especially with external brain networks (ex.frontal to parietal regions) in adults scoring high on non-verbal tests compared to average performers [66].
The structural brain changes shown by the DTI measurements were examined or extracted using the atlases, and nearly all studies (92%) examined the fractional anisotropy, which represents the directionality and organization of tissue microstructure; other studies examined some of its variants, including RA, QA, and HMOA, providing additional information to FA.The studies also examined the following measures: the AD measure, which can show changes in the density or integrity of axons within white matter pathways; the RD measure, whose increase frequently denotes disruptions in the microstructure of white matter, such as demyelination or axonal damage; and the MD measure, which, along with some variations like Dav and MDe, typically shows higher values that indicate decreased tissue integrity and increased diffusion.
In children before reading acquisition who had a positive family history of dyslexia, the structural changes analyzed by DTI metrics were shown mainly by the FA alterations.The decrease in FA was predominant in the left of the AF, as well as for IFOF, a result also reported in a recent study with dyslexic children with this profile [21], and a pattern of increased FA occurred in few studies, and only in the right hemisphere of the SLF and sCC, which may suggest possible early neural compensatory mechanisms in the right hemisphere [17].This pattern was also similar in reading children, with low FA values mainly in the left hemisphere involving the AF, but also high FA values in the right hemisphere of AF, showing more neuroplasticity signals than the other young group.In addition, in these reader groups the structural changes covered more areas, still with a predominance of the left hemisphere, and were identified in other DTI metrics, such as low AD values in AF, SLF, IFOF, UF, ThR, ILF, CG, and CS, as well as a high MD values in AF, and low in UF and CR.Children submitted to reading intervention show an increase in MD values in AF and other reading brain circuitry such as the left ILF and posterior CC [67].Variations of anisotropy, such as RA and HMOA, also showed low values in brain structural changes reported in the articles with reader children, but only FA showed high values in the following tracts AF, SLF, CC, ThR, ILF, and CG.
In dyslexic adults, the percentage of structural changes reported was much lower than those reported in children.It seems that with brain development, in adulthood, the structural differences of dyslexia become less evident due to neuroadaptation.However, the pattern of decreased FA measurements in dyslexic adults remained the same, with predominance in the left hemisphere of AF and the region that comprises the AF (lateral geniculate nucleus, and temporo-occipital cortex), as well as bilaterally SLF, CC, CS, and the left of UF and ThR by the QA measurements.This anisotropy variation, the QA measure, also showed an increase in the left side of SLF, VOF, and CS, which may represent neuroplasticity in adulthood.Nonetheless, due to literature scarcity on dyslexia in adults it is difficult to compare these results found in the review with other studies.
Another form of result widely explored between studies was the analysis of association, be it through correlation or the prediction of structural data with demographic or neuropsychological data, and this occurred with a greater incidence of significant findings between studies than the actual comparison of structural changes.These association analyses are normally applied to assist in the interpretation of results, mainly in studies with adults when there were no structural difference between the groups, but measures mainly of FA were positively or negatively correlated with the results of neuropsychological tests.In children, the structural changes helped predict some performance in neuropsychological tests, especially in children with a family risk of dyslexia.
In addition to the structural results found in participants with dyslexia and controls, this review considered how DTI data were acquired, processed, and extracted as an analysis.In general, all studies included in this review took good care to ensure good image and data quality, reducing bias in the interpretation of results.
The DTI acquired in high magnetic field equipment, such as 3 Tesla, can increase the capacity to acquire higher resolution scans more quickly, with higher b-values and thinner slices, as well as to increase tissue contrast and reduce background noise (thereby increasing the signal-to-noise ratio and contrast-to-noise ratio) [68]; this magnetic field was used in 88% of the studies (49% Siemens, 35% Philips, and 4% GE), in the acquisition of both DKI and DTI sequences to study tractography.The latter sequence is the more traditional sequence and was used in most of the selected articles in our systematic review.The difference between the two is that DKI significantly reduces the error of dODF orientation estimates compared to DTI and makes it possible to detect crossing fibers, which leads to a noticeable improvement in tractography across regions with complex fiber bundle geometries [69,70].DKI-based tractography has potential benefits, especially in clinical contexts when time is of the essence [71].
The acquisition parameters of the DTI is important because they affect values of white matter (WM) scalar metrics, including FA, MD, Signal Noise Ratio (SNR), and even entire brain tractography investigations.These acquisition parameters include the diffusion sensitivity coefficient (b-value), which is a factor that reflects the strength and timing of those gradients used to generate DWI, as well as the reliability of DTI results concerning image and data quality; diffusion directions, in which there are more directions the longer the acquisition time; and voxel size (the smaller the size, the higher the quality [72,73]).
The adequate b-value for diffusion imaging quality evaluation in dyslexia was unclear and varied according to the equipment's magnetic field.In a heath brain, a greater b-value usually results in a poorer SNR and image quality because increased signal attenuation owing to diffusion as well as increased TE (and therefore additional signal loss due to T2 decay), would lead to the decrease of MD, AD, and RD when the gradient directions and voxel resolution remained constant [73,74].In the literature, this is more evident at the low field strength of the MR scanner as 1.5 T [75], and only 10% of the selected studies of this review used this field, less evident in high field strength (3 T and 7 T) due to their relatively high SNR, in which the increase in b value has little influence on the decrease in SNR, showing the best image and data quality with the respective b-values, 200 and 900 s/mm 2 [73].The selected studies included in this review reported normally more than one b-value and the MRI of 1.5T used b-values ranging from 800 to 1000 s/mm 2 and the 3 T from 700 to 5000 s/mm 2 [73,74].
The increased number of diffusion-encoding gradient directions can also improve DTI quality by averaging and strengthening the tensor estimation, and the opposite can reduce all DTI scalar values' accuracy and precision; a minimum of 18 diffusion directions is advised to produce trustworthy DTI scalar results using the TBSS toolbox of FSL software [76].In our study, just a small number of the selected studies (11.8%) used fewer than 30 directions, and of these, 7.8% used less than 18 directions, whereas the majority (78.4%) used 30 directions or more (41.3%used more than 60 directions).
Considering the impact of MRI acquisition parameters on the values of DTI measures, changes in the number of gradients and voxel resolution have the greatest impact on the FA, but variations in the b-value have a special effect on MD [72].Another study also showed that the number of gradient directions was more relevant than the spatial resolution in some quantitative measures of DTI, such as tract volume, median fiber density, and mean FA, but this did not occur for all tracts evaluated in the same way, only for SLF and IFOF [77].
One of the most defining and defiant components of building a tracking algorithm is determining the underlying model that connects the raw dMRI images to the local fiber orientations, and presently, there is a wide range of software packages that incorporate higher-order fiber-tracking techniques that can calculate the relative contributions and orientations of several fiber populations within each voxel, which are easily applied to clinically relevant data sets [78].
A wide range of processing functions are offered by these software packages, such as tensor calculations, fiber tracking, visualization, statistical analysis, quantitative measure extraction from DTI datasets, and integration with additional neuroimaging tools.They vary, though, when it comes to the tractography algorithms that are applied, such as probabilistic, which produces a vast collection or distribution of potential trajectories from each seed point, and deterministic, which assumes a unique fiber orientation estimate in each voxel [79]; as well as the local approach, which is a fast and widely used method, it follows the local orientations of previously extracted fibers independently of each other, but the sum of small errors in these local orientations can significantly affect the final result, making it a very weak predictor of data with little quantitative significance or biological [78].On the other hand, the global methods offer improved stability concerning noise and imaging artifacts and a greater agreement with the real dMRI data that was recorded; however, they rely on stochastic optimization approaches and hence do not guarantee convergence to a globally optimal solution [80].
The choice of software to analyze the DTI normally depends on the specific analysis requirements, familiarity with the software, and preference for user interface and workflow, and the studies normally used more than one software to employ specific tools to conduct the entire analysis.
Several image adjustments are usually conducted before DTI analysis to enhance the data quality and reduce common artifacts that may occur during data acquisition, allowing for appropriate interpretation and trustworthy outcomes.Most of the adjustments reported by the studies were Eddy current correction (74%) which aligns the DWI to a reference image acquired without diffusion weighting [81], and motion correction (74%); few explore structural analyses but they realign the image to compensate for subject motion, ensuring that the diffusion measurements are accurate and consistent across the dataset [82].However, in low frequency the Gibbs ringing correction (4%) of the discontinuities in the kspace data caused by undersampling during image acquisition was also reported, resulting in obscure anatomical structures and affecting diffusion measurements, as well as EPI distortion correction (10%), considering the spatial variations in the strength and direction of the magnetic field gradients used for diffusion encoding.Unhappily, 14% of the studies did not report any adjustment before DTI analyses.
Regarding the type of quantitative DTI analyses reported by the studies, most (69.2%)used voxel-wise, which analyzes each voxel to identify differences in diffusion properties between groups or conditions, providing more detailed spatial information about diffusion metrics at a voxel-level; this is present in some of the software such as FSL, SPM, MRItrix, and ExploreDTI.Interestingly, a lot of articles reported this analysis as the whole brain since voxel-based analysis includes all the cerebro voxels.Also, ROI-based analysis used in 61.5% of studies is frequently conducted after voxel-based, making it challenging to determine which method was used in each study.
The outcome description of the studies was based on anatomical regions of interest or specific tracts outlined by a well-established atlas that facilitates the interpretation of imaging data by providing standardized anatomical labeling and spatial coordinates.While the atlases share the goal of delineating brain structures and regions, they differ in several aspects, including their origins, resolutions, and intended applications.
The study's limitations stem mainly from the differences in how dyslexia was defined across studies.The use of the terms dyslexia, developmental dyslexia, reading disorders, or never reading difficulties point to the lack of a common terminology and diagnosis criteria.The impact of the findings of this study could be that the results may be possibly drawn from a very heterogeneous sample.However, the cognitive testing of participants may ameliorate this and emerge as a potential tool mainly in internationally normed tests.
Another limitation of the study was the influence of language on dyslexia's reading acquisition history.It is a well-known fact that readers in irregular language systems have longer reading acquisition and struggling readers may develop different compensation strategies.
Studies on neuroimaging may be limited by variations in image acquisition parameterization; however, despite this wide range, all studies used optimal acquisition and analysis parameters that did not affect the comparability of results, even in cases where certain fundamental information was not stated.In terms of results, a certain study focused on describing the anatomical regions examined rather than the tract as most studies did, taking into account the tracts involved in the regions described.

Conclusions
This systematic review of structural alterations in the brain associated with dyslexia revealed that over the past ten years studies on children have outnumbered those on adults, primarily focusing on boys and the English language.The studies also showed that brain changes concentrated in FA reduction in the fasciculus arcuate of the left hemisphere at all ages, and in the left superior longitudinal fascicle for reading in children and adults, as well as an increase in the right hemisphere, which may indicate signs of neuroadaptation.A better understanding of structural brain changes of dyslexia and neuroadaptations can be a guide for future interventions.

Figure 1 .
Figure 1.PRISMA flowchart of this systematic review study, identifying at each stage the number of studies included and the reasons for excluding studies until the final stage of inclusion of studies.

Figure 2 .
Figure 2. Spider graphic of the diffusion tensor image (DTI) outcomes percentage distributed by the main tracts reported in the systematic review and their DTI metrics behavior found according to tract and age group of dyslexia participants (pre-reader children in green, reader children in red, and reader adults in purple).The arrows indicate the increase or decrease in DTI metrics in the dyslexia group in comparison to the control group.Abbreviations: AF: arcuate fasciculus; SLF: superior longitudinal fasciculus; IFOF: inferior fronto-occipital fasciculus; CC: corpus callosum; UF: uncinate fasciculus; ThR: thalamic radiations; ILF: inferior longitudinal fasciculus; CG: cingulate cortex; CS: corticospinal fasciculus; CR: corona radiata; OR: optic radiation; Forceps: forceps major and minor; VOF: ventral occipital fasciculus; FA: fractional anisotropy; AD: axial diffusivity; MD: mean diffusivity; QA: quantitative anisotropy; RA: relative anisotropy; and HMOA: hindrance-modulated oriented anisotropy.

Table 1 .
Demographic and neuropsychological assessment.
NR Note: Bold font indicates significant group differences.* time effect in DYX group, ** only 1 control had a high school diploma.Abbreviations: NA: Not Applicable; NR: Not Reported; FRD+: Children with Familial Risk for Dyslexia; FRD−: Children without Familial Risk for Dyslexia; TR: Typical reading; rg: reading gain; nrg: no rg; m: months; RAN: rapid automatized naming tasks; DYX: children with dyslexia; RD: Reading Disorder; RI: Reading Impairment; PreR: pre-reader children; BR: older reader children; FR: fluent reader children; PIQ: Performance IQ; FSIQ: Full-Scale Intelligence Quotient; CFT 20-R: Cattell's Fluid Intelligence Test, Scale 2; TrR: Treatment Responders; EF: Executive Functions; MD: Mood Disorders; WRD: word recognition deficits; RLD: reading and learning disabilities; SI: spelling impaired children; SRI: children with spelling and reading impairment; CFP: readers with comprehension or fluency problems; PA: Phonological awareness; PPVT: Peabody Picture Vocabulary Test; TONI or TONI-4: Test of Nonverbal Intelligence; CTONI: Comprehensive TONI; CTOPP: Comprehensive Test of Phonological Processing; TOWRE or TOWRE-2: Test of Word Reading Efficiency; WJ: the Reading Fluency subtest of the Woodcock-Johnson Test; WISC or WISC-IV: Wechsler Intelligence Scale for Children, 4th edition; WASI: Wechsler Abbreviated Scale of Intelligence; WAIS: Wechsler Adult Intelligence Scale, 3rd edition; WIAT: Wechsler Individual Achievement Test; BRIEF: Behaviour Rating Inventory of Executive Function; KBIT or KBIT-2: Kaufman Brief Intelligence Test; WRMT-R NU: Woodcock Reading Mastery Tests-Revised, Normative Update; WA: Word attention; SLRT-II: The Salzburg Reading and Spelling Test; YARC: York Assessment of Reading for Comprehension; GAPS: Grammar and Phonology Screening; CELF or CELF-4: Clinical Evaluation of Language Fundamentals; KTEA-II: Kaufman Test of Educational Achievement-Second Edition; CVLT: California Verbal Learning Test; WID: word identification; LWID: letter and WID; VAS: Visual Attention Span; SB4: Stanford-Binet Intelligence Scales-Fouth Edition; HLE: Home Literacy Environment; HOME: the Home Observation for Measurement of the Environment; DAT: Dyslexia Assessment Test; ASRS: Adult Self Report Scale for ADHD clinical assessment; TOC: Test of Orthographic Competence; TOMAL2: Test of Memory and Learning -Second Edition; MSEL: Mullen Scales of Early Learning; ARHQ: Adult Reading History Questionnaire; ADC: Adult Dyslexia Checklist; HSP: Hamburger-Schreibprobe; SLS: Salzburger-Lese-Sreening; FFQ: five factor questionnaire; DAWBA: Diagnostic and Well-Being Assessment; SWE: Sight Word Efficiency; DS: digit span; VSA: visual spatial attention; BISC: Bielefeld screening of literacy precursor abilities; DERET: German spelling test; SETK 3-5: a developmental German language test for children between 3 and 5 years of age; BAKO: Test of basic reading and spelling skills; PDE: Phonemic Decoding Efficiency; TOSWRF or TOSWRF-2: Test of Silent Word Reading Fluency, Second Edition; GORT: Grey Oral Reading Test; PC: Passage comprehension; ODEDYS: dyslexia screening tool; DKEFS: Delis-Kaplan Executive Function System; BALE: Analytic Battery of Written Language; DRT-3: Spelling percentile; 3DM: differential diagnostics for dyslexia; HRT: Heidelberger Rechentest; WRT: Weingartener Grundwortschatz Rechtschreibtest; PROLEC: Text Comprehension task; WRAT: Wide Range Achievement Test; TDE: test for School Achievement; DASH: Detailed Assessment of Speed of Handwriting; FLI: Fundamental Literacy Index; WIST: Word Identification and Spelling Test; ELFE: standardized achievement tests; TEMA: Test di Memoria e Apprendimento; BVDDE: Battery for the Assessment of Developmental Reading and Spelling Disabilities; and WM: working memory.