Factorization Vision Transformer: Modeling Long-Range Dependency With Local Window Cost.

Transformers have astounding representational power but typically consume considerable computation which is quadratic with image resolution. The prevailing Swin transformer reduces computational costs through a local window strategy. However, this strategy inevitably causes two drawbacks: 1) the local window-based self-attention (WSA) hinders global dependency modeling capability and 2) recent studies point out that local windows impair robustness. To overcome these challenges, we pursue a preferable trade-off between computational cost and performance. Accordingly, we propose a novel factorization self-attention (FaSA) mechanism that enjoys both the advantages of local window cost and long-range dependency modeling capability. By factorizing the conventional attention matrix into sparse subattention matrices, FaSA captures long-range dependencies, while aggregating mixed-grained information at a computational cost equivalent to the local WSA. Leveraging FaSA, we present the factorization vision transformer (FaViT) with a hierarchical structure. FaViT achieves high performance and robustness, with linear computational complexity concerning input image spatial resolution. Extensive experiments have shown FaViT's advanced performance in classification and downstream tasks. Furthermore, it also exhibits strong model robustness to corrupted and biased data and hence demonstrates benefits in favor of practical applications. In comparison to the baseline model Swin-T, our FaViT-B2 significantly improves classification accuracy by 1% and robustness by 7% , while reducing model parameters by 14% . Our code will soon be publicly available: at https://github.com/q2479036243/FaViT.

Symptom aggravation after withdrawal of metal chelating agent therapy in patients with Wilson's disease.

OBJECTIVE: To study the aggravation of clinical symptoms after discontinuation of metal chelating agent therapy in Wilson's disease (WD) patients, analyze the causes of aggravation, and observe the prognosis. METHODS: 40 WD patients (cerebral type 30 cases and hepatic type 10 cases) who stopped using metal chelating agent were selected, 40 WD patients with normal therapy, and 10 normal control cases were selected. All patients underwent neurological symptom evaluation using modified Young scale, Child-Pugh liver function grading, metal metabolism, and disease typing. Magnetic sensitivity imaging (SWI), diffusion tensor imaging (DTI), and magnetic resonance spectroscopy imaging (MRS) were performed. According to the imaging results, WD patients were divided into metal deposition stage, fiber damage stage, and neuron necrosis stage. All patients were treated with metal chelating agent for 6 months. RESULTS: The score of modified Young scale in drug withdrawal group was lower than that in normal treatment group before drug withdrawal (p = .032). The score of modified Young scale was higher after drug withdrawal than before (p = .011). The number of Child-Pugh B-grade patients after drug withdrawal was more than that before drug withdrawal and in normal treatment group. The proportion of patients in the stage of neuronal necrosis after drug withdrawal (25%) was higher than that before drug withdrawal (10%) (p = .025). After drug withdrawal, urine copper was significantly higher than that before drug withdrawal and in the normal treatment group (p = .032, .039). After the withdrawal group resumed metal chelating agent treatment, 34.2% of neurological symptoms worsened. CONCLUSIONS: WD patients showed neurological symptoms aggravation and increased liver injury after metal chelating agent withdrawal. Increased metal deposition and new nerve injury occurred in the brain. After re-treatment, the aggravated neurological symptoms of WD patients are difficult to reverse.

Injury factors and pathological features of toxic milk mice during different disease stages.

OBJECTIVE: To evaluate different injury factors and pathological characteristics of the brain at different disease stages in toxic milk (TX) mice, an animal model of Wilson's disease (WD). METHODS: Thirty TX mice (10 each at 3, 6 and 12 months old) and 30 age-matched C57 mice were used in this study. Corrected phase (CP) values were determined from susceptibility-weighted images. Myelin content was determined by measuring inhibition optical density values of Luxol fast blue-stained sections. Neurofilament protein 68 kDa (NF68), ß-amyloid precursor protein (ß-APP), and myelin basic protein (MBP) levels, as well as copper and iron content, in brain nuclei of the TX mouse were evaluated. Gene amplification ratios for catalase (CAT), GSH peroxidase (GSH-PX), nitric oxide synthase (NOS), and superoxide dismutase (SOD) in mouse brain were also determined. RESULTS: Compared with C57 mice, neuronal cell counts were decreased in 12-months-old TX mice (p = .011). Myelin content was decreased in the lenticular nucleus (p = .029), thalamus (p = .030), and brainstem (p = .034) of 6-months-old TX mice; decreases in the corresponding nuclei (p = .044, .037, and .032, respectively) were also found in 12-months-old TX mice. MBP values were lower in the lenticular nucleus and thalamus (p = .027 and .016, respectively) of 6-months-old TX mice and in the corresponding nuclei (p = .24 and .040) of 12-months-old TX mice. NF-68 values were lower in the lenticular nucleus and thalamus (p = .034 and .037, respectively) of 6-months-old TX mice and in the corresponding nuclei (p = .006 and .012) of 12-months-old TX mice. ß-APP values were higher in the thalamus of 6-months-old (p = .037) and 12-months-old (p = .012) TX mice. Iron content was higher in the lenticular nucleus, thalamus, and cerebellum (p = .044, .038, and .029, respectively) of 6-months-old TX mice and in the corresponding nuclei (p = .017, .024, and .029) of 12-months-old TX mice. The NOS gene amplification multiple was higher (p = .039), whereas the SOD1 gene amplification multiple was lower (p = .041) in 12-months-old TX mice. There was no correlation between metal content or oxidation index and pathological index. CONCLUSIONS: The pathological characteristics of the brains of TX mice may differ at different ages. Different pathogenic factors, including copper and iron deposition and abnormal oxidative stress, are present at different stages.

The asymmetry of neural symptoms in Wilson's disease patients detecting by diffusion tensor imaging, resting-state functional MRI, and susceptibility-weighted imaging.

Objective: To investigate the cause of the motor asymmetry in Wilson's disease (WD) patients using functional MRI. Methods: Fifty patients with WD and 20 age-matched healthy controls were enrolled. Neurological symptoms were scored using the modified Young Scale. All study subjects underwent diffusion tensor imaging (DTI), susceptibility-weighted imaging (SWI), and resting-state functional MRI (rs-fMRI) of the brain. Six regions of interest (ROI) were chosen. Fiber volumes between ROIs on DTI, corrected phase (CP) values on SWI, amplitude of low-frequency fluctuation (ALFF), and regional homogeneity (REHO) values on rs-fMRI were determined. Asymmetry index (right or left value/left or right value) was evaluated. Results: Asymmetry of rigidity, tremor, choreic movement, and gait abnormality (asymmetry index = 1.33, 1.39, 1.36, 1.40), fiber tracts between the GP and substantia nigra (SN), GP and PU, SN and thalamus (TH), SN and cerebellum, head of the caudate nucleus (CA) and SN, PU and CA, CA and TH, TH and cerebellum (asymmetry index = 1.233, 1.260, 1.269, 1.437, 1.503, 1.138, 1.145, 1.279), CP values in the TH, SN (asymmetry index = 1.327, 1.166), ALFF values, and REHO values of the TH (asymmetry index = 1.192, 1.233) were found. Positive correlation between asymmetry index of rigidity and fiber volumes between the GP and SN, SN and TH (r = .221, .133, p = .043, .036), and tremor and fiber volumes between the CA and TH (r = .045, p = .040) was found. Conclusions: The neurological symptoms of patients with WD were asymmetry. The asymmetry of fiber projections may be the main cause of motor asymmetry in patients with WD.

Diffusion tensor imaging of the extracorticospinal network in the brains of patients with Wilson disease.

OBJECTIVE: To evaluate damage to the extracorticospinal tract in Wilson disease (WD) patients using diffusion tensor imaging (DTI). METHODS: 70 patients with WD, including 50 with cerebral type and 20 with hepatic type, and 20 age-matched healthy controls were enrolled. Neurological symptoms were scored using the modified Young Scale. Patients with cerebral type WD were divided into four subgroups: those with (1) hypokinesia, (2) parkinsonism, (3) mouth and throat dystonia, and (4) psychiatric symptoms. All study subjects underwent DTI of the brain. Five regions of interest (ROIs) were chosen. Fractional anisotropy (FA) and fiber volumes between ROIs were determined, and the relationships between DTI metrics and clinical status were evaluated. RESULTS: FA values and fiber volumes between subcortical nuclei were lower in WD patients. Fiber volumes between the putamen (PU) and the globus pallidus (GP), substantia nigra (SN), and thalamus (TH); between the head of the caudate nucleus (CA) and the GP and TH; and between the TH and cerebellum were lower in group 1 than in the other groups of WD patients. Fiber volumes between the GP and the SN and TH were lower in group 2, and fiber volumes between the SN and TH were lower in group 3. DTI metrics differed between patients with the cerebral and hepatic types of WD. CONCLUSIONS: DTI can reconstruct the network of the extracorticospinal tract. Fiber projection between subcortical nuclei was abnormal in WD patients. Damage to fiber connections may correlate with neurological symptoms in WD patients.

Characterizing brain mineral deposition in patients with Wilson disease using susceptibility-weighted imaging.

AIMS: The aim of this study was to evaluate the feasibility of characterizing the brain-mineral deposition in patients with Wilson disease (WD) using susceptibility-weighted imaging (SWI). MATERIALS AND METHODS: The study enrolled 30 WD patients and 20 age-matched healthy controls. Neurological symptoms were scored using the modified Young Scale. The hepatic function indices, serum and urinary copper content, and serum iron content were determined. All study objects received the magnetic resonance imaging (MRI) and SWI test of the brain. The values of corrected phase (CP) were calculated on SWI. The relationship between CP values and the clinical status were evaluated. RESULTS: The serum iron content of WD patients was higher than the normal. The CP values of substantia nigra, caudate nucleus, and globus pallidus of WD were lower than the normal values, while the CP value of substantia nigra was the lowest. No correlations were determined between the CP values and the iron and copper parameters. There was negative correlation between the scores of dysarthria and the CP values of the globus pallidus. There was negative correlation between the scores of tremor and the CP values of caudate nucleus. Some regions, which had high signals on T2-weighted image, had low signals on SWI. CONCLUSIONS: There might be abnormal iron metabolism in patients with WD. The decreased CP values might reflect a deposition of both copper and iron. SWI may be more sensitive than the ordinary MRI. The mineral deposition may contribute to the neural symptoms.