PGKD-Net: Prior-guided and Knowledge Diffusive Network for Choroid Segmentation.

The thickness of the choroid is considered to be an important indicator of clinical diagnosis. Therefore, accurate choroid segmentation in retinal OCT images is crucial for monitoring various ophthalmic diseases. However, this is still challenging due to the blurry boundaries and interference from other lesions. To address these issues, we propose a novel prior-guided and knowledge diffusive network (PGKD-Net) to fully utilize retinal structural information to highlight choroidal region features and boost segmentation performance. Specifically, it is composed of two parts: a Prior-mask Guided Network (PG-Net) for coarse segmentation and a Knowledge Diffusive Network (KD-Net) for fine segmentation. In addition, we design two novel feature enhancement modules, Multi-Scale Context Aggregation (MSCA) and Multi-Level Feature Fusion (MLFF). The MSCA module captures the long-distance dependencies between features from different receptive fields and improves the model's ability to learn global context. The MLFF module integrates the cascaded context knowledge learned from PG-Net to benefit fine-level segmentation. Comprehensive experiments are conducted to evaluate the performance of the proposed PGKD-Net. Experimental results show that our proposed method achieves superior segmentation accuracy over other state-of-the-art methods. Our code is made up publicly available at: https://github.com/yzh-hdu/choroid-segmentation.

Construction and implementation of wide range parameter switchable chaotic system.

Research on switchable chaotic systems with a large range of parameters is scarce. To explore the chaotic characteristics of such systems, this paper proposes new switchable methods by modifying the nonlinear term in the system, resulting in a chaotic system with different nonlinear terms. The unknown parameters in the nonlinear term exhibit different numerical relationships under various combined conditions, and some parameters may tend towards positive infinity. The chaos characteristics are verified by applying a specific switching method to the unified chaotic system. The pseudo-randomness of the random sequence generated by the dissipative system is verified using the NIST test. Finally, the circuit simulation of the system under various switching conditions is performed by selecting different circuit components and adjusting the resistance values.The switching chaotic system is implemented physically on FPGA and breadboard, and the effectiveness of the system is verified.

Multi-scale consistent self-training network for semi-supervised orbital tumor segmentation.

PURPOSE: Segmentation of orbital tumors in CT images is of great significance for orbital tumor diagnosis, which is one of the most prevalent diseases of the eye. However, the large variety of tumor sizes and shapes makes the segmentation task very challenging, especially when the available annotation data is limited. METHODS: To this end, in this paper, we propose a multi-scale consistent self-training network (MSCINet) for semi-supervised orbital tumor segmentation. Specifically, we exploit the semantic-invariance features by enforcing the consistency between the predictions of different scales of the same image to make the model more robust to size variation. Moreover, we incorporate a new self-training strategy, which adopts iterative training with an uncertainty filtering mechanism to filter the pseudo-labels generated by the model, to eliminate the accumulation of pseudo-label error predictions and increase the generalization of the model. RESULTS: For evaluation, we have built two datasets, the orbital tumor binary segmentation dataset (Orbtum-B) and the orbital multi-organ segmentation dataset (Orbtum-M). Experimental results on these two datasets show that our proposed method can both achieve state-of-the-art performance. In our datasets, there are a total of 55 patients containing 602 2D images. CONCLUSION: In this paper, we develop a new semi-supervised segmentation method for orbital tumors, which is designed for the characteristics of orbital tumors and exhibits excellent performance compared to previous semi-supervised algorithms.

Clinical evaluation of deep learning systems for assisting in the diagnosis of the epiretinal membrane grade in general ophthalmologists.

BACKGROUND: Epiretinal membrane (ERM) is a common age-related retinal disease detected by optical coherence tomography (OCT), with a prevalence of 34.1% among people over 60 years old. This study aims to develop artificial intelligence (AI) systems to assist in the diagnosis of ERM grade using OCT images and to clinically evaluate the potential benefits and risks of our AI systems with a comparative experiment. METHODS: A segmentation deep learning (DL) model that segments retinal features associated with ERM severity and a classification DL model that grades the severity of ERM were developed based on an OCT dataset obtained from three hospitals. A comparative experiment was conducted to compare the performance of four general ophthalmologists with and without assistance from the AI in diagnosing ERM severity. RESULTS: The segmentation network had a pixel accuracy (PA) of 0.980 and a mean intersection over union (MIoU) of 0.873, while the six-classification network had a total accuracy of 81.3%. The diagnostic accuracy scores of the four ophthalmologists increased with AI assistance from 81.7%, 80.7%, 78.0%, and 80.7% to 87.7%, 86.7%, 89.0%, and 91.3%, respectively, while the corresponding time expenditures were reduced. The specific results of the study as well as the misinterpretations of the AI systems were analysed. CONCLUSION: Through our comparative experiment, the AI systems proved to be valuable references for medical diagnosis and demonstrated the potential to accelerate clinical workflows. Systematic efforts are needed to ensure the safe and rapid integration of AI systems into ophthalmic practice.

Loss of Biliverdin Reductase Increases Oxidative Stress in the Cyanobacterium Synechococcus sp. PCC 7002.

Oxygenic photosynthesis requires metal-rich cofactors and electron-transfer components that can produce reactive oxygen species (ROS) that are highly toxic to cyanobacterial cells. Biliverdin reductase (BvdR) reduces biliverdin IXα to bilirubin, which is a potent scavenger of radicals and ROS. The enzyme is widespread in mammals but is also found in many cyanobacteria. We show that a previously described bvdR mutant of Synechocystis sp. PCC 6803 contained a secondary deletion mutation in the cpcB gene. The bvdR gene from Synechococcus sp. PCC 7002 was expressed in Escherichia coli, and recombinant BvdR was purified and shown to reduce biliverdin to bilirubin. The bvdR gene was successfully inactivated in Synechococcus sp. PCC 7002, a strain that is naturally much more tolerant of high light and ROS than Synechocystis sp. PCC 6803. The bvdR mutant strain, BR2, had lower total phycobiliprotein and chlorophyll levels than wild-type cells. As determined using whole-cell fluorescence at 77 K, the photosystem I levels were also lower than those in wild-type cells. The BR2 mutant had significantly higher ROS levels compared to wild-type cells after exposure to high light for 30 min. Together, these results suggest that bilirubin plays an important role as a scavenger for ROS in Synechococcus sp. PCC 7002. The oxidation of bilirubin by ROS could convert bilirubin to biliverdin IXα, and thus BvdR might be important for regenerating bilirubin. These results further suggest that BvdR is a key component of a scavenging cycle by which cyanobacteria protect themselves from the toxic ROS byproducts generated during oxygenic photosynthesis.

A novel prognostic model based on ferritin and nomogram-revised risk index could better stratify patients with extranodal natural killer/T-cell lymphoma.

BACKGROUND: Extranodal natural killer (NK)/T-cell lymphoma (ENKTCL) is an aggressive lymphoma with marked heterogeneity, resulting in a distinct prognosis even in patients with the same disease stage. The nomogram-revised risk index (NRI) has been proposed to stratify patients with ENKTCL. Numerous reports have revealed the prognostic role of serum ferritin in various cancers. PURPOSE: We aimed to evaluate the role of NRI in our single cohort of patients with ENKTCL treated uniformly, explore the prognostic value of ferritin, and establish a new prognostic model to better stratify patients with ENKTCL. METHODS: We included 326 patients with ENKTCL with detailed data regarding clinical characteristics and survival outcomes. All patients were treated with asparaginase-based chemotherapy with or without radiotherapy. Multiple R packages were used to analyze the prognostic factors and derive a novel prognostic model. RESULTS: In the training cohort comprising 236 patients with ENKTCL, NRI significantly correlated with progression-free survival (PFS) and overall survival (p < 0.0001). Using a ferritin level of 400 µg/L as the cutoff value, patients with high ferritin levels had significantly inferior PFS (p = 0.00028). Integrating the NRI score and four easily accessible clinical parameters, namely ferritin, hemoglobin, albumin, and D-dimer, a new prognostic model was constructed, stratifying patients with ENKTCL into three risk groups. This new prognostic model was independent of disease stage and NRI and performed better than NRI. Furthermore, this model helped to stratify patients within the same NRI risk groups. Finally, the role of this novel prognostic model was validated in the external validation cohort comprising 90 patients with ENKTCL. CONCLUSIONS: Serum ferritin level could be a novel prognostic factor in patients with ENKTCL. The new prognostic model combining NRI and clinical parameters could better predict the prognosis of ENKTCL, thereby warranting further validation and potentially guiding individualized treatment in future prospective clinical trials.

Dose-adjusted EPOCH-R vs. R-CHOP in frontline management of Waldeyer's ring diffuse large B-cell lymphoma: a retrospective study from a single institution.

BACKGROUND: To compare the efficacy and safety of dose-adjusted etoposide, prednisone, vincristine, cyclophosphamide, and doxorubicin plus rituximab (DA-EPOCH-R) with standard rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) in Waldeyer's ring diffuse large B-cell lymphoma (WR-DLBCL) at a single institution. METHODS: This retrospective study included 115 newly diagnosed patients with WR-DLBCL, of whom 68 patients received R-CHOP, and 47 patients received DA-EPOCH-R as their first-line treatment. The baseline features of the two groups were well balanced using a 1:1 propensity score matching method, and a total of 84 cases were obtained, including respective 42 cases in the R-CHOP and DA-EPOCH-R groups, for further survival and prognosis analysis. The primary objectives included progression-free survival (PFS) and overall survival (OS). RESULTS: During a median follow-up of 45 months, there were nine (21.4%) deaths in the R-CHOP group and two (4.8%) in the DA-EPOCH-R group. Kaplan-Meier analysis showed statistically significant improvements in PFS and OS in patients with DA-EPOCH-R compared with those treated with R-CHOP (log-rank test, P  = 0.025 and P  = 0.035, respectively). The 2-year PFS and OS rates in the DA-EPOCH-R group were 90.1% (95% confidence interval [CI]: 81.4-99.8%) and 95.2% (95% CI: 89.0-100.0%), respectively, and 80.5% (95% CI: 69.3-93.6%) and 90.5% (95% CI: 52.8-99.8%) in the R-CHOP group. Patients without B symptoms and elevated lactate dehydrogenase levels had a higher PFS in the DA-EPOCH-R group, with P values of 0.038 (hazard ratio [HR]: 0.11; 95% CI: 0.01-0.88) and 0.042 (HR: 0.19; 95% CI: 0.04-0.94), respectively. There were no statistically significant differences in clinical responses and treatment-related toxicities between the two groups. CONCLUSION: Compared with patients received R-CHOP, those treated by DA-EPOCH-R had superior PFS, OS, and controlled toxicity in patients with WR-DLBCL.

Glutaraldehyde fixation promotes palladium and gold nanoparticles formation in yeast and enhances their catalytic activity in 4-nitrophenol reduction.

When observing biosynthesized metal nanoparticles in microorganisms, glutaraldehyde is commonly used as a fixative to prepare TEM ultra-thin sections. However, as a chemical reagent with aldehyde groups, its reduction potential on metal ions has yet to be studied elaborately. Herein, we explored the influences of glutaraldehyde on yeast-synthesized gold nanoparticles (AuNPs), palladium nanoparticles (PdNPs) and their catalytic performance. A modified method for ultra-thin section preparation without glutaraldehyde fixation was developed to exclude its influence on AuNPs/PdNPs observation. It was confirmed that glutaraldehyde could promote the biosynthesis of AuNPs and PdNPs extra- and intracellularly, without altering their crystal structure and chemical state. The adsorption and reduction of Au(III)/Pd(II) were attributed to the different components of the yeast cell. Specifically, the amines and carboxyl groups in proteins and polysaccharides were involved in adsorption, while the reducing sugars hydrolyzed from polysaccharides were responsible for Au(III)/Pd(II) reduction. After glutaraldehyde fixation, the catalytic activities of Au/Pd-loaded yeast in 4-nitrophenol reduction were enhanced as well. Therefore, the influence of chemical fixatives in biosynthesized metal nanoparticles should be taken into consideration in regard to SEM, TEM observation and catalytic performance.

iERM: An Interpretable Deep Learning System to Classify Epiretinal Membrane for Different Optical Coherence Tomography Devices: A Multi-Center Analysis.

Background: Epiretinal membranes (ERM) have been found to be common among individuals >50 years old. However, the severity grading assessment for ERM based on optical coherence tomography (OCT) images has remained a challenge due to lacking reliable and interpretable analysis methods. Thus, this study aimed to develop a two-stage deep learning (DL) system named iERM to provide accurate automatic grading of ERM for clinical practice. Methods: The iERM was trained based on human segmentation of key features to improve classification performance and simultaneously provide interpretability to the classification results. We developed and tested iERM using a total of 4547 OCT B-Scans of four different commercial OCT devices that were collected from nine international medical centers. Results: As per the results, the integrated network effectively improved the grading performance by 1−5.9% compared with the traditional classification DL model and achieved high accuracy scores of 82.9%, 87.0%, and 79.4% in the internal test dataset and two external test datasets, respectively. This is comparable to retinal specialists whose average accuracy scores are 87.8% and 79.4% in two external test datasets. Conclusion: This study proved to be a benchmark method to improve the performance and enhance the interpretability of the traditional DL model with the implementation of segmentation based on prior human knowledge. It may have the potential to provide precise guidance for ERM diagnosis and treatment.

High expression of AP2M1 correlates with worse prognosis by regulating immune microenvironment and drug resistance to R-CHOP in diffuse large B cell lymphoma.

BACKGROUND: First-line treatment with R-CHOP has cured 50%-60% patients of diffuse large B cell lymphoma (DLBCL), and more than one-third patients will eventually progressed to relapsed/refractory disease with dismal outcomes. Adaptor Related Protein Complex 2 Subunit Mu 1 (AP2M1) is required for the activity of a vacuolar ATPase and may also play an important role in regulating the intracellular trafficking and function of CTLA-4 protein. Herein, using both public databases and our own tumor samples, we aimed to demonstrate the prognostic role of AP2M1 and the potential tumor-promoting mechanisms in DLBCL. METHOD: Using public datasets of DLBCL from both GEO and TCGA databases, we analyzed the role of AP2M1 in mediating chemoresistance to R-CHOP and its correlation with various clinical parameters and prognosis. By using various R packages, we evaluated the role of AP2M1 on regulating tumor immune microenvironment. Moreover, tumor samples of DLBCL from Beijing TongRen Hospital were used to validate our findings by immunohistochemistry staining. RESULT: Expression of AP2M1 was significantly increased in DLBCL, which was correlated with poor prognosis and a variety of clinical indicators. On the basis of enrichment analysis, it was found that AP2M1 may be related to intracellular receptor signaling pathway. Through immune analysis and drug prediction, we found that the expression of AP2M1 affected the immune environment and drug response of DLBCL, which further revealed the important role of AP2M1 in DLBCL. By analyzing 61 patients treated uniformly with R-CHOP regimen in our center, we validated the above findings that high expression of AP2M1 correlated with inferior survival outcomes and affected sensitivity to R-CHOP treatment. CONCLUSION: Expression of AP2M1 may affect the prognosis of DLBCL patients probably by affecting the immune environment and the responses to many drugs in treating DLBCL, indicating AP2M1 as a potential therapy target in DLBCL.

Long-term survival outcomes of patients with primary ocular adnexal MALT lymphoma: A large single-center cohort study.

BACKGROUND: Primary ocular adnexal extranodal marginal zone mucosa-associated lymphoid tissue lymphoma (OAML) is a rare subtype of non-Hodgkin's lymphoma, and no consensus has been defined concerning the optimal treatment strategies. This study aims to investigate the associations of disease characteristics and different treatments with long-term outcomes of patients with localized OAML. METHODS: A large retrospective cohort study was conducted in a single-center of China, and 166 patients with newly diagnosed primary localized OAML were enrolled. Detailed data of disease characteristics at diagnosis and treatments were collected for all patients. We compared treatment response and progression-free survival (PFS) among patients with different characteristics and treatments. RESULTS: Of the 166 patients, 52 received complete resection of neoplasm, whereas 114 had residual lesion after surgery. Among the 114 patients, 61 underwent watchful waiting and 53 received further treatment including localized radiotherapy, chemotherapy, or combined radiotherapy and chemotherapy. Median follow-up was 49 months. A total of 31 patients had disease progression or relapse, including four patients with such event more than five years after initial treatment. The 5-year PFS was 73.9%, 70.6%, and 85.9%, whereas the 10-year PFS was 69.3%, 59.2%, and 79.3%, among patients with complete resection of neoplasm, patients in the watchful waiting group and patients with further treatment, respectively. Patients with further treatment had longer PFS, compared with patients in the watchful waiting group (p = 0.011). Bilateral involvement at diagnosis was associated with significantly inferior PFS (p = 0.029), whereas age, IPI score, or TNM staging were not associated with PFS. No serious adverse reaction was reported among patients with further treatment. CONCLUSIONS: Bilateral involvement was associated with poor prognosis. Among patients with residual lesions after surgery, further treatment was associated with improved survival. Patients with OAML might experience disease progression or relapse more than five years after initial treatment.

The phycoerythrobilin isomerization activity of MpeV in Synechococcus sp. WH8020 is prevented by the presence of a histidine at position 141 within its phycoerythrin-I ß-subunit substrate.

Marine Synechococcus efficiently harvest available light for photosynthesis using complex antenna systems, called phycobilisomes, composed of an allophycocyanin core surrounded by rods, which in the open ocean are always constituted of phycocyanin and two phycoerythrin (PE) types: PEI and PEII. These cyanobacteria display a wide pigment diversity primarily resulting from differences in the ratio of the two chromophores bound to PEs, the green-light absorbing phycoerythrobilin and the blue-light absorbing phycourobilin. Prior to phycobiliprotein assembly, bilin lyases post-translationally catalyze the ligation of phycoerythrobilin to conserved cysteine residues on α- or ß-subunits, whereas the closely related lyase-isomerases isomerize phycoerythrobilin to phycourobilin during the attachment reaction. MpeV was recently shown in Synechococcus sp. RS9916 to be a lyase-isomerase which doubly links phycourobilin to two cysteine residues (C50 and C61; hereafter C50, 61) on the ß-subunit of both PEI and PEII. Here we show that Synechococcus sp. WH8020, which belongs to the same pigment type as RS9916, contains MpeV that demonstrates lyase-isomerase activity on the PEII ß-subunit but only lyase activity on the PEI ß-subunit. We also demonstrate that occurrence of a histidine at position 141 of the PEI ß-subunit from WH8020, instead of a leucine in its counterpart from RS9916, prevents the isomerization activity by WH8020 MpeV, showing for the first time that both the substrate and the enzyme play a role in the isomerization reaction. We propose a structural-based mechanism for the role of H141 in blocking isomerization. More generally, the knowledge of the amino acid present at position 141 of the ß-subunits may be used to predict which phycobilin is bound at C50, 61 of both PEI and PEII from marine Synechococcus strains.

Enhanced Ultrafast Broadband Reverse Saturable Absorption in Twistacenes with Enlarged π-Conjugated Central Bridge.

Optical nonlinearities of two all-carbon twistacenes, DPyA and DPyN, with the different π-conjugated central bridges were investigated. The nonlinear absorption properties of these compounds were measured using the femtosecond Z-scan with wavelengths between 650 and 900 nm. It has been found that the nonlinear absorption originated from two-photon absorption (TPA) and TPA-induced excited state absorption (ESA), wherein DPyA demonstrates higher performance than DPyN. The TPA cross section of DPyA (4300 GM) is nearly 4.3 times larger than that of DPyN at 650 nm. Moreover, the different central structures modulate the intensity of ESA at 532 nm, and DPyA exhibits an excellent ESA at 532 nm with multi-pulse excitation. Meanwhile, the result of data fitting and quantum chemistry calculation shows that the enhancement of nonlinear absorption in DPyA is due to the extended π- conjugated bridge and improved delocalization of π-electrons. These all-carbon twistacenes could yield potential applications in optical power limiting (OPL) technology.

Quantifying the Solution Structure of Metal Nanoclusters Using Small-Angle Neutron Scattering.

Metal nanoclusters are a unique class of synthetic material, as their crystal structures can be resolved using X-ray diffraction, and their chemical formula can be precisely determinated from mass spectroscopy. However, a complete structure characterization by these two techniques is often a challenging task. Here, we utilize small-angle neutron scattering (SANS) to directly quantify the key structure parameters of a series of silver and gold nanoclusters in solution. The results not only correlate well to their crystallographic structures, but also allow the quantification of the counterions layer surrounding charged nanoclusters in solution. Furthermore, when combining with X-ray scattering, it is possible to estimate the molecular weight of both the metal core and the ligand shell of nanoclusters. This work offers an alternative characterization tool for nanoclusters without the requirement of crystallization or gas phase ionization.

Malignant clonal evolution drives multiple myeloma cellular ecological diversity and microenvironment reprogramming.

BACKGROUND: Multiple myeloma (MM) is a heterogeneous disease with different patterns of clonal evolution and a complex tumor microenvironment, representing a challenge for clinicians and pathologists to understand and dissect the contribution and impact of polyclonality on tumor progression. METHODS: In this study, we established a global cell ecological landscape of the bone marrow (BM) from MM patients, combining single-cell RNA sequencing and single-molecule long-read genome sequencing data. RESULTS: The malignant mutation event was localized to the tumor cell clusters with shared mutation of ANK1 and IFITM2 in all malignant subpopulations of all MM patients. Therefore, these two variants occur in the early stage of malignant clonal origin to mediate the malignant transformation of proplasmacytes or plasmacytes to MM cells. Tumor cell stemness index score and pseudo-sequential clonal evolution analysis can be used to divide the evolution model of MM into two clonal origins: types I and IX. Notably, clonal evolution and the tumor microenvironment showed an interactive relationship, in which the evolution process is not only selected by but also reacts to the microenvironment; thus, vesicle secretion enriches immune cells with malignant-labeled mRNA for depletion. Interestingly, microenvironmental modification exhibited significant heterogeneity among patients. CONCLUSIONS: This characterization of the malignant clonal evolution pattern of MM at the single-cell level provides a theoretical basis and scientific evidence for a personalized precision therapy strategy and further development of a potential new adjuvant strategy combining epigenetic agent and immune checkpoint blockade.

A curriculum learning-based fully automated system for quantification of the choroidal structure in highly myopic patients.

Objective.An automated tool for choroidal segmentation and quantitative analysis under pathological conditions is currently lacking, hindering the exploration of choroidal structural changes in fundus diseases. This study aims to create a fully automated deep learning system for the quantitative analysis of the choroid with pathological changes, and to apply the system in analyzing the correlation between the choroidal structure and the severity of high myopia.Approach. A total of 2590 optical coherence tomography B-scan images of 1424 eyes of 1029 patients of high myopia from 3 hospitals were collected. We developed a curriculum learning-based system, including a two-stage U-net (TSU-net) and a post-process module for segmentation of the choroid, to calculate mean choroidal thickness (MCT) and choroidal vascularity index (CVI). The output of the images was statistically analyzed to explore the associations among MCT, CVI and the clinical characteristics of the patients.Main results. The Dice coefficient and IoU measures of choroid segmentation were 0.9221 and 0.8575, respectively. In a human-machine comparison, the system performed faster and better than a senior ophthalmologist. Statistical analysis demonstrated that, MCT is correlated with age, scan region, axial length, maculopathy type, and CVI, and CVI is correlated with scan region and MCT.Significance. A fully automated choroidal structural quantification system was developed. Clinical evaluation demonstrated that severity of high myopia is closely related to MCT but shows only a low correlation with CVI, suggesting that CVI may have little applicability in eyes with large anatomical structural variations. Future quantitative analysis of choroidal structure of large samples will enable exploration of the pathogenesis of additional fundus diseases.

CD56-Negative Extranodal Natural Killer/T-Cell Lymphoma: A Retrospective Study in 443 Patients Treated by Chemotherapy With or Without Asparaginase.

Objective: Extranodal natural killer/T cell lymphoma (NKTCL) is an aggressive EBV-related lymphoma, originating from NK cells or T cells. Previous study demonstrated that CD56 negative NKTCL should be recognized as a distinct subtype. In this study, the value of CD56 in NKTCL is validated in the era of asparaginase, and genomic analysis was done to dissect the differences between CD56-negative and positive NKTCL. Methods: 443 patients with newly diagnosed NKTCL were enrolled in this retrospective study, and correlation between CD56 positivity and survival outcomes was analyzed. The gene sequencing data was downloaded (http://www.biosino.org/node/project/detail/OEP000498), and bioinformatics analysis was done to delineate the tumor microenvironment and differentially expressed genes. Results: CD56 was expressed in 337 patients (76.1%). Within a median follow-up time of 51 months, the 5-year overall survival (OS) and progression free survival (PFS) rates were 63.8% and 51.9%, respectively. For the whole cohort, patients who were CD56-positive had superior OS (5-year OS, 86.2% vs. 51.9%, p=0.019) and PFS (5-year PFS, 55.9% vs. 40.1%, p=0.016). For patients in early stage disease, CD56 positivity was associated with superior OS and PFS (p=0.008 and 0.005, respectively). In patients who received non-asparaginase-based chemotherapy, CD56-negative was associated with shorter OS and PFS (p<0.001), and in patients who received asparaginase-based chemotherapy, CD56-negative was not related to inferior OS and PFS (p=0.093 and p=0.829, respectively). The genomic analysis demonstrated that CD56 positive NKTCL probably originated from NK cells and CD56 negative NKTCL originated from T cells. CD56 positive NKTCL had significantly higher proportion of resting NK cells, activated NK cells, and activated CD8+ and CD4+ T cells in the tumor microenvironment. Conclusions: CD56 negative NKTCL differs from CD56 positive NKTCL in both the tumor microenvironment and survival outcomes, and asparaginase-based treatment may overcome the poor prognosis brought by CD56 negativity.

Automated diagnosis of age-related macular degeneration using multi-modal vertical plane feature fusion via deep learning.

PURPOSE: To develop a computer-aided diagnostic (CADx) system of age-related macular degeneration (AMD) through feature fusion between infrared reflectance (IR) and optical coherence tomography (OCT) modalities in order to explore the superiority of multi-modality CADx system and the optimal feature fusion patterns between multi-modality inputs. METHODS: This is a dual center retrospective study. We retrospectively collected 2006 pairs of IR and OCT images to develop the algorithms. Two single-modality models and three multi-modality models were constructed for the comparison of the diagnostic efficacy. The multi-modality models were designed utilizing a novel feature fusion method, namely, vertical plane feature fusion (VPFF). The results were validated using an independent external validation dataset and compared by three ophthalmologists. RESULTS: In the test set of the ZJU dataset, our best model named OCT_MAIN demonstrated diagnostic efficiency with an overall accuracy of 0.9608 and area under the curve of 0.9944 for the normal category, 0.9659 for the dry AMD category, and 0.9930 for the wet AMD category. The external validation exhibited an overall accuracy of 0.9159. Its diagnostic efficiency was comparable to that of the senior ophthalmologist. CONCLUSIONS: The VPFF method was successfully employed to develop a multi-modal intelligent diagnostic system for the AMD classification. This is a valuable complement and optimization to the existing CADx system, which reveals a wide application prospect and research potential.

End-to-end diabetic retinopathy grading based on fundus fluorescein angiography images using deep learning.

PURPOSE: To develop and validate a deep learning system for diabetic retinopathy (DR) grading based on fundus fluorescein angiography (FFA) images. METHODS: A total of 11,214 FFA images from 705 patients were collected to form the internal dataset. Three convolutional neural networks, namely VGG16, RestNet50, and DenseNet, were trained using a nine-square grid input, and heat maps were generated. Subsequently, a comparison between human graders and the algorithm was performed. Lastly, the best model was tested on two external datasets (Xian dataset and Ningbo dataset). RESULTS: VGG16 performed the best, with a maximum accuracy of 94.17%, and had an AUC of 0.972, 0.922, and 0.994 for levels 1, 2, and 3, respectively. For Xian dataset, our model reached the accuracy of 82.47% and AUC of 0.910, 0.888, and 0.976 for levels 1, 2, and 3. As for Ningbo dataset, the network performed with the accuracy of 88.89% and AUC of 0.972, 0.756, and 0.945 for levels 1, 2, and 3. CONCLUSIONS: A deep learning system for DR staging was trained based on FFA images and evaluated through human-machine comparisons as well as external dataset testing. The proposed system will help clinical practitioners to diagnose and treat DR patients, and lay a foundation for future applications of other ophthalmic or general diseases.