Enhancing psychiatric rehabilitation outcomes through a multimodal multitask learning model based on BERT and TabNet: An approach for personalized treatment and improved decision-making.

Evaluating the rehabilitation status of individuals with serious mental illnesses (SMI) necessitates a comprehensive analysis of multimodal data, including unstructured text records and structured diagnostic data. However, progress in the effective assessment of rehabilitation status remains limited. Our study develops a deep learning model integrating Bidirectional Encoder Representations from Transformers (BERT) and TabNet through a late fusion strategy to enhance rehabilitation prediction, including referral risk, dangerous behaviors, self-awareness, and medication adherence, in patients with SMI. BERT processes unstructured textual data, such as doctor's notes, whereas TabNet manages structured diagnostic information. The model's interpretability function serves to assist healthcare professionals in understanding the model's predictive decisions, improving patient care. Our model exhibited excellent predictive performance for all four tasks, with an accuracy exceeding 0.78 and an area under the curve of 0.70. In addition, a series of tests proved the model's robustness, fairness, and interpretability. This study combines multimodal and multitask learning strategies into a model and applies it to rehabilitation assessment tasks, offering a promising new tool that can be seamlessly integrated with the clinical workflow to support the provision of optimized patient care.

Shanghai Community-Based Schizophrenia Cohort (SCS): a protocol for establishing a longitudinal cohort and research database of patients with schizophrenia receiving community-based mental health treatment.

INTRODUCTION: Drivers for remission, relapse and violence-related behaviour among patients with schizophrenia are the most complicated issue. METHODS AND ANALYSIS: This study aims to recruit a longitudinal cohort of patients with schizophrenia. Two suburban districts and two urban districts were randomly selected according to health service facilities, population, geographical region and socioeconomic status. Individuals (>18 years old) who received a diagnosis of schizophrenia following the International Classification of Diseases (10th edition) criteria within the past 3 years will be invited as participants. Assessments will be carried out in local community health centres. Data will be used to (1) establish a community-based schizophrenia cohort and biobank, (2) prospectively determine the course of multidimensional functional outcomes of patients with schizophrenia who are receiving community-based mental health treatment, and (3) map the trajectories of patients with schizophrenia and prospectively determine the course of multidimensional outcomes based on the differential impact of potentially modifiable moderators. ETHICS AND DISSEMINATION: The study has been reviewed and approved by the Human Research Ethics Committee of Shanghai Mental Health Center (2021-67). Results of the study will be disseminated through peer-reviewed journals. If effective, related educational materials will be released to the public.

Distribution and correlation of refractive parameters in children with different corneal curvatures in southeast China.

AIM: To analyze the distribution of refractive status in school-age children with different corneal curvatures (CC) and the correlation between CC and refractive status. METHODS: A total of 2214 school-aged children of grade 4 in Hangzhou who were screened for school myopia were included. Uncorrected distance visual acuity (UCDVA), non-cycloplegic refraction, axial length (AL), horizontal and vertical corneal curvature (K1, K2) were measured and spherical equivalent (SE), corneal curvature radius (CCR) and axial length/corneal radius of curvature ratio (AL/CR) were calculated. UCDVA<5.0 and SE≤-0.50 D were classified as school-screening myopia. According to the different CCRs, the patients were divided into the lower corneal curvature (LCC) group (CCR≥7.92) and the higher corneal curvature (HCC) group (CCR<7.92). Each group was further divided into the normal AL subgroup and the long AL subgroup. The refractive parameters were compared to identify any differences between the two groups. RESULTS: Both SE and AL were greater in the LCC group (P=0.013, P<0.001). The prevalence of myopia was 38% in the LCC group and 44% in the HCC group (P<0.001). The proportion of children without screening myopia was higher in the LCC group (62%) than in the HCC group (56%). Among these children without screening myopia, the proportion of long AL in the LCC group (24%) was significantly higher than that in the HCC group (0.012%; P<0.001). The change of SE in the LCC group was less affected by the increase of AL than that in the HCC group. CONCLUSION: School-aged children in the LCC group have a lower incidence of screening myopia and longer AL. Low CC can mask SE reduction and AL growth to some extent, and the change of AL growth change more in children with low CC than high CC. Before the onset of myopia, its growth rate is even faster than that after the onset of myopia.

Mechano-driven intervertebral bone bridging via oriented mechanical stimulus in a twist metamaterial cage: An in silico study.

Stiffer cages provide sufficient mechanical support but fail to promote bone ingrowth due to stress shielding. It remains challenging for fusion cage to satisfy both bone bridging and mechanical stability. Here we designed a fusion cage based on twist metamaterial for improved bone ingrowth, and proved its superiority to the conventional diagonal-based cage in silico. The fusion process was numerically reproduced via an injury-induced osteogenesis model and the mechano-driven bone remodeling algorithm, and the outcomes fusion effects were evaluated by the morphological features of the newly-formed bone and the biomechanical behaviors of the bone-cage composite. The twist-based cages exhibited oriented bone formation in the depth direction, in comparison to the diagonal-based cages. The axial stiffness of the bone-cage composites with twist-based cages was notably higher than that with diagonal-based cages; meanwhile, the ranges of motion of the twist-based fusion segment were lower. It was concluded that the twist metamaterial cages led to oriented bone ingrowth, superior mechanical stability of the bone-cage composite, and less detrimental impacts on the adjacent bones. More generally, metamaterials with a tunable displacement mode of struts might provide more design freedom in implant designs to offer customized mechanical stimulus for osseointegration.

Sub-MIC vancomycin enhances the antibiotic tolerance of vancomycin-intermediate Staphylococcus aureus through downregulation of protein succinylation.

Bacteria develop tolerance after transient exposure to antibiotics, and tolerance is a significant driver of resistance. The purpose of this study is to evaluate the mechanisms underlying tolerance formation in vancomycin-intermediate Staphylococcus aureus (VISA) strains. VISA strains were cultured with sub-minimum inhibitory concentrations (sub-MICs) of vancomycin. Enhanced vancomycin tolerance was observed in VISA strains with distinct genetic lineages. Western blot revealed that the VISA protein succinylation (Ksucc) levels decreased with the increase in vancomycin exposure. Importantly, Ksucc modification, vancomycin tolerance, and cell wall synthesis were simultaneously affected after deletion of SacobB, which encodes a desuccinylase in S. aureus. Several Ksucc sites were identified in MurA, and vancomycin MIC levels of murA mutant and Ksucc-simulated (MurA(K69E) and MurA(K191E)) mutants were reduced. The vancomycin MIC levels of K65-MurA(K191E) in particular decreased to 1 mg/L, converting VISA strain K65 to a vancomycin-susceptible S. aureus strain. We further demonstrated that the enzymatic activity of MurA was dependent on Ksucc modification. Our data suggested the influence of vancomycin exposure on bacterial tolerance, and protein Ksucc modification is a novel mechanism in regulating vancomycin tolerance.

SALW-Net: a lightweight convolutional neural network based on self-adjusting loss function for spine MR image segmentation.

Segmentation of intervertebral discs and vertebrae from spine magnetic resonance (MR) images is essential to aid diagnosis algorithms for lumbar disc herniation. Convolutional neural networks (CNN) are effective methods, but often require high computational costs. Designing a lightweight CNN is more suitable for medical sites lacking high-computing power devices, yet due to the unbalanced pixel distribution in spine MR images, the segmentation is often sub-optimal. To address this issue, a lightweight spine segmentation CNN based on a self-adjusting loss function, which is named SALW-Net, is proposed in this study. For SALW-Net, the self-adjusting loss function could dynamically adjust the loss weights of the two branches according to the differences in segmentation results and labels during the training; thus, the ability for learning unbalanced pixels is enhanced. Two separate datasets are used to evaluate the proposed SALW-Net. Specifically, the proposed SALW-Net has fewer parameter numbers than U-net (only 2%) but achieves higher evaluation scores than that of U-net (the average DSC score of SALW-Net is 0.8781, and that of U-net is 0.8482). In addition, the practicality validation for SALW-Net is also proceeding, including deploying the model on a lightweight device and producing an aid diagnosis algorithm based on segmentation results. This means our SALW-Net has clinical application potential for assisted diagnosis in low computational power scenarios.

OR11H1 Missense Variant Confers the Susceptibility to Vogt-Koyanagi-Harada Disease by Mediating Gadd45g Expression.

Vogt-Koyanagi-Harada (VKH) disease is a severe autoimmune disease. Herein, whole-exome sequencing (WES) study are performed on 2,573 controls and 229 VKH patients with follow-up next-generation sequencing (NGS) in a collection of 2,380 controls and 2,278 VKH patients. A rare c.188T>C (p Val63Ala) variant in the olfactory receptor 11H1 (OR11H1) gene is found to be significantly associated with VKH disease (rs71235604, Pcombined = 7.83 × 10-30 , odds ratio = 3.12). Functional study showes that OR11H1-A63 significantly increased inflammatory factors production and exacerbated barrier function damage. Further studies using RNA-sequencing find that OR11H1-A63 markedly increased growth arrest and DNA-damage-inducible gamma (GADD45G) expression. Moreover, OR11H1-A63 activates the MAPK and NF-κB pathways, and accelerates inflammatory cascades. In addition, inhibiting GADD45G alleviates inflammatory factor secretion, likely due to the regulatory effect of GADD45G on the MAPK and NF-κB pathways. Collectively, this study suggests that the OR11H1-A63 missense mutation may increase susceptibility to VKH disease in a GADD45G-dependent manner.

GlnR activated transcription of nitrogen metabolic pathway genes facilitates biofilm formation by mycobacterium abscessus.

OBJECTIVES: Nitrogen is indispensable for the synthesis of biomacromolecules. The correlation between nitrogen metabolism and Mycobacterium abscessus (M. abscessus) biofilm formation is unclear. This study constructed global nitrogen regulator gene GlnR (Mab_0744) knockout (ΔglnR) and complementation (ΔglnR::glnR) M. abscessus strains. METHODS: Global nitrogen regulator gene glnR (Mab_0744) knockout (ΔglnR) and complementation (ΔglnR::glnR) M. abscessus strains were constructed. Sauton's medium was used to culture M. abscessus pellicle biofilm. To test the antibiotic susceptibility of pellicle biofilm, clarithromycin, amikacin, cefoxitin or imipenem was added to the medium under biofilms after 14 days of incubation. RT-qPCR and ChIP-qPCR were performed to analyse the transcriptional regulatory function of GlnR. RESULTS: GlnR knockout decreased the growth rate of planktonic cells, reduced biofilm mass and wrinkle formation, and diminished the resistance of biofilms to antibiotics. However, the susceptibility of planktonic cells to antibiotics was not changed by glnR knockout. The growth rate of planktonic ΔglnR cells was accelerated by adding nitrogen sources to the medium; the addition of glutamine or sodium glutamate rescued ΔglnR biofilm morphology and resistance to amikacin, cefoxitin, clarithromycin and imipenem. GlnR bound the promoter region and activated the transcription of eight nitrogen metabolic pathway genes (i.e. glnA, amt, ansP, nirB, nirD, glnD, glnK and narK3), which are closely related to glutamine/glutamate biosynthesis and, thus, regulate biofilm formation. CONCLUSION: This study provides insights into the mechanisms of M. abscessus biofilm formation and its resistance to antibiotics.

A polyglutamic acid/tannic acid-based nano drug delivery system: Antibacterial, immunoregulation and sustained therapeutic strategies for oral ulcers.

Oral ulcers are a common inflammatory mucosal ulcer, and the moist and dynamic environment in the oral cavity makes topical pharmacological treatment of oral ulcers challenging. Herein, oral ulcer tissue adhesion nanoparticles were prepared by using esterification reaction between polyglutamic acid and tannic acid, and at the same time doxycycline hydrochloride was loaded into the nanoparticles. The obtained slow drug release effect of the drug-loaded nanoparticles reduced the toxicity of the drug, and by penetrating into the fine crevice region of the wound tissue and adhering to it, they could in-situ release the carried drug more effectively and thus have shown significant antibacterial effects. In addition, tannic acid in the system conferred adhesion, antioxidant and immune regulation activities to the nanocarriers. A rat oral ulcer model based on fluorescent labeling was established to investigate the retention of nanoparticles at the ulcer, and the results showed that the retention rate of drug-loaded nanoparticles at the ulcer was 17 times higher than that of pure drug. Due to the antibacterial and immune regulation effects of the drug-loaded nanoparticles, the healing of oral ulcer wounds was greatly accelerated. Such application of doxycycline hydrochloride loaded polyglutamic acid/tannic acid nanoparticles is a novel and effective treatment strategy for oral ulcer.

A de novo missense mutation in MPP2 confers an increased risk of Vogt-Koyanagi-Harada disease as shown by trio-based whole-exome sequencing.

Vogt-Koyanagi-Harada (VKH) disease is a leading cause of blindness in young and middle-aged people. However, the etiology of VKH disease remains unclear. Here, we performed the first trio-based whole-exome sequencing study, which enrolled 25 VKH patients and 50 controls, followed by a study of 2081 VKH patients from a Han Chinese population to uncover detrimental mutations. A total of 15 de novo mutations in VKH patients were identified, with one of the most important being the membrane palmitoylated protein 2 (MPP2) p.K315N (MPP2-N315) mutation. The MPP2-N315 mutation was highly deleterious according to bioinformatic predictions. Additionally, this mutation appears rare, being absent from the 1000 Genome Project and Genome Aggregation Database, and it is highly conserved in 10 species, including humans and mice. Subsequent studies showed that pathological phenotypes and retinal vascular leakage were aggravated in MPP2-N315 mutation knock-in or MPP2-N315 adeno-associated virus-treated mice with experimental autoimmune uveitis (EAU). In vitro, we used clustered regularly interspaced short palindromic repeats (CRISPR‒Cas9) gene editing technology to delete intrinsic MPP2 before overexpressing wild-type MPP2 or MPP2-N315. Levels of cytokines, such as IL-1ß, IL-17E, and vascular endothelial growth factor A, were increased, and barrier function was destroyed in the MPP2-N315 mutant ARPE19 cells. Mechanistically, the MPP2-N315 mutation had a stronger ability to directly bind to ANXA2 than MPP2-K315, as shown by LC‒MS/MS and Co-IP, and resulted in activation of the ERK3/IL-17E pathway. Overall, our results demonstrated that the MPP2-K315N mutation may increase susceptibility to VKH disease.

Cascaded 2D Micromirror with Application to LiDAR.

This paper introduced a novel approach to enhance the vertical scanning angle of a large aperture 2D electromagnetic micromirror through the utilization of a cascaded torsional beam design. The primary objective was to increase the vertical scanning angle without compromising the robustness, which was achieved by optimizing the trade-off between the rotation angle and the first mode of resonant frequency. The cascaded design provides flexibility to either increase the outer frame's rotation angle without sacrificing torsional stiffness or enhance the torsion beam's stiffness while maintaining the same rotation angle, thus elevating the first-mode resonant frequency and overall robustness. The effectiveness of the cascaded design was demonstrated through a comparative study with a non-cascaded 2D micromirror possessing the same aperture size, torque, and mass moment of inertia. Theoretical analysis and finite-element simulation are employed to determine critical parameters such as the stiffness ratio between the cascaded torsion beams, and to predict improvements in the scanning angle and primary resonant frequency brought by the cascaded design. Prototypes of both cascaded and non-cascaded designs are fabricated using a flexible printed circuit board combined with Computer numerical control (CNC) machining of a Ti-alloy thin film, confirming the superior performance of the cascaded 2D micromirror. The cascaded design achieved vertical scanning angles up to 26% higher than the traditional design when both were actuated at close resonance frequencies. Additionally, the micromirror was successfully integrated into a 3D LiDAR system. The light detection and ranging (LiDAR) system was modelled in Zemax OpticStudio to find the optimized design and assembly positions.

Global lactylome reveals lactylation-dependent mechanisms underlying TH17 differentiation in experimental autoimmune uveitis.

Dysregulation of CD4+ T cell differentiation is linked to autoimmune diseases. Metabolic reprogramming from oxidative phosphorylation to glycolysis and accumulation of lactate are involved in this process. However, the underlying mechanisms remain unclear. Our study showed that lactate-derived lactylation regulated CD4+ T cell differentiation. Lactylation levels in CD4+ T cells increased with the progression of experimental autoimmune uveitis (EAU). Inhibition of lactylation suppressed TH17 differentiation and attenuated EAU inflammation. The global lactylome revealed the landscape of lactylated sites and proteins in the CD4+ T cells of normal and EAU mice. Specifically, hyperlactylation of Ikzf1 at Lys164 promoted TH17 differentiation by directly modulating the expression of TH17-related genes, including Runx1, Tlr4, interleukin-2 (IL-2), and IL-4. Delactylation of Ikzf1 at Lys164 impaired TH17 differentiation. These findings exemplify how glycolysis regulates the site specificity of protein lactylation to promote TH17 differentiation and implicate Ikzf1 lactylation as a potential therapeutic target for autoimmune diseases.

Towards Building a Trustworthy Deep Learning Framework for Medical Image Analysis.

Computer vision and deep learning have the potential to improve medical artificial intelligence (AI) by assisting in diagnosis, prediction, and prognosis. However, the application of deep learning to medical image analysis is challenging due to limited data availability and imbalanced data. While model performance is undoubtedly essential for medical image analysis, model trust is equally important. To address these challenges, we propose TRUDLMIA, a trustworthy deep learning framework for medical image analysis, which leverages image features learned through self-supervised learning and utilizes a novel surrogate loss function to build trustworthy models with optimal performance. The framework is validated on three benchmark data sets for detecting pneumonia, COVID-19, and melanoma, and the created models prove to be highly competitive, even outperforming those designed specifically for the tasks. Furthermore, we conduct ablation studies, cross-validation, and result visualization and demonstrate the contribution of proposed modules to both model performance (up to 21%) and model trust (up to 5%). We expect that the proposed framework will support researchers and clinicians in advancing the use of deep learning for dealing with public health crises, improving patient outcomes, increasing diagnostic accuracy, and enhancing the overall quality of healthcare delivery.

A Nature-Inspired Versatile Bio-Adhesive.

The application of most hydrogel bio-adhesives is greatly limited due to their high swelling, low underwater adhesion, and single function. Herein, a spatial multi-level physical-chemical and bio-inspired in-situ bonding strategy is proposed, to develop a multifunctional hydrogel bio-glue using polyglutamic acid (PGA), tyramine hydrochloride (TYR), and tannic acid (TA) as precursors and 4-(4,6-dimethoxytriazine-2-yl) -4-methylmorpholine hydrochloride(DMTMM) as condensation agent, which is used for tissue adhesion, hemostasis and repair. By introducing TYR and TA into the PGA chain, it is demonstrated that not only can the strong adhesion of bio-glue to the surface of various fresh tissues and wet materials be realized through the synergistic effect of spatial multi-level physical and chemical bonding, but also this glue can be endowed with the functions of anti-oxidation and hemostasis. The excellent performance of such bio-glue in the repair of the wound, liver, and cartilage is achieved, showing a great potential in clinical application for such bio-glue. This study will open up a brand-new avenue for the development of multifunctional hydrogel biological adhesive.

Using Nitroxides to Enhance Carbon Fiber Interfacial Adhesion and as an Anchor for "Graft to" Surface Modification Strategies.

Nitroxide groups covalently grafted to carbon fibers are used as anchoring sites for TEMPO-terminated polymers (poly-n-butylacrylate and polystyrene) in a "graft to" surface modification strategy. All surface-modified fibers are evaluated for their physical properties, showing that several treatments have enhanced the tensile strength and Young's modulus compared to the control fibers. Up to an 18% increase in tensile strength and 12% in Young's modulus are observed. Similarly, the evaluation of interfacial shear strength in an epoxy polymer shows improvements of up to 144% relative to the control sample. Interestingly, the polymer-grafted surfaces show smaller increases in interfacial shear strength compared to surfaces modified with a small molecule only. This counterintuitive result is attributed to the incompatibility, both chemical and physical, of the grafted polymers to the surrounding epoxy matrix. Molecular dynamics simulations of the interface suggest that the diminished increase in mechanical shear strength observed for the polymer grafted surfaces may be due to the lack of exposed chain ends, whereas the small molecule grafted interface exclusively presents chain ends to the resin interface, resulting in good improvements in mechanical properties.

FTO-mediated m6A modification alleviates autoimmune uveitis by regulating microglia phenotypes via the GPC4/TLR4/NF-κB signaling axis.

Uveitis, a vision-threatening inflammatory disease worldwide, is closely related to resident microglia. Retinal microglia are the main immune effector cells with strong plasticity, but their role in uveitis remains unclear. N6-methyladenosine (m6A) modification has been proven to be involved in the immune response. Therefore, we in this work aimed to identify the potentially crucial m6A regulators of microglia in uveitis. Through the single-cell sequencing (scRNA-seq) analysis and experimental verification, we found a significant decrease in the expression of fat mass and obesity-associated protein (FTO) in retinal microglia of uveitis mice and human microglia clone 3 (HMC3) cells with inflammation. Additionally, FTO knockdown was found to aggravate the secretion of inflammatory factors and the mobility/chemotaxis of microglia. Mechanistically, the RNA-seq data and rescue experiments showed that glypican 4 (GPC4) was the target of FTO, which regulated microglial inflammation mediated by the TLR4/NF-κB pathway. Moreover, RNA stability assays indicated that GPC4 upregulation was mainly regulated by the downregulation of the m6A "reader" YTH domain family protein 3 (YTHDF3). Finally, the FTO inhibitor FB23-2 further exacerbated experimental autoimmune uveitis (EAU) inflammation by promoting the GPC4/TLR4/NF-κB signaling axis, and this could be attenuated by the TLR4 inhibitor TAK-242. Collectively, a decreased FTO could facilitate microglial inflammation in EAU, suggesting that the restoration or activation of FTO function may be a potential therapeutic strategy for uveitis.

Neutrophil-Derived IL-6 Potentially Drives Ferroptosis Resistance in B Cells in Lupus Kidney.

Ferroptosis resistance is vital for B cell development, especially in inflammatory diseases, yet the underlying mechanism is still unclear. In this study, based on the scRNA-seq technique and flow cytometry, we discovered a proportion of neutrophils exhibited upregulated expression of the IL-6 and correlated with the expression of IL-6 receptor and SLC7A11 from B cells in lupus kidney. Moreover, we identified that in lupus kidney, neutrophils could provide IL-6 to facilitate ferroptosis resistance in B cells via SLC7A11, and inhibition of SLC7A11 could significantly enhance ferroptosis in B cells and could decrease B cell proliferation. This study helps understand the crosstalk between neutrophils and B cells in the kidney in the development of lupus.

A spatially resolved single-cell genomic atlas of the adult human breast.

The adult human breast is comprised of an intricate network of epithelial ducts and lobules that are embedded in connective and adipose tissue1-3. Although most previous studies have focused on the breast epithelial system4-6, many of the non-epithelial cell types remain understudied. Here we constructed the comprehensive Human Breast Cell Atlas (HBCA) at single-cell and spatial resolution. Our single-cell transcriptomics study profiled 714,331 cells from 126 women, and 117,346 nuclei from 20 women, identifying 12 major cell types and 58 biological cell states. These data reveal abundant perivascular, endothelial and immune cell populations, and highly diverse luminal epithelial cell states. Spatial mapping using four different technologies revealed an unexpectedly rich ecosystem of tissue-resident immune cells, as well as distinct molecular differences between ductal and lobular regions. Collectively, these data provide a reference of the adult normal breast tissue for studying mammary biology and diseases such as breast cancer.

Factors Influencing Patient and Caregiver Preferences for Antipsychotic Treatment of Schizophrenia in China: A Discrete Choice Experiment.

Purpose: This study aimed to quantify the preferences of Chinese patients with schizophrenia and their caregivers for antipsychotic treatment. Patients and Methods: Patients with schizophrenia (aged 18-35) and their caregivers were recruited via six outpatient mental health clinics in Shanghai, People's Republic of China. In a discrete choice experiment (DCE), participants chose between two hypothetical treatment scenarios that varied regarding the type of treatment, rate of hospitalization, severity of positive symptoms, treatment cost and rates of improvement in daily and social functioning. Data for each group were analyzed using the modelling approach that yielded the lower deviance information criterion. The relative importance score (RIS) for each treatment attribute was also determined. Results: A total of 162 patients and 167 caregivers participated. Frequency of hospital admission was the most important treatment attribute for patients (average scaled RIS=27%), followed by mode and frequency of treatment administration (24%). Improvement in ability to carry out daily activities (8%) and improvement in social functioning (8%) were least important. Patients in full-time employment placed more importance on the frequency of hospital admission than unemployed patients (p<0.01). Frequency of hospital admission was also the most important attribute for caregivers (RIS=33%), followed by improvement in positive symptoms (20%), while improvement in daily activities (7%) was the least important. Conclusion: Patients with Schizophrenia in China prefer treatments that help reduce the number of times they are admitted to hospital, as do their caregivers. These results may bring insight for physicians and health authorities in China regarding the treatment characteristics that patients value the most.

A Lightweight Convolutional Neural Network Based on Dynamic Level-Set Loss Function for Spine MR Image Segmentation.

BACKGROUND: Spine MR image segmentation is important foundation for computer-aided diagnostic (CAD) algorithms of spine disorders. Convolutional neural networks segment effectively, but require high computational costs. PURPOSE: To design a lightweight model based on dynamic level-set loss function for high segmentation performance. STUDY TYPE: Retrospective. POPULATION: Four hundred forty-eight subjects (3163 images) from two separate datasets. Dataset-1: 276 subjects/994 images (53.26% female, mean age 49.02 ± 14.09), all for disc degeneration screening, 188 had disc degeneration, 67 had herniated disc. Dataset-2: public dataset with 172 subjects/2169 images, 142 patients with vertebral degeneration, 163 patients with disc degeneration. FIELD STRENGTH/SEQUENCE: T2 weighted turbo spin echo sequences at 3T. ASSESSMENT: Dynamic Level-set Net (DLS-Net) was compared with four mainstream (including U-net++) and four lightweight models, and manual label made by five radiologists (vertebrae, discs, spinal fluid) used as segmentation evaluation standard. Five-fold cross-validation are used for all experiments. Based on segmentation, a CAD algorithm of lumbar disc was designed for assessing DLS-Net's practicality, and the text annotation (normal, bulging, or herniated) from medical history data were used as evaluation standard. STATISTICAL TESTS: All segmentation models were evaluated with DSC, accuracy, precision, and AUC. The pixel numbers of segmented results were compared with manual label using paired t-tests, with P < 0.05 indicating significance. The CAD algorithm was evaluated with accuracy of lumbar disc diagnosis. RESULTS: With only 1.48% parameters of U-net++, DLS-Net achieved similar accuracy in both datasets (Dataset-1: DSC 0.88 vs. 0.89, AUC 0.94 vs. 0.94; Dataset-2: DSC 0.86 vs. 0.86, AUC 0.93 vs. 0.93). The segmentation results of DLS-Net showed no significant differences with manual labels in pixel numbers for discs (Dataset-1: 1603.30 vs. 1588.77, P = 0.22; Dataset-2: 863.61 vs. 886.4, P = 0.14) and vertebrae (Dataset-1: 3984.28 vs. 3961.94, P = 0.38; Dataset-2: 4806.91 vs. 4732.85, P = 0.21). Based on DLS-Net's segmentation results, the CAD algorithm achieved higher accuracy than using non-cropped MR images (87.47% vs. 61.82%). DATA CONCLUSION: The proposed DLS-Net has fewer parameters but achieves similar accuracy to U-net++, helps CAD algorithm achieve higher accuracy, which facilitates wider application. EVIDENCE LEVEL: 2 TECHNICAL EFFICACY: Stage 1.