Several tissue structures include a number of myofibers aligned along multiple spatial axes at the microscopic scale. However, the DTI cannot uncover such structural characteristics due to its failure to detect multidirectional fibers. Therefore, one of the important applications of DSI tractography is primarily focused on revealing the complex microstructure of the tissue. For instance, for an anterior slice of the lingual core in bovine tongue, DTI depicted it solely as a region with low anisotropy, whereas DSI revealed two different fiber populations with an explicit orthogonal relationship to each other [48]. Dai et al. found that DSI tractography revealed that the cingulum bundle was less mature when cat myelination was incomplete, whereas DTI tractography tended to terminate in such areas, possibly due to the existence of crossing fibers [49]. By using DSI, Schmahmann et al. identified the major features of 10 long association fiber bundles that matched the observations in the isotope material using autoradiographic histological tract tracing in the monkey brain, whereas the DTI did not observe such precise structural characteristics due to its inability to visualize the crossing fibers and details of the origins, course, and terminations of the white-matter pathways [50]. In human brain studies, Wedeen et al. used DSI to clarify the relationships of adjacency and crossing between cerebral fiber pathways in four nonhuman primate species and humans [51]. They first found that the cerebral fiber pathways formed a rectilinear three-dimensional grid continuous with the three principal axes of development, and cortico-cortical pathways formed parallel sheets of interwoven paths in the longitudinal and medial–lateral axes, in which major pathways were local condensations. Because of the limitation in DTI, the subcomponents and connectivity of the inferior fronto-occipital fasciculus (IFOF) and the superior longitudinal fasciculus (SLF) in human brain are still controversial. The DSI shows high-quality fiber tractography and fewer partial volume effects and false continuation artifacts, and thereby it has been used to reveal more complete connectivity patterns and anatomical details of the IFOF I-V subcomponents [52] and of the SLF I-III subcomponents [53], which are connected to different cortical regions. Similar DSI tractography applications have been reported in recent studies in the tractography of other white-matter pathways, such as the thalamic–prefrontal peduncle [54], pyramidal tracts [55], anterior commissure [56], and corpus callosum [57]. Collectively, these findings suggest a powerful potential of DSI in enhancing tractography for the complexed white-matter fibers.

The DTI has limited angular resolution and cannot adequately assess the cortical regions. Another application of the DSI in resolving the tissue microstructure is cortical parcellation and connectivity reconstruction. Recent studies used the DSI to segment the ventral [58] and dorsal premotor areas [59] (VPM/DPM); they found that the VPM consists of four subregions, 6v, 4, 3b, and 3a, and the DPM is divided into three areas, 6a, 6d, and 6v. These brain regions showed consistent inter-hemispheric connection but different intra-hemispheric connection patterns. Based on DSI tractography, two recent studies characterized the connections of the middle frontal gyrus (MFG) and inferior temporal gyrus (ITG) to other cortical areas, respectively. The MFG included two major connections of the superior longitudinal fasciculus (which connected the MFG to parts of the inferior parietal lobule, posterior temporal lobe, and lateral occipital cortex) and the inferior fronto-occipital fasciculus (which connected the MFG to the lingual gyrus and cuneus) [60]. The ITG is connected to five major fibers: the U-fiber, the inferior longitudinal fasciculus, the vertical occipital fasciculus, the arcuate fasciculus, and the uncinate fasciculus [61]. Furthermore, a recent study used DSI tractography to organize “pyramid-shaped crossings” of converged U-fibers, which are key anatomical structures to construct the neural network for intricate communications throughout the entire cerebrum [62]. Using the same method, another study delineated the decussating dentato-rubro-thalamic tract, in which the afferent regions were found mainly in the posterior cerebellum, and the efferent fibers were mainly projected to the contralateral frontal cortex, suggesting segregated and parallel cerebellar outputs to cerebral regions [63] (see Table 1).

With the improvement in DSI data acquisition, more attention has recently been paid to its application in clinical diseases, including attention-deficit/hyperactivity disorder (ADHD) [2,64,65], schizophrenia [66,67,68], stroke [69,70,71,72], Parkinson’s disease [73,74], hypertension [75], autism [76], epilepsy [77], and gliomas [78] (see Table 1).

The quantitative diffusion scalars of the DSI, especially the track density imaging (TDI) of the crura of fornix (FORX) and the parahippocampal radiation of the cingulum (PHCR), are sensitive enough to define the ipsilateral side for epilepsy patients, with a sensitivity of 89.5% and specificity of 100.0% for PHCR_TDI (AUC = 0.93), and a sensitivity of 95.0% and specificity of 100% for FORX_TDI (AUC = 0.95) [77]. The DSI-based quantitative anisotropy (QA) values of corticospinal tracts (CSTs) in patients with idiopathic normal pressure hydrocephalus (iNPH) were significantly lower than those in healthy controls (HCs), but no significant differences were found between iNPH patients and HCs in the DTI-based FA values, suggesting the DSI may provide more information that can improve the present understanding of the disease mechanism [79]. Another study found that the QA value was correlated with the neuronal diameter/density in the cortical layer IIIc, and its asymmetry showed an overall favorable accuracy (sensitivity = 90.9%, specificity = 89.5%, AUC = 0.96) in the diagnostic testing of hippocampal sclerosis patients [80]. The DSI parameters also showed a good performance, with an accuracy of 83%, sensitivity of 78%, and specificity of 86% in discriminating patients with mild and severe visual defects [81].

A prior study demonstrated that the DSI could be helpful for the preoperative prediction of human epidermal growth factor receptor 2 (HER2) in patients with breast cancer, with the finding that the AUC values of the DSI quantitative parameters (range from 0.67 to 0.72) were higher than that of the DTI metric apparent diffusion coefficient (AUC = 0.57) [82]. Another study found that the anatomic integrity of the pyramidal tract (PT) with DSI tractography effectively predicted the postoperative motor function after hemispherectomy; they reported that the AUC of the DSI tractography was 0.84, and the cutoff value of the PT asymmetric ratio was 11.5%, with 100% sensitivity and 75% specificity [83]. The DSI-derived GFA in the ipsilateral medial geniculate body was related to prognosis (sensitivity = 64.7%; specificity = 85.7%; and AUC = 0.80) in patients with unilateral idiopathic sudden sensorineural hearing loss, indicating the GFA value of the ipsilateral medial geniculate body may help to predict recovery outcomes [84]. Another study reconstructed a local connectome matrix from DSI data in patients with aphasia after stroke, and their findings challenged dual-stream accounts that denied a role for the arcuate fasciculus in semantic processing and ventral-stream pathways in language production and illuminated limbic contributions to both semantic and phonological processing for word production [69]. A recent DSI study demonstrated for the first time that distinct aspects of the cortical structural reserve enable basal and complex motor control after stroke. In particular, the recovery of basal motor control may be supported via an alternative route through contralesional M1 and non-crossing fibers of the contralesional CST [72]. Taken together, these findings suggest the DSI is a very potential and powerful technology in studying the mechanism, diagnosis, and progressive prediction of clinical diseases.

First Author (Ref. #)	Type	Subject	Main Findings
Lacerda et al., 2016 [46]	Methodology	Optimization of postprocessing method	This study proposed a new way of including biophysical constraints to compute the ODF, which removed most of the artifacts due to fast diffusion components like those from pathological tissues and offered improved angular resolution.
Tian et al., 2019 [47]	Methodology	Optimization of postprocessing method	This study proposed a generalized DSI framework to compute the ensemble average propagator, which could be used to elucidate the contribution and combination of q-space signals to the diffusion ODF.
Gilbert et al., 2006 [48]	Application	DSI tractography for white-matter fibers	Diffusion tensor imaging (DTI) depicted the anterior slice of the lingual core in bovine tongue solely as a region with low anisotropy, whereas DSI revealed two different fiber populations with an explicit orthogonal relationship to each other.
Dai et al., 2016 [49]	Application	DSI tractography for white-matter fibers	The cingulum bundle was less mature when cat myelination was incomplete, whereas the DTI tractography tended to terminate in such areas.
Schmahmann et al., 2007 [50]	Application	DSI tractography for white-matter fibers	This study identified 10 major long association fiber bundles that matched the observations in autoradiographic histological tract tracing in the monkey brain, and such precise structural characteristics were not observed by DTI.
Wedeen et al., 2012 [51]	Application	DSI tractography for white-matter fibers	This study first clarified the relationships of adjacency and crossing between cerebral fiber pathways in four nonhuman primate species and humans.
Wu et al., 2016 [52]; Wang et al., 2016 [53]	Application	DSI tractography for white-matter fibers	The DSI revealed a more complete connectivity pattern and anatomical details of the IFOF I-V subcomponents and of the SLF I-III subcomponents.
Sun et al., 2018 [54]; Suo et al., 2021 [55]; Liu et al., 2022 [56]; Wei et al., 2017 [57]	Application	DSI tractography for white-matter fibers	The DSI identified detailed and completed white-matter pathways, including the thalamic–prefrontal peduncle, pyramidal tracts, anterior commissure, and corpus callosum.
Sheets et al., 2020, 2021 [58,59]	Application	Cortical parcellation	The DSI segmented the ventral premotor area into four subregions of 6v, 4, 3b, and 3a and the dorsal premotor area into three areas of 6a, 6d, and 6v.
Briggs et al., 2021 [60]; Lin et al., 2020 [61]	Application	Cortical connectivity reconstruction	The MFG included two major connections of the superior longitudinal fasciculus and inferior fronto-occipital fasciculus. The ITG connected to five major fibers: the U-fiber, inferior longitudinal fasciculus, vertical occipital fasciculus, arcuate fasciculus, and uncinate fasciculus.
Chiang et al., 2020, 2023 [2,64];Tsai et al., 2021 [65]	Application	Attention deficit and hyperactivity disorder (ADHD)	Participants with ADHD showed more rapid development of generalized fractional anisotropy (GFA) in the frontal tracts and showed higher axial diffusivity values in the perpendicular fasciculus, superior longitudinal fasciculus I, corticospinal tract, and corpus callosum compared to the control group.
Wen et al., 2020 [73]; Papageorgiou et al., 2021 [74]	Application	Parkinson’s disease (PD)	The PD patients showed impaired global efficiency and characteristic path length in the DSI-based connected network, which were associated with executive function and episodic memory.
Wang et al., 2020, 2022 [77,80]; Zhang et al., 2023 [82]	Application	Epilepsy	The AUC of the asymmetric indices of the DSI-derived QA value to the lateralization of epilepsy was 0.96, with 0.91 sensitivity and 0.90 specificity; The AUC of DSI tractography was 0.84, with 100% sensitivity and 75% specificity in discriminating patients with epilepsy from healthy controls.
Ni et al., 2020 [76]	Application	Autism spectrum disorder (ASD)	A higher GFA of the tracts was implicated in memory, attention, sensorimotor processing, and perception associated with less dysregulation in TDC but worse dysregulation in ASD.
Zhang et al., 2021 [79]	Application	Idiopathic normal-pressure hydrocephalus (iNPH)	The DSI-based QA values of corticospinal tracts (CSTs) in patients with Inph were lower than those in healthy controls (HCs), but such differences in DTI-based FA were observed between iNPH patients and HCs.
Liang et al., 2021 [81]	Application	Pituitary adenomas	The DSI parameters also showed a good performance, with an accuracy of 0.83, sensitivity of 0.78, and specificity of 0.86 in discriminating patients with mild and severe visual defects
Mao et al., 2022 [82]	Application	Breast cancer	DSI could be helpful for the preoperative prediction of human epidermal growth factor receptor 2 (HER2) in patients with breast cancer, with the findings that the AUC values of DSI quantitative parameters (0.67~0.72) were higher than those of apparent diffusion coefficient (0.57) from DTI.
Zhang et al., 2021 [84]	Application	Idiopathic sudden sensorineural hearing loss	The DSI-derived GFA in the ipsilateral medial geniculate body was related to the prognosis (sensitivity = 64.7%; specificity = 85.7%; AUC = 0.796) in patients with unilateral idiopathic sudden sensorineural hearing loss.
Paul et al., 2023 [72]	Application	Stroke	This study used DSI to demonstrate for the first time that recovery of basal motor control may be supported via an alternative route through contralesional M1 and non-crossing fibers of the contralesional CST.
Salisbury et al., 2023 [68]	Application	First-episode psychosis	White-matter tracts showing associations between QA from DSI and auditory hallucinations were associated with frontal–parietal–temporal connectivity in the cingulum bundle and in the prefrontal interhemispheric connectivity.