OBJECTIVE: To utilize Spinal Cord Toolbox (SCT) to measure the morphological parameters of healthy cervical spinal cords and explore the impact of gender, age and vertebral levels on them. METHODS: A retrospective study was conducted, collecting cervical spinal cord magnetic resonance images of 184 healthy adults. Then, cross-sectional area (CSA), eccentricity, antero-posterior diameter (AP) and right-left diameter (RL) were each computed for every subject. Then, impacts of gender, age and vertebral level on CSA and eccentricity were explored. Moreover, linear correlation analysis was conducted among CSA, AP and RL. RESULTS: The CSA of the cervical spinal cord in males was significantly larger than that in females at C2~C7 segments (P＜0.05) . For C2~C6 segments, there was no statistically significant difference in CSA between the young and middle-aged groups, but both were significantly larger than that of the elderly group. CSA of C4 was the largest whereas CSA of C7 was the smallest. Eccentricity values of C4 and C5 were significantly larger than those of other vertebral levels (P＜0.05). Linear correlation was established between CSA, AP and RL at C2~C7 segments. CONCLUSION The results allow for minimizing inter-subject variability using normalization, and thereby highlighting the importance of morphological parameters as biomarkers in the research of cervical spinal lesions.
Humans ; Female ; Male ; Retrospective Studies ; Middle Aged ; Adult ; Cervical Cord/anatomy & histology* ; Magnetic Resonance Imaging ; Young Adult ; Aged ; Cervical Vertebrae/anatomy & histology*

Objective To investigate the impact of Toxoplasma gondii type I, II and III rhoptry protein 16 (ROP16) on programmed cell death ligand 1 (PD-L1) expression in lung adenocarcinoma cells, and to examine the effects of T. gondii type I ROP16 protein on the relative PD-L1 expression, the relative PD-L1 distribution on the cell membrane surface, and the binding of programmed cell death 1 (PD-1) to PD-L1 in lung adenocarcinoma cells. Methods Lentiviral vectors overexpressing T. gondii type I, II and III ROP16 proteins were generated, and transfected into the human lung adenocarcinoma A549 cell line. A549 cells were used as a blank control group, and A549 cells transfected with an empty lentiviral expression vector were used as a negative control group, while A549 cells transfected with lentiviral vectors overexpressing T. gondii type I, II and III ROP16 proteins served as experimental groups. Stably transfected cells were selected with puromycin and verified using Western blotting, quantitative real-time PCR (RT-qPCR), and immunofluorescence assays. The PD-L1 expression was quantified at translational and transcriptional levels using Western blotting and RT-qPCR assays in A549 cells in the five groups, and the relative PD-L1 distribution was detected on the A549 cell membrane surface using flow cytometry. In addition, the effect of T. gondii type I ROP16 protein on the PD-1/PD-L1 binding was measured in A549 cells using enzyme-linked immunosorbent assay (ELISA). Results The relative ROP16 protein expression was 0, 0, 1.546 ± 0.091, 1.822 ± 0.047 and 2.334 ± 0.089 in the blank control group, negative control group, and the T. gondii type I, II and III ROP16 protein overexpression groups (F = 1 339.00,P < 0.001), and the relative ROP16 mRNA expression was 2.153 ± 0.949, 2.436 ± 1.614, 14.343 ± 0.020, 12.577 ± 0.285 and 15.090 ± 0.420 in the blank control group, negative control group and the T. gondii type I, II and III ROP16 protein overexpression groups, respectively (F = 483.50,P < 0.001). The ROP16 expression was higher in the T. gondii type I, II and III ROP16 protein overexpression groups than in the blank control group at both translational and transcriptional levels (allP values < 0.001). Immunofluorescence assay revealed that T. gondii type I, II and III ROP16 proteins were predominantly localized in A549 cell nuclei. Western blotting showed that the relative PD-L1 protein expression was 0.685 ± 0.109, 0.589 ± 0.114, 1.007 ± 0.117, 0.572 ± 0.151, and 0.426 ± 0.116 in the blank control group, negative control group, and the T. gondii type I, II and III ROP16 protein overexpression groups (F = 9.46,P < 0.05), and RT-qPCR assay quantified that the relative PD-L1 mRNA expression was 1.012 ± 0.190, 1.281 ± 0.465, 1.950 ± 0.175, 0.889 ± 0.251, and 0.230 ± 0.192 in the blank control group, negative control group, and the T. gondii type I, II and III ROP16 protein overexpression groups (F = 14.18,P < 0.05). The PD-L1 expression was higher in the T. gondii type IROP16 protein overexpression group than in the blank control group at both translational and transcriptional levels (both P values < 0.05). Flow cytometry detected that the relative distributions of PD-L1 protein were (10.83 ± 0.60)%, (11.23 ± 0.20)%, and (14.61 ± 0.50)% on the A549 cell membrane surface (F = 28.31, P < 0.05), and the relative distribution of PD-L1 protein was higher in the T. gondii type IROP16 protein overexpression group than in the blank control group and negative control group (both P values < 0.001). ELISA measured significant differences in the absorbance (A) value among the T. gondii type IROP16 protein overexpression group, the blank control group and the negative control group if the concentrations of the recombinant PD-1 protein were 0.04 (F = 10.45, P < 0.05), 0.08 μg/mL (F = 11.68, P < 0.05) and 0.12 μg/mL (F = 52.68, P < 0.05), and the A value was higher in the T. gondii type IROP16 protein overexpression group than in the blank control group and the negative control group (both P values < 0.05), indicating that T. gondii type IROP16 protein promoted the PD-L1/PD-1 binding in A549 cells in a concentration-dose manner. Conclusions T. gondii type IROP16 protein overexpression may up-regulate PD-L1 expression in A549 cells at both transcriptional and translational levels and the relative PD-L1 distribution on the A549 cell membrane surface, and affect the PD-1/PD-L1 binding in a concentration-dependent manner.

AIM: To evaluate whether Wnt pathway-related genes previously implicated in high myopia(HM)could serve as candidate genes for HM in the Chinese Han population, and to identify risk loci associated with HM susceptibility.METHODS: A case-control association analysis was conducted, involving 530 HM patients(HM group)and 1 087 healthy controls. The test efficacy was estimated using Quanto software. Peripheral blood DNA was extracted using the magnetic bead method, and seven candidate single nucleotide polymorphisms(SNPs)were genotyped using the Sequenom MassARRAY system, including HIVEP3 rs17365632, rs35134694, rs11210537, CTNNB1 rs13072632, CAMK2N1 rs10753502, TCF4 rs41396445 and Wnt7B rs73175083. Differences in allele and genotype frequencies between the HM and healthy control groups were compared under different inheritance models. Haplotype analysis was performed using SHEsis plus.RESULTS: All 7 SNPs had a genotyping detection rate exceeding 90%, and were in Hardy-Weinberg equilibrium(P>0.05). The test efficacy of the sample size was above 90.13%, indicating that the samples were representative of the population. In the HM group, the A allele frequency of HIVEP3 rs11210537 was significantly reduced(Pc=0.003, OR=0.889). Conversely, the G allele frequency was significantly elevated(Pc=0.003, OR=1.176). In an additive genetic model(AA vs GG), the AA genotype frequency was significantly lower than the GG genotype frequency(Pc=0.003, OR=0.583). Additionally, the frequency of the CCA haplotype of rs17365632, rs35134694, and rs11210537 in HIVEP3 was decreased in the HM group compared to the control group(Pc=0.008, OR=0.791).CONCLUSION: The SNP locus rs11210537 in the HIVEP3 gene is associated with genetic susceptibility to HM in the Chinese Han population, with the G allele identified as risk genetic markers. The CCA haplotype of rs17365632, rs35134694, and rs11210537 in the HIVEP3 gene represents a protection haplotype for HM.

Purpose: The genomic characteristics of uterine sarcomas have not been fully elucidated. This study aimed to explore the genomic landscape of the uterine sarcomas (USs). Materials and Methods: Comprehensive genomic analysis through RNA-sequencing was conducted. Gene fusion, differentially expressed genes (DEGs), signaling pathway enrichment, immune cell infiltration, and prognosis were analyzed. A deep learning model was constructed to predict the survival of US patients. Results: A total of 71 US samples were examined, including 47 endometrial stromal sarcomas (ESS), 18 uterine leiomyosarcomas (uLMS), three adenosarcomas, two carcinosarcomas, and one uterine tumor resembling an ovarian sex-cord tumor. ESS (including high-grade ESS [HGESS] and low-grade ESS [LGESS]) and uLMS showed distinct gene fusion signatures; a novel gene fusion site, MRPS18A–PDC-AS1 could be a potential diagnostic marker for the pathology differential diagnosis of uLMS and ESS; 797 and 477 uterine sarcoma DEGs (uDEGs) were identified in the ESS vs. uLMS and HGESS vs. LGESS groups, respectively. The uDEGs were enriched in multiple pathways. Fifteen genes including LAMB4 were confirmed with prognostic value in USs; immune infiltration analysis revealed the prognositic value of myeloid dendritic cells, plasmacytoid dendritic cells, natural killer cells, macrophage M1, monocytes and hematopoietic stem cells in USs; the deep learning model named Max-Mean Non-Local multi-instance learning (MMN-MIL) showed satisfactory performance in predicting the survival of US patients, with the area under the receiver operating curve curve reached 0.909 and accuracy achieved 0.804. Conclusion USs harbored distinct gene fusion characteristics and gene expression features between HGESS, LGESS, and uLMS. The MMN-MIL model could effectively predict the survival of US patients.

Purpose: The genomic characteristics of uterine sarcomas have not been fully elucidated. This study aimed to explore the genomic landscape of the uterine sarcomas (USs). Materials and Methods: Comprehensive genomic analysis through RNA-sequencing was conducted. Gene fusion, differentially expressed genes (DEGs), signaling pathway enrichment, immune cell infiltration, and prognosis were analyzed. A deep learning model was constructed to predict the survival of US patients. Results: A total of 71 US samples were examined, including 47 endometrial stromal sarcomas (ESS), 18 uterine leiomyosarcomas (uLMS), three adenosarcomas, two carcinosarcomas, and one uterine tumor resembling an ovarian sex-cord tumor. ESS (including high-grade ESS [HGESS] and low-grade ESS [LGESS]) and uLMS showed distinct gene fusion signatures; a novel gene fusion site, MRPS18A–PDC-AS1 could be a potential diagnostic marker for the pathology differential diagnosis of uLMS and ESS; 797 and 477 uterine sarcoma DEGs (uDEGs) were identified in the ESS vs. uLMS and HGESS vs. LGESS groups, respectively. The uDEGs were enriched in multiple pathways. Fifteen genes including LAMB4 were confirmed with prognostic value in USs; immune infiltration analysis revealed the prognositic value of myeloid dendritic cells, plasmacytoid dendritic cells, natural killer cells, macrophage M1, monocytes and hematopoietic stem cells in USs; the deep learning model named Max-Mean Non-Local multi-instance learning (MMN-MIL) showed satisfactory performance in predicting the survival of US patients, with the area under the receiver operating curve curve reached 0.909 and accuracy achieved 0.804. Conclusion USs harbored distinct gene fusion characteristics and gene expression features between HGESS, LGESS, and uLMS. The MMN-MIL model could effectively predict the survival of US patients.

Objective:To observe the imaging features of extramacular retinoschisis (EMRS) and paravascular abnormalities (PVA) in myopic patients, and preliminary analyze the differences in age, best corrected visual acuity (BCVA), spherical equivalent (SE), axial length (AL), and subfoveal choroidal thickness (SFCT).Methods:A cross-sectional clinical study. A total of 60 myopia patients with EMRS who were admitted to Department of Ophthalmology of The First Affiliated Hospital of Zhengzhou University from January 2023 to June 2024 were included in the study. There were 18 male cases with 18 eyes and 42 female cases with 42 eyes. Age was (37.57±17.14) years; SE was (-10.76±4.66) D; AL was (28.36±1.87) mm. According to the characteristics of ultra-wide-angle optical coherence tomography images, PVA was divided into perivascular cysts (PC), perivascular microfolds (PM) and perivascular lamellar holes (PLH). According to the splitting level, EMRS can be divided into inner layer, middle layer and outer layer. According to SE, the affected eyes were divided into low myopia group, moderate myopia group and high myopia group. The occurrence of EMRS near optic disc, supratemporal, suprasal and subnasal, as well as the clinical characteristics of patients with EMRS at different locations, levels and forms of PVA were observed. Age, BCVA, SE, AL and SFCT of EMRS patients at different locations and levels were compared by independent sample t test. χ2 test or Fisher exact probability test were used to compare the categorical variables between groups. Results:In 60 eyes, EMRS were located in supratemporal, infratemporal, supranasal, subnasal, and paratopic discs in 36, 43, 15, 13, and 14 eyes, respectively. The EMRS in the inner and outer layers were 59 (98.3%, 59/60) and 35 (58.3%, 35/60) eyes, respectively. PVA was present in 47 eyes (78.3%, 47/60). Among them, PC, PM and PLH were 45, 39 and 18 eyes, respectively. The age of those with paratopic splitting was older than those without paratopic splitting ( t=2.720). Those with temporal splitting had worse BCVA and longer AL than those without splitting ( t=2.139, 2.119). Those with subnasal splitting had worse BCVA, higher myopia, longer AL and thinner SFCT than those without splitting. The differences were statistically significant ( t=2.926, -2.640, 2.635, -3.938; P<0.05). Compared with other types of EMRS, patients with inner EMRS had younger age ( t=-2.383), better BCVA ( t=-4.825), shorter AL ( t=-4.767), lower myopia ( t=4.791), and thicker SFCT ( t=4.791); patients with full-layer EMRS were older ( t=2.419), worse BCVA ( t=3.656), longer AL ( t=2.677), higher degree of myopia ( t=-2.755), and thinner SFCT ( t=-3.283), with statistical significance ( P<0.05). There was significant difference in SFCT among patients with or without PC ( t=-2.396, P<0.05). Compared with eyes without PM and PLH, eyes with PM had worse BCVA, longer AL, higher myopia, and thinner SFCT, and the differences were statistically significant (PM: t=2.514, 3.078, -2.811, -4.205; P<0.05; PLH: t=2.514, 2.992, -2.949, -1.773; P<0.05). Conclusions:EMRS primarily occurs in the temporal side, with the highest frequency in the inner layer. Patients with inner-layer EMRS are younger, have better BCVA, shorter AL, lower myopia, and thicker SFCT, whereas patients with full-layer EMRS exhibit the opposite characteristics.

Objective:To observe and analyze the imaging features of pathologic myopic paravascular abnormalities (PVA) and macular retinoschisis (MRS) and their relationship with clinical features.Methods:A retrospective case series study. A total of 371 eyes of 224 patients with pathological myopia with PVA and/or MRS diagnosed by examination in Department of Ophthalmology of the First Affiliated Hospital of Zhengzhou University from September 2021 to December 2023 were included in the study. There were 74 eyes in 48 males and 297 eyes in 176 females. Age were 54 (49, 61) years; equivalent spherical lens (SE) was -13.375 (-18.00,-10.00) D. Axial length (AL) was 29.84 (28.27, 31.24) mm. According to the features of ultra-wide-angle sweep source optical coherence tomography, PVA morphology was divided into blood vessels and paravascular microfolds, paravascular retinal cysts, paravascular retinal splits and paravascular lamellar macular holes. MRS was divided into inner layer, outer layer and mixed layer according to splitting level, and grouped accordingly. The presence of dome-shaped macula (DSM), internal and external lamellar macular hole, and full lamellar macular hole (FTMH) were recorded. According to whether PVA combined with MRS or not, the affected eyes were divided into PVA group and PVA combined with MRS group. According to whether MRS involved the fovea, the affected eyes were divided into two groups: MRS not involved the fovea group and MRS involved the fovea group. Mann-Whitney U test was used to compare age, SE and AL among different groups. Qualitative data were compared by χ2 test. Results:In 371 eyes, there were 120 eyes in the simple PVA group; in the MRS group, there were 251 eyes, of which 208 eyes were in the PVA combined with MRS group. There were 33, 27, 3, 14, 12, 56, 28, and 7 eyes with or without retinal detachment, choroidal neovascularization, DSM, preretinal membrane, and simple superficial detachment of neuroepithelium in inner lamina, outer lamina, and FTMH, respectively. In 328 eyes with PVA, blood vessels and paravascular microfolds, paravascular retinal cysts, paravascular retinal splits and paravascular lamellar macular holes were 151, 236, 202 and 72 eyes, respectively. There were 142 eyes with single lesion (43.29%, 142/328). There were 186 eyes with 2 or more lesions (56.71%, 186/328). There were 34 eyes in the inner MRS group, 92 eyes in the outer MRS group and 125 eyes in the mixed MRS group, respectively. It involved 155 eyes in the fovea group; 96 eyes in the fovea group were not involved. There were significant differences in the number of PVA eyes between the outer MRS group, the inner MRS group and the mixed MRS group ( χ2=30.614, 28.379; P<0.001). Compared with PVA group, PVA group combined with MRS group was more likely to have two or more PVA lesions, the difference was statistically significant ( χ2=30.535, P<0.001). Compared with the inner MRS group and mixed MRS group, the age of the simple PVA group was younger, and the age of the outer MRS group was older, and the differences were statistically significant ( P<0.05). Compared with the inner MRS group and the outer MRS group, the simple PVA group had short AL, less myopia and better optimal corrected visual acuity (BCVA), while the mixed MRS group had long AL, more myopia and worse BCVA, and the differences were statistically significant ( P<0.05). Compared with the PVA group and the inner MRS group, the fovea was more involved in the outer MRS group and the mixed MRS group, and the difference was statistically significant ( χ2=3.906, 10.836; P<0.05). Those with MRS involved in fovea were older, AL was longer, myopia was more severe, and BCVA was worse ( P<0.001). DSM-associated splits were less likely to involve macula ( P<0.001). Conclusion:PVA and MRS exhibit a variety of presentations, and their imaging features correlate with clinical features.

Objective:To compare and observe the fundus imaging characteristics of eyes with glucocorticoid-related central serous chorioretinopathy (CSC).Methods:A retrospective clinical study. A total of 149 CSC patients with 166 eyes diagnosed at Department of Ophthalmology of The First Affiliated Hospital of Zhengzhou University from March 2021 to October 2024 were included in the study. The duration of the disease from the appearance of symptoms to treatment was less than 3 months. All affected eyes underwent best-corrected visual acuity (BCVA), fundus color photography, swept-source optical coherence tomography (SS-OCT), and fundus fluorescein angiography (FFA) examinations. BCVA was tested using an international standard vision chart and converted into logarithm of the minimum angle of resolution (logMAR) visual acuity for statistical analysis. The SS-OCT instrument measured subfoveal choroidal thickness (SFCT), central macular thickness (CMT), choroidal vascular volume (CVV), and the width and height of flat irregular pigment epithelial detachment (FIPED). FIPED, subretinal fibrin, and choroidal layer strong reflective spots were identified from SS-OCTA B-scan images; multiple leakages (leak points >3) were identified from FFA images. Based on the presence or absence of a clear history of glucocorticoid administration before the onset, patients were divided into glucocorticoid-related and non-glucocorticoid-related groups, comprising 41 patients with 53 eyes and 108 patients with 113 eyes, respectively. Clinical and fundus imaging characteristics of the two groups were compared. The comparison of quantitative data between the two groups was performed using independent samples t test or non-parametric independent samples Wilcoxon test; the comparison of qualitative data was performed using χ2 test. Results:Compared with the non-glucocorticoid-related group, the glucocorticoid-related group had a smaller male-to-female ratio and a higher bilateral incidence, and the differences were statistically significant ( χ2=4.925, 17.849; P<0.05). The logMAR BCVA for the glucocorticoid-related and non-glucocorticoid-related groups were 0.45±0.33 and 0.21±0.21, respectively; SFCT were (644.43±131.91) and (507.26±121.79) μm; CMT were (389.51±233.45) and (362.59±140.85) μm; CVV were (4.44±1.07) and (3.67±0.82) mm 3; FIPED incidence were 58.49% (31/58) and 20.35% (23/113), respectively; FIPED width and height were (1 122.01±533.98) and (742.90±388.79) μm, and (99.13±92.17) and (33.01±15.99) μm; subretinal fibrin were observed in 24 (45.28%, 24/53) and 15 (13.27%, 15/113) eyes; choroidal strong reflections were found in 38 (71.70%, 38/53) and 45 (39.82%, 45/113) eyes; multiple leak points were identified in 35 (66.03%, 35/53) and 40 (35.40%, 40/113) eyes, respectively. Compared with the non-glucocorticoid-related group, the glucocorticoid-related group had worse BCVA ( Z=-4.984), thicker SFCT ( t=6.586), larger CVV ( t=5.160), higher incidence of FIPED ( χ2=23.908), and greater width and height of FIPED ( t=2.895, Z=-3.703). The glucocorticoid-related group also had a significantly increased incidence of subretinal fibrin, choroidal strong reflections, and multiple leak points, with all differences being statistically significant ( χ2=20.565, 14.663, 13.675; P<0.05); however, the comparison of CMT showed no statistically significant difference ( Z=-0.651, P>0.05). Conclusion:Compared with non-glucocorticoid-related CSC, glucocorticoid-related CSC patients have poorer vision, are more likely to affect both eyes, show no gender bias; choroidal vascular dilation is more significant, and damage to the outer retina and retinal pigment epithelium is more severe.

Objective:To investigate the relationship between single nucleotide polymorphisms (SNP) in the CHRM1 gene and genetic susceptibility to high myopia (HM) in the Han population of Henan Province. Methods:A retrospective case-control study. From January 2021 to April 2023, 576 HM patients (HM group) and 768 healthy volunteers (control group) were recruited from the Department of Ophthalmology, the First Affiliated Hospital of Zhengzhou University. All participants were of Han ethnicity from Henan Province. SNP data for the CHRM1 gene in the Northern Han Chinese population were downloaded from the 1000 Genomes Project Online Website, with screening criteria of Hardy-Weinberg equilibrium P>0.05 and minor allele frequency> 0.05. Haploview software was used to analyze HapMap genotypes, identifying 5 tagSNP: rs55885673, rs544978, rs56995061, rs1942499, and rs2075748. MassARRAY system was employed for genotyping the 5 tagSNP loci. The SHEsis online software was employed to analyze the distribution differences of genotypes and allele frequencies between the two groups. Linkage disequilibrium coefficient D' was used to evaluate the recombination events between SNP loci, and haplotypes with frequencies exceeding 3% were constructed for statistical analysis. Results:The age of the HM group was significantly lower than that of the control group ( t=18.515, P<0.05), while no significant difference was observed in gender distribution ( χ2=2.869, P=0.087). Compared with the control group, the HM group showed significantly higher frequencies of the C allele [odds ratio ( OR)=1.44, 95% confidence interval ( CI): 1.09-1.91, Pc=0.045)] and CC genotype ( OR=1.50, 95% CI: 1.11-2.02, Pc=0.038) at the rs56995061 locus, and significantly lower frequencies of the T allele ( OR=0.69, 95% CI: 0.52-0.91, Pc=0.045) and CT genotype ( OR=0.67, 95% CI: 0.49-0.91, Pc=0.045). Additionally, the CT genotype frequency at the rs2075748 locus was significantly lower in the HM group ( OR=0.66, 95% CI: 0.53-0.84, Pc=0.002). The haplotype G-T-A-A formed by rs55885673-rs56995061-rs1942499-rs544978 was significantly less frequent in the HM group ( OR=0.71, 95% CI: 0.54-0.94, P=0.170). Conclusions:The polymorphisms at the SNP loci rs56995061 and rs2075748 in the CHRM1 gene are associated with the genetic susceptibility to high myopia in the Chinese Han population. The G-T-A-A haplotype composed of rs55885673-rs56995061-rs1942499-rs544978 reduces the susceptibility to high myopia.

Purpose: The genomic characteristics of uterine sarcomas have not been fully elucidated. This study aimed to explore the genomic landscape of the uterine sarcomas (USs). Materials and Methods: Comprehensive genomic analysis through RNA-sequencing was conducted. Gene fusion, differentially expressed genes (DEGs), signaling pathway enrichment, immune cell infiltration, and prognosis were analyzed. A deep learning model was constructed to predict the survival of US patients. Results: A total of 71 US samples were examined, including 47 endometrial stromal sarcomas (ESS), 18 uterine leiomyosarcomas (uLMS), three adenosarcomas, two carcinosarcomas, and one uterine tumor resembling an ovarian sex-cord tumor. ESS (including high-grade ESS [HGESS] and low-grade ESS [LGESS]) and uLMS showed distinct gene fusion signatures; a novel gene fusion site, MRPS18A–PDC-AS1 could be a potential diagnostic marker for the pathology differential diagnosis of uLMS and ESS; 797 and 477 uterine sarcoma DEGs (uDEGs) were identified in the ESS vs. uLMS and HGESS vs. LGESS groups, respectively. The uDEGs were enriched in multiple pathways. Fifteen genes including LAMB4 were confirmed with prognostic value in USs; immune infiltration analysis revealed the prognositic value of myeloid dendritic cells, plasmacytoid dendritic cells, natural killer cells, macrophage M1, monocytes and hematopoietic stem cells in USs; the deep learning model named Max-Mean Non-Local multi-instance learning (MMN-MIL) showed satisfactory performance in predicting the survival of US patients, with the area under the receiver operating curve curve reached 0.909 and accuracy achieved 0.804. Conclusion USs harbored distinct gene fusion characteristics and gene expression features between HGESS, LGESS, and uLMS. The MMN-MIL model could effectively predict the survival of US patients.