Deep Learning-Based Synthesis of Contrast-Enhanced MRI for Automated Delineation of Primary Gross Tumor Volume in Radiotherapy of Nasopharyngeal Carcinoma

PURPOSE/OBJECTIVE(S)

Contrast-enhanced MRIs are necessary to delineate the primary gross tumor volume (GTVp) in radiotherapy of nasopharyngeal carcinoma (NPC). However, using contrast agents to scan contrast-enhanced MRIs is not applicable to some patients due to metal implants or their allergy, and it increases the treatment cost of patients. To address these problems, this work aims at synthesizing contrast-enhance MRIs from unenhanced MRIs by implementing generative adversarial network (GAN).

MATERIALS/METHODS

In this work, 324 MRI datasets of patients with NPC were retrospectively collected between September 2016 and September 2017 from a single institute. MRI examinations were performed with un-enhanced T1-weighted (T1) and T2-weighted (T2) sequences, and contrast-enhanced T1-weighted (T1C) and fat-suppressed T1-weighted (T1FSC) sequences. We designed and developed a modified pix2pix network to synthesize T1C (sT1C) and T1FSC (sT1FSC) from real T1. The end of the generator in this network was assembled with multiple heads (the classification head and gradient head) to learn more representation information and features from real images, the discriminator in this network distinguished whether the synthesized image is real and fake and supervised that the generator outputs more realistic synthesized image. We verified the performance of the synthesized images for automated delineation of GTVp. In an independent testing set of 11 patients, the synthesized sT1C and sT1FSC were inputted into the segmentation deep learning network along with their corresponding T1 and T2 sequences to generate GTVp contours. Delineation performance of the synthesized images and real images for automated delineation were evaluated by dice similarity coefficient (DSC), and average surface distance (ASD), using human expert contours as the ground truth.

RESULTS

In automated contouring of GTVp for NPC, the segmentation deep learning network using one or two synthesized MRIs showed equivalent performance when compared with the automated contours which generated from four real MRI sequences. Mean DSCs between automated contours by sT1C-replaced or sT1C and sT1FSC-replaced network and ground truth contours were 0.726 ± 0.143 and 0.711 ± 0.157, respectively, slightly inferior to that of contours generated from four real MRI sequences (0.740 ± 0.154, both P >0.05). In terms of mean ASD, there was also no significant difference between automated contours generated from synthesized images and real images (3.056 ± 4.216 mm and 3.537 ± 4.793 mm vs. 3.124 ± 4.637 mm; both P > 0.05).

CONCLUSION

We proposed an MRI-synthesis method based on GAN and the synthesized contrast-enhanced MRIs performed equivalent as the real contrast-enhanced MRIs in the automated delineation of gross tumor volume for radiotherapy of NPC.

A Feasible Study for Auto Planning and Auto Re-Planning for Nasopharyngeal Carcinoma (NPC) Adaptive Radiation Therapy (ART)

PURPOSE/OBJECTIVE(S)

To explore the necessity of Image Guided Radiation Therapy (IGRT) for Nasopharyngeal Carcinoma (NPC) adaptive radiation therapy (ART), and evaluate the consistency and robustness of auto re-planning during ART.

MATERIALS/METHODS

Eleven NPC patients were enrolled in this study at one institution. We used a CT-integrated linear accelerator, which integrates a 16-slice helical CT to acquire diagnostic-grade fan-beam CT (FBCT) for IGRT. Electron density accuracy from FBCT provides a solid foundation for accurate radiation dose calculation. PGTVp, PTV1 and PTV2 prescription dose are 69.96 Gy, 60.06 Gy and 54.12 Gy with 33 fractions. All ROIs, including Targets and OARs, were auto delineated via a treatment planning system (TPS), and modified by a senior physician with more than 15-year experience to confirm that they follow the clinical requirement. An initial plan (Plan A) was automatically generated based on the first CT-Sim images on the TPS. Another adjusted re-plan (Plan B) was also automatically generated based on the second CT-Sim images after 20 fractions of treatment for ART implementation. During the whole course of the 33 fractions delivery, there are 20 fractions Plan A (with 4 weekly IGRTi, i = 1∼4) and 13 fractions Plan B (with 3 weekly IGRTj, j = 5∼7). After carefully rigid registration between the CT-Sim images and their following weekly FBCT images, we copied Plan A and Plan B to IGRTi and IGRTj, respectively. Plan_IGRT would be re-calculated for dose evaluation. In addition, the Plan A was copied to the second CT-Sim (Plan A_2nd CT-Sim) after first CT-Sim and second CT-Sim rigid images registration.

RESULTS

There is a significant target volume change of -5%±4%, -3%±3%, and -5%±3% from Plan A to Plan B, for PGTVp, PTV1 and PTV2 (p<0.05), respectively. All the Plan A and Plan B could be generated within 210.2s±1.4s, which is more time-saving than manual planning greatly, and there is no statistical difference between Plan A and Plan B of the plan quality index (p>0.05). The plans for IGRT7 are inferior to the plans for IGRT5 with higher V110% for PGTVp (4.40%±8.60% for Plan A, 2.37%±8.91% for Plan B). PlanA_2nd CT-Sim for each patient is inferior to Plan B, with higher V110% for PGTVp (19.12%±18.91%), lower V100% for PTV2 (-2.84%±2.89%) and higher Dmax for Brainstem (315.88 cGy ± 190.39 cGy) statistically. Furthermore, all the Plan B_IGRTj are superior to Plan A_IGRTj, with the dose index difference of -17.50% ± 23.15%/-15.47% ± 14.85%, 2.45% ± 3.23%/2.31% ± 3.09% and -194.03 cGy ± 221.91 cGy/-170.07 cGy ± 168.41 cGy for V110% of PGTVp, V100% for PTV2 and Dmax of Brainstem for j = 5/7 (p<0.05), respectively.

CONCLUSION

The world's first integrated CT-Linac platform, equipped with FBCT, can provide a diagnostic-quality FBCT for achieve offline ART. It is necessary for NPC patients to have the IGRT, ART and re-planning after 20 fractions treatment, for the target volumes change sharply. Auto planning and auto re-planning for NPC ART are able to maintain the plan consistency and robustness while shorten the planning time.

METTL3 promotes drug resistance to oxaliplatin in gastric cancer cells through DNA repair pathway

Gastric cancer (GC) poses a significant threat to human health and remains a prevalent form of cancer. Despite clinical treatments, the prognosis for Gastric cancer patients is still unsatisfactory, largely due to the development of multidrug resistance. Oxaliplatin (OXA), a second-generation platinum drug, is commonly recommended for adjuvant and palliative chemotherapy in Gastric cancer; however, the underlying mechanisms of acquired resistance to Oxaliplatin in Gastric cancer patients are not yet fully understood. In this study, we aimed to explore the potential mechanisms of Oxaliplatin resistance in Gastric cancer by employing bioinformatics analysis and conducting in vitro experiments. Specifically, we focused on investigating the role of methyltransferase-like 3 (METTL3). Our findings revealed that the knockdown of METTL3 significantly impeded the proliferation and migration of Gastric cancer cells. METTL3 knockdown induced apoptosis in OXA-resistant Gastric cancer cells and enhanced their sensitivity to Oxaliplatin. Furthermore, we found that DNA repair pathways were significantly activated in OXA-resistant Gastric cancer cells, and METTL3 knockdown significantly inhibited DNA repair pathways. Another important finding is that METTL3 knockdown and OXA-induced Gastric cancer cell death are additive, and the targeted METTL3 can assist Oxaliplatin treatment. Collectively, our findings suggest that METTL3 knockdown can augment the sensitivity of Gastric cancer cells to Oxaliplatin by impeding DNA repair processes. Consequently, targeting METTL3 holds great promise as a viable adjuvant strategy in the treatment of Gastric cancer patients.

On fast simulation of dynamical system with neural vector enhanced numerical solver

The large-scale simulation of dynamical systems is critical in numerous scientific and engineering disciplines. However, traditional numerical solvers are limited by the choice of step sizes when estimating integration, resulting in a trade-off between accuracy and computational efficiency. To address this challenge, we introduce a deep learning-based corrector called Neural Vector (NeurVec), which can compensate for integration errors and enable larger time step sizes in simulations. Our extensive experiments on a variety of complex dynamical system benchmarks demonstrate that NeurVec exhibits remarkable generalization capability on a continuous phase space, even when trained using limited and discrete data. NeurVec significantly accelerates traditional solvers, achieving speeds tens to hundreds of times faster while maintaining high levels of accuracy and stability. Moreover, NeurVec's simple-yet-effective design, combined with its ease of implementation, has the potential to establish a new paradigm for fast-solving differential equations based on deep learning.

Clonal Propagation and Assessment of Biomass Production and Saponin Content of Elite Accessions of Wild Paris polyphylla var. yunnanensis

Paris polyphylla var. yunnanensis is an endangered medicinal plant endemic to China with great economic importance for the pharmaceutical industry. Two significant barriers to its commercial development are the long duration of its seed germination and the frequency of interspecific hybridization. We developed a method for clonal propagation of Paris polyphylla var. yunnanensis and successfully applied it to selected elite wild plants, which could become cultivar candidates based on their biomass production and saponin content. In comparison to the traditional method, somatic embryogenesis produced an average of 63 somatic embryos per gram of callus in just six weeks, saving 12 to 15 months in plantlet production. The produced in vitro plantlets were strong and healthy and 94% survived transplanting to soil. Using this method, four candidate cultivars with diverse morphologies and geographic origins were clonally reproduced from selected elite wild accessions. In comparison to those obtained with the traditional P. polyphylla propagation technique, they accumulated higher biomass and polyphyllin levels in rhizomes plus adventitious roots during a five-year period. In conclusion, somatic embryogenesis-based methods offer an alternate approach for the rapid and scaled-up production of P. polyphylla, as well as opening up species conservation options.

CIPL: Counterfactual Interactive Policy Learning to Eliminate Popularity Bias for Online Recommendation

Popularity bias, as a long-standing problem in recommender systems (RSs), has been fully considered and explored for offline recommendation systems in most existing relevant researches, but very few studies have paid attention to eliminate such bias in online interactive recommendation scenarios. Bias amplification will become increasingly serious over time due to the existence of feedback loop between the user and the interactive system. However, existing methods have only investigated the causal relations among different factors statically without considering temporal dependencies inherent in the online interactive recommendation system, making them difficult to be adapted to online settings. To address these problems, we propose a novel counterfactual interactive policy learning (CIPL) method to eliminate popularity bias for online recommendation. It first scrutinizes the causal relations in the interactive recommender models and formulates a novel temporal causal graph (TCG) to guide the training and counterfactual inference of the causal interactive recommendation system. Concretely, TCG is used to estimate the causal relations of item popularity on prediction score when the user interacts with the system at each time during model training. Besides, it is also used to remove the negative effect of popularity bias in the test stage. To train the causal interactive recommendation system, we formulated our CIPL by the actor-critic framework with an online interactive environment simulator. We conduct extensive experiments on three public benchmarks and the experimental results demonstrate that our proposed method can achieve the new state-of-the-art performance.

Postoperative pulmonary complications in older patients undergoing elective surgery with a supraglottic airway device or tracheal intubation

The two most commonly used airway management techniques during general anaesthesia are supraglottic airway devices and tracheal tubes. In older patients undergoing elective non-cardiothoracic surgery under general anaesthesia with positive pressure ventilation, we hypothesised that a composite measure of in-hospital postoperative pulmonary complications would be less frequent when a supraglottic airway device was used compared with a tracheal tube. We studied patients aged ≥ 70 years in 17 clinical centres. Patients were allocated randomly to airway management with a supraglottic airway device or a tracheal tube. Between August 2016 and April 2020, 2900 patients were studied, of whom 2751 were included in the primary analysis (1387 with supraglottic airway device and 1364 with a tracheal tube). Pre-operatively, 2431 (88.4%) patients were estimated to have a postoperative pulmonary complication risk index of 1-2. Postoperative pulmonary complications, mostly coughing, occurred in 270 of 1387 patients (19.5%) allocated to a supraglottic airway device and 342 of 1364 patients (25.1%) assigned to a tracheal tube (absolute difference -5.6% (95%CI -8.7 to -2.5), risk ratio 0.78 (95%CI 0.67-0.89); p < 0.001). Among otherwise healthy older patients undergoing elective surgery under general anaesthesia with intra-operative positive pressure ventilation of their lungs, there were fewer postoperative pulmonary complications when the airway was managed with a supraglottic airway device compared with a tracheal tube.

O, N Coordination-Mediated Nickel Single-Atom Catalysts for High-Efficiency Generation of H2O2

Electrochemical selective two-electron oxygen reduction shows great potential for on-site electrochemical production of hydrogen peroxide (H₂O₂). Herein, we demonstrated the synthesis of Ni single-atom sites coordinated by three oxygen atoms and one nitrogen atom (i.e., Ni-N₁O₃) supported by oxidized carbon black (OCB) by pyrolyzing nickel-(pyridine-2,5-dicarboxylate) coordination complexes. Aberration-corrected scanning transmission electron microscopy (AC-STEM) combined with X-ray absorption spectroscopy (XAS) proves the presence of atomically dispersed Ni atoms attached on OCB (labeled as Ni-SACs@OCB), in which Ni single atoms are stabilized by a N, O-mediated coordination configuration. This Ni-SACs@OCB catalyst shows high H₂O₂ selectivity (95%) in a range of 0.2-0.7 V undergoing a two-electron oxygen reduction process, with a kinetic current density of 2.8 mA cm^-2 and a mass activity of 24 A g_cat.^-1 at 0.65 V (vs RHE). In practice, H-cells with Ni-SACs@OCB as catalysts displayed a high H₂O₂ production rate of 98.5 mmol g_cat.^-1 h^-1 with negligible current loss during testing, suggesting the high H₂O₂ generation efficiency and robust stability. DFT theoretical calculations revealed that Ni single-atom sites coordinated by O, N coordination exhibit advantages in oxygen adsorption and increased reactivity toward the intermediate species, *OOH, which is beneficial to high selectivity for H₂O₂ production. This work provides a novel N, O-mediated four-coordinate nickel single-atom catalyst as a promising candidate for practical decentralized production of H₂O₂.

[IL-8 Links NF-κB and Wnt/β-Catenin Pathways in Persistent Inflammatory Response Induced by Chronic Helicobacter pylori Infection]

Helicobacter pylori (H. pylori) infection can cause persistent inflammatory response in human gastric mucosal epithelial cells, which may result in the occurrence of cancer. However, the underlying mechanism of carcinogenesis has not been elucidated yet. Herein, we established the models of chronic H. pylori infection in GES-1 cells and C57BL/6J mice. Interleukin 8 (IL-8) level was detected by ELISA. The expression of NF-κB p65, IL-8, Wnt2 and β-catenin mRNA and proteins was evaluated by real-time PCR, Western blotting, immunofluorescence staining, and immunohistochemistry. The infection of H. pylori in mice was evaluated by rapid urease test, H&E staining and Warthin-Starry silver staining. The morphological changes of gastric mucosa were observed by electron microscopy. Our results showed that in H. pylori infected gastric mucosal cells along with activation of NF-κB signaling pathway and increase of IL-8 level, the expression of Wnt2 was also increased significantly, which preliminarily indicates that IL-8 can positively regulate the expression of Wnt2. Studies in chronic H. pylori infected C57BL/6J mice models showed that there was an increased incidence of premalignant lesions in the gastric mucosa tissue. Through comparing changes of gastric mucosal cell ultrastructure and analyzing the relationship between NF-κB signaling pathway and Wnt2 expression, we found that H. pylori infection activated NF-κB signal pathways, and the massive release of IL-8 was positively correlated with the high expression of Wnt2 protein. Subsequently, the activated Wnt/β-catenin signal pathways may be involved in the malignant transformation of gastric mucosal cells. Collectively, H. pylori chronic infection may continuously lead to persistent inflammatory response: activate NF-κB pathway, promote IL-8 release and thereby activate Wnt/β-catenin pathway. IL-8 probably plays an important role of a linker in coupling these two signal pathways.

LC-MS/MS-Based Absolute Quantitation of Hemoglobin Subunits from Dried Blood Spots Reveals Novel Biomarkers for α-Thalassemia Silent Carriers

Identification of α-thalassemia silent carriers is challenging with conventional phenotype-based screening methods. A liquid chromatography tandem mass spectrometry (LC-MS/MS)-based approach may offer novel biomarkers to address this conundrum. In this study, we collected dried blood spot samples from individuals with three α-thalassemia subtypes for biomarker discovery and validation. We observed differential expression patterns of hemoglobin subunits among various α-thalassemia subtypes and normal controls through proteomic profiling of 51 samples in the discovery phase. Then, we developed and optimized a multiple reaction monitoring (MRM) assay to measure all detectable hemoglobin subunits. The validation phase was conducted in a cohort of 462 samples. Among the measured hemoglobin subunits, subunit μ was significantly upregulated in all the α-thalassemia groups with distinct fold changes. The hemoglobin subunit μ exhibits great potential as a novel biomarker for α-thalassemia, especially for silent α-thalassemia. We constructed predictive models based on the concentrations of hemoglobin subunits and their ratios to classify the various subtypes of α-thalassemia. In the binary classification problems of silent α-thalassemia vs normal, non-deletional α-thalassemia vs normal, and deletional α-thalassemia vs normal, the best performance of the models achieved average ROCAUCs of 0.9505, 0.9430, and 0.9976 in the cross-validation, respectively. In the multiclass model, the best performance achieved an average ROCAUC of 0.9290 in cross-validation. The performance of our MRM assay and models demonstrated that the hemoglobin subunit μ would play a vital role in screening silent α-thalassemia in clinical practice.

Incidence and characteristics of aspiration pneumonia in adults in Beijing, China, 2011-2017

OBJECTIVES

This study aimed to estimate aspiration pneumonia (AP) incidence and describe comorbid characteristics and mortality in Beijing, China.

STUDY DESIGN

A historical cohort study was conducted based on medical claim records.

METHODS

Patients admitted with a primary diagnosis of AP were identified from approximately 12 million adults who enrolled in the Urban Employee Basic Medical Insurance program in Beijing, China, from January 2011 to December 2017. The incidences of AP and pneumonia with risk factors for aspiration (PRFA) were estimated by a Poisson distribution. The estimated annual percentage change was reported to represent the average percentage change in incidence per year. Characteristics and 6-month and 1-year all-cause mortality rates for AP and suspected AP patients were described and compared with community-acquired pneumonia (CAP).

RESULTS

The incidence rates of hospitalized AP and PRFA were 9.4 (95% confidence interval [CI]: 7.6, 11.3) and 102.9 (95% CI: 95.8, 110.3) per 100,000 person-years, respectively. The incidences increased rapidly with age and were stable across the observed years. Patients with AP and PRFA possessed a greater burden of comorbidities than CAP (mean age-adjusted Charlson comorbidity indices for AP: 7.72, PRFA: 7.83, and CAP: 2.84). The 6-month and 1-year all-cause mortality rates for those with AP and PRFA were higher than those for patients with CAP (6-month mortality, AP: 35.2%, PRFA: 21.8%, CAP: 11.1%; 1-year mortality, AP: 42.7%, PRFA: 26.6%, CAP: 13.2%).

CONCLUSIONS

The incidence of AP and PRFA in Beijing was reported, presenting a full picture of the disease burden. The results provide baseline information for AP prevention.

Decomposing the Response Time in Amblyopia: A Drift Diffusion Model Analysis

Purpose

Amblyopes are known to have delayed response times (RT) in various visual tasks. We aim to investigate whether any factor other than the sensory deficit contributes to the delayed RT in amblyopia.

Method

Fifteen amblyopic (26.0 ± 4.50 years) and 15 normal (25.6 ± 2.90 years) participants took part in this study. The responses and RTs in an orientation identification task were collected for each participant with stimulus contrast adjusted to the multiples of individual's threshold. A drift diffusion model was used to fit to the response and RT data and to estimate the RT components.

Result

There was a significant difference in the RT between the amblyopic and normal groups (F(1, 28) = 6.75, P = 0.015) but no difference in the accuracy (F(1, 28) = 0.028, P = 0.868). The drift rate function in the amblyopic eye had a larger threshold (P = 0.001) and shallower slope (P = 0.006) than that of the fellow eye. The amblyopic group has a longer non-decision time than the normal group (F(1, 28) = 8.02, P = 0.008). The drift rate threshold correlated with the contrast sensitivity (P = 1.71 × 10-18) but the non-decision time did not (P = 0.393).

Conclusions

Both sensory and post-sensory factors contributed to the delayed RT in amblyopia. The effect of the sensory loss in V1 on RT can be compensated by increasing stimulus contrast, and the post-sensory delay provides evidence for higher-level deficits in amblyopia.

Reward-Adaptive Reinforcement Learning: Dynamic Policy Gradient Optimization for Bipedal Locomotion

Controlling a non-statically bipedal robot is challenging due to the complex dynamics and multi-criterion optimization involved. Recent works have demonstrated the effectiveness of deep reinforcement learning (DRL) for simulation and physical robots. In these methods, the rewards from different criteria are normally summed to learn a scalar function. However, a scalar is less informative and may be insufficient to derive effective information for each reward channel from the complex hybrid rewards. In this work, we propose a novel reward-adaptive reinforcement learning method for biped locomotion, allowing the control policy to be simultaneously optimized by multiple criteria using a dynamic mechanism. The proposed method applies a multi-head critic to learn a separate value function for each reward component, leading to hybrid policy gradients. We further propose dynamic weight, allowing each component to optimize the policy with different priorities. This hybrid and dynamic policy gradient (HDPG) design makes the agent learn more efficiently. We show that the proposed method outperforms summed-up-reward approaches and is able to transfer to physical robots. The MuJoCo results further demonstrate the effectiveness and generalization of HDPG.

Fashion Retrieval via Graph Reasoning Networks on a Similarity Pyramid

Matching clothing images from customers and online shopping stores has rich applications in e-commerce. Existing algorithms mostly encode an image as a global feature vector and perform retrieval via global representation matching. However, distinctive local information on clothing is immersed in this global representation, resulting in sub-optimized performance. To address this issue, we propose a novel graph reasoning network (GRNet) on a similarity pyramid, which learns similarities between a query and a gallery cloth by using both initial pairwise multi-scale feature representations and matching propagation for unaligned representations. The query local representations at each scale are aligned with those of the gallery via an adaptive window pooling module. The similarity pyramid is represented by a similarity graph, where nodes represent similarities between clothing components at different scales, and the final matching score is obtained by message propagation along edges. In GRNet, graph reasoning is solved by training a graph convolutional network, enabling the alignment of salient clothing components to improve clothing retrieval. To facilitate future research, we introduce a new benchmark, i.e. FindFashion, containing rich annotations of bounding boxes, views, occlusions, and cropping. Extensive experiments show that GRNet obtains new state-of-the-art results on three challenging benchmarks, e.g. pushing the accuracy of top-1, top-20, and top-50 on DeepFashion to 27, 66, and 75 percent (i.e. 6, 12, and 10 percent absolute improvements), outperforming competitors with large margins. On FindFashion, GRNet achieves considerable improvements on all empirical settings.

Routing User-Interest Markov Tree for Scalable Personalized Knowledge-Aware Recommendation

To facilitate more accurate and explainable recommendation, it is crucial to incorporate side information into user-item interactions. Recently, knowledge graph (KG) has attracted much attention in a variety of domains due to its fruitful facts and abundant relations. However, the expanding scale of real-world data graphs poses severe challenges. In general, most existing KG-based algorithms adopt exhaustively hop-by-hop enumeration strategy to search all the possible relational paths, this manner involves extremely high-cost computations and is not scalable with the increase of hop numbers. To overcome these difficulties, in this article, we propose an end-to-end framework Knowledge-tree-routed UseR-Interest Trajectories Network (KURIT-Net). KURIT-Net employs the user-interest Markov trees (UIMTs) to reconfigure a recommendation-based KG, striking a good balance for routing knowledge between short-distance and long-distance relations between entities. Each tree starts from the preferred items for a user and routes the association reasoning paths along the entities in the KG to provide a human-readable explanation for model prediction. KURIT-Net receives entity and relation trajectory embedding (RTE) and fully reflects potential interests of each user by summarizing all reasoning paths in a KG. Besides, we conduct extensive experiments on six public datasets, our KURIT-Net significantly outperforms state-of-the-art approaches and shows its interpretability in recommendation.

Changes in the mammary gland during aging and its links with breast diseases

The functional capacity of organisms declines in the process of aging. In the case of breast tissue, abnormal mammary gland development can lead to dysfunction in milk secretion, a primary function, as well as the onset of various diseases, such as breast cancer. In the process of aging, the terminal duct lobular units (TDLUs) within the breast undergo gradual degeneration, while the proportion of adipose tissue in the breast continues to increase and hormonal levels in the breast change accordingly. Here, we review changes in morphology, internal structure, and cellular composition that occur in the mammary gland during aging. We also explore the emerging mechanisms of breast aging and the relationship between changes during aging and breast-related diseases, as well as potential interventions for delaying mammary gland aging and preventing breast disease.

Template-Based Contrastive Distillation Pretraining for Math Word Problem Solving

Since math word problem (MWP) solving aims to transform natural language problem description into executable solution equations, an MWP solver needs to not only comprehend the real-world narrative described in the problem text but also identify the relationships among the quantifiers and variables implied in the problem and maps them into a reasonable solution equation logic. Recently, although deep learning models have made great progress in MWPs, they ignore the grounding equation logic implied by the problem text. Besides, as we all know, pretrained language models (PLM) have a wealth of knowledge and high-quality semantic representations, which may help solve MWPs, but they have not been explored in the MWP-solving task. To harvest the equation logic and real-world knowledge, we propose a template-based contrastive distillation pretraining (TCDP) approach based on a PLM-based encoder to incorporate mathematical logic knowledge by multiview contrastive learning while retaining rich real-world knowledge and high-quality semantic representation via knowledge distillation. We named the pretrained PLM-based encoder by our approach as MathEncoder. Specifically, the mathematical logic is first summarized by clustering the symbolic solution templates among MWPs and then injected into the deployed PLM-based encoder by conducting supervised contrastive learning based on the symbolic solution templates, which can represent the underlying solving logic in the problems. Meanwhile, the rich knowledge and high-quality semantic representation are retained by distilling them from a well-trained PLM-based teacher encoder into our MathEncoder. To validate the effectiveness of our pretrained MathEncoder, we construct a new solver named MathSolver by replacing the GRU-based encoder with our pretrained MathEncoder in GTS, which is a state-of-the-art MWP solver. The experimental results demonstrate that our method can carry a solver's understanding ability of MWPs to a new stage by outperforming existing state-of-the-art methods on two widely adopted benchmarks Math23K and CM17K. Code will be available at https://github.com/QinJinghui/tcdp.

Regulation of Interfacial Lattice Oxygen Activity by Full-Surface Modification Engineering towards Long Cycling Stability for Co-Free Li-Rich Mn-Based Cathode

The construction of a protective layer for stabilizing anion redox reaction is the key to obtaining long cycling stability for Li-rich Mn-based cathode materials. However, the protection of the exposed surface/interface of the primary particles inside the secondary particles is usually ignored and difficult, let alone the investigation of the impact of the surface engineering of the internal primary particles on the cycling stability. In this work, an efficient method to regulate cycling stability is proposed by simply adjusting the distribution state of the boron nickel complexes coating layer. Theoretical calculation and experimental results display that the full-surface boron nickel complexes coating layer can not only passivate the activity of interface oxygen and improve its stability but also play the role of sharing voltage and protective layer to gradually activate the oxygen redox reaction during cycling. As a result, the elaborately designed cobalt-free Li-rich Mn-based cathode displays the highest discharge-specific capacity retentions of 91.1% after 400 cycles at 1 C and 94.3% even after 800 cycles at 5 C. In particular, the regulation strategy has well universality and is suitable for other high-capacity Li-rich cathode materials.

[Efficacy and safety of epidermal growth factor receptor tyrosine kinase inhibitors combined with chemotherapy as first-line treatment for epidermal growth factor receptor-mutant advanced non-small cell lung cancer]

Objective: To observe the clinical efficacy and safety of epidermal growth factor receptor tyrosine kinase inhibitors (EGFR-TKIs) combined with chemotherapy as first-line treatment for EGFR mutant advanced non-small cell lung cancer (NSCLC). Methods: It was a retrospective, single-arm real-world study and a total of 39 patients with stage ⅢB to Ⅳ EGFR mutant NSCLC diagnosed in Cancer Hospital of Chinese Academy of Medical Sciences from July 2018 to December 2020 were collected. There were 16 males and 23 females, the age ranged from 25 to 73 years, with a median age of 53 years. All patients received EGFR-TKIs synchronously combined with pemetrexed and platinum-containing chemotherapy for 4-6 cycles as first-line treatment, followed by EGFR-TKI monotherapy with or without pemetrexed maintenance therapy. The objective response rate (ORR), disease control rate (DCR), progression-free survival (PFS) and adverse reactions were evaluated. Median follow-up time was 18.6 months (95%CI: 16.2-21.0 months). The Kaplan-Meier method was used for survival analysis. Results: The ORR was 61.5% (24/39), the DCR was 94.9% (37/39) and the median PFS was 16.4 months (95%CI: 12.1-20.7 months). The main adverse reactions were liver function injury (59.0%, 23/39), myelosuppression (43.6%, 17/39), skin reaction (25.6%, 10/39), gastrointestinal reaction (17.9%, 7/39), fatigue (12.8%, 5/39) and kidney injury (5.1%, 2/39). Most of the patients had grade 1-2 adverse reactions, and the rate of grade 3 adverse events were 12.8%(5/39), which were effectively alleviated after symptomatic support treatment, no grade 4 serious adverse events occurred. Conclusion: EGFR-TKIs synchronously combined with chemotherapy followed by EGFR-TKI monotherapy with or without pemetrexed maintenance therapy has a certain therapeutic effect and fairly good safety, which can prolong PFS in patients with EGFR mutated advanced NSCLC.

Abstract 3187: Improved anti-tumor immune functions of iPSC-derived NK cells with TGFβR2 knock-out and/or IL-15 knock-in by TALEN® editing for use alone or in combination with GPC3 Flex-NKTM bispecific antibody

Background: Induced Pluripotent Stem Cell (iPSC)-derived NK cells (iNK) offer an opportunity to generate unlimited homogenous NK cells as allogeneic off-the-shelf therapies. Interleukin-15 (IL-15) signaling enhances proliferation, persistence, cytotoxicity, and metabolic fitness of NK cells. Activation of TGF-β signaling suppresses anti-tumor functions of immune cells, including NK cells, in the tumor microenvironment (TME). In addition, previously, we have reported that our Flex-NKTM bispecific antibody that engages NK cells through NKp46 can enhance the cytotoxicity of the non-edited iNK cells. Therefore, we hypothesized that iNK cells with IL-15 knock-in (KI) and/or TGFβR2 knock-out (KO) could exhibit improved immune function and overcome the immunosuppressive TME. Furthermore, the activity of these edited universal iNK cells could be enhanced when combined with CYT-303, a Flex-NKTM bispecific antibody NK engager targeting GPC3 expressed on many solid tumors including hepatocellular carcinoma (HCC).

Methods: IL15 was knocked-in and/or TGFβR2 was knocked-out in iPSC using Cellectis TALEN® and the verified edited iPSC clones were differentiated and expanded into NK cells. The functional significance of these edits in iNKs were assessed in IL-15 and TGF-β dependent NK cell assays evaluating survival and proliferation, expression of activating receptors, as well as cytolysis of these iNK cells against HCC tumor cells. Cytotoxic activity of these edited cells was also tested in a serial killing assay with or without CYT-303 in the absence or presence of TGF-β.

Results: Compared to the non-edited iNK cells, iNKs with IL-15 KI can extend persistence in vitro in the absence of exogeneous cytokines. In the presence of TGF-β, the expression level of a number of NK cell activating receptors, such as NKG2D, DNAM-1, and NKp30, was decreased and lower cytotoxicity against HCC tumor cells was observed. However, this TGF-β-mediated immune suppression was reversed in TGFβR2 KO iNKs cells which also presented enhanced cytotoxicity against HCC cells. Furthermore, the anti-HCC cytotoxic activity of either single (IL-15 KI) or double edited (TGFβR2 KO and Il-15 KI) iNK cells were further enhanced by CYT-303, even in the presence of TGF-β. Serial killing assays against HCC tumor cells showed that the iNK cell dysfunction observed in later rounds of killing could still be reversed by the combination of these gene edits and CYT-303.

Conclusions: This work demonstrates that KI of IL-15 and KO of TGFβR2 is a promising strategy for TALEN®-engineered iNK cell therapies to overcome the immunosuppressive TME and mount a potent and persistent anti-tumor immune response. The data also provide a solid foundation for combining these edited iNK cells with CYT-303 to address the immunosuppressive TME towards a cure for HCC.

Citation Format: Liang Lin, Peng Gao, An-Ping Chen, Hao-Ming Chang, Preeti Ashok, David Zou, Marshall Chao Ma, Justin Eyquem, Alex Boyne, Alexandre Juillerat, Laurent Poirot, Philippe Duchateau, Armin Rath, Elisabetta Burchi, Daniel Teper, Antonio Arulanandam, Wei Li. Improved anti-tumor immune functions of iPSC-derived NK cells with TGFβR2 knock-out and/or IL-15 knock-in by TALEN® editing for use alone or in combination with GPC3 Flex-NKTM bispecific antibody [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3187.

[Analysis of the clinical features and the risk factors of severe human metapneu movirus-associated community acquired pneumonia in children]

Objective: To investigate the clinical characteristics and the risk factors of severe human metapneumovirus (hMPV)-associated community acquired pneumonia (CAP) in children. Methods: A retrospective case summary was conducted. From December 2020 to March 2022, 721 children who were diagnosed with CAP and tested positive for hMPV nucleic acid by PCR-capillary electrophoresis fragment analysis of nasopharyngeal secretions at the Yuying Children's Hospital, the Second Affiliated Hospital of Wenzhou Medical University were selected as the research objects. The clinical characteristics, epidemiological characteristics and mixed pathogens of the two groups were analyzed. According to CAP diagnostic criteria, the children were divided into the severe group and the mild group. Chi-square test or Mann-Whitney rank and contrast analysis was used for comparison between groups, while multivariate Logistic regression was applied to analyze the risk factors of the severe hMPV-associated CAP. Results: A total of 721 children who were diagnosed with hMPV-associated CAP were included in this study, with 397 males and 324 females. There were 154 cases in the severe group. The age of onset was 1.0 (0.9, 3.0) years, <3 years old 104 cases (67.5%), and the length of hospital stay was 7 (6, 9) days. In the severe group, 67 children (43.5%) were complicated with underlying diseases. In the severe group, 154 cases (100.0%) had cough, 148 cases (96.1%) had shortness of breath and pulmonary moist rales, and 132 cases (85.7%) had fever, 23 cases (14.9%) were complicated with respiratory failure. C-reactive protein (CRP) was elevated in 86 children (55.8%), including CRP≥50 mg/L in 33 children (21.4%). Co-infection was detected in 77 cases (50.0%) and 102 strains of pathogen were detected, 25 strains of rhinovirus, 17 strains of Mycoplasma pneumoniae, 15 strains of Streptococcus pneumoniae, 12 strains of Haemophilus influenzae and 10 strains of respiratory syncytial virus were detected. Six cases (3.9%) received heated and humidified high flow nasal cannula oxygen therapy, 15 cases (9.7%) were admitted to intensive care unit, and 2 cases (1.3%) received mechanical ventilation. In the severe group, 108 children were cured, 42 children were improved, 4 chlidren were discharged automatically without recovery and no death occurred. There were 567 cases in the mild group. The age of onset was 2.7 (1.0, 4.0) years, and the length of hospital stay was 4 (4, 6) days.Compared with the mild group, the proportion of children who age of disease onset <6 months, CRP≥50 mg/L, the proportions of preterm birth, congenital heart disease, malnutrition, congenital airway malformation, neuromuscular disease, mixed respiratory syncytial viruses infection were higher (20 cases (13.0%) vs. 31 cases (5.5%), 32 cases (20.8%) vs. 64 cases (11.3%), 23 cases (14.9%) vs. 44 cases (7.8%), 11 cases (7.1%) vs. 18 cases (3.2%), 9 cases (5.8%) vs. 6 cases (1.1%), 11 cases (7.1%) vs. 12 cases (2.1%), 8 cases (5.2%) vs. 4 cases (0.7%), 10 cases (6.5%) vs. 13 cases (2.3%), χ²=0.42, 9.45, 7.40, 4.94, 11.40, 8.35, 3.52, 6.92, all P<0.05). Multivariate Logistic regression analysis showed that age<6 months (OR=2.51, 95%CI 1.29-4.89), CRP≥50 mg/L (OR=2.20, 95%CI 1.36-3.57), prematurity (OR=2.19, 95%CI 1.26-3.81), malnutrition (OR=6.05, 95%CI 1.89-19.39) were the independent risk factors for severe hMPV-associated CAP. Conclusions: Severe hMPV-associated CAP is most likely to occur in infants under 3 years old and has a higher proportion of underlying diseases and co-infection. The main clinical manifestations are cough, shortness of breath and pulmonary moist rales, fever. The overall prognosis is good. Age<6 months, CRP≥50 mg/L, preterm birth, malnutrition are the independent risk factors for severe hMPV-associated CAP.

Prognostic value of apparent diffusion coefficient in neuroendocrine carcinomas of the uterine cervix

Objectives

This research was designed to examine the associations between the apparent diffusion coefficient (ADC) values and clinicopathological parameters, and to explore the prognostic value of ADC values in predicting the International Federation of Gynecology and Obstetrics (FIGO) stage and outcome of patients suffering from neuroendocrine carcinomas of the uterine cervix (NECCs).

Methods

This retrospective study included 83 patients with NECCs, who had undergone pre-treatment magnetic resonance imaging (MRI) between November 2002 and June 2019. The median follow-up period was 50.7 months. Regions of interest (ROIs) were drawn manually by two radiologists. ADC values in the lesions were calculated using the Functool software. These values were compared between different clinicopathological parameters groups. The Kaplan–Meier approach was adopted to forecast survival rates. Prognostic factors were decided by the Cox regression method.

Results

In the cohort of 83 patients, nine, 42, 23, and nine patients were in stage I, II, III, and IV, respectively. ADCmean, ADCmax, and ADCmin were greatly lower in stage IIB–IVB than in stage I–IIA tumours, as well as in tumours measuring ≥ 4 cm than in those < 4 cm. ADCmean, FIGO stage, and age at dianosis were independent prognostic variables for the 5-year overall survival (OS). ADCmin, FIGO stage, age at diagnosis and para-aortic lymph node metastasis were independent prognostic variables for the 5-year progression-free survival (PFS) in multivariate analysis. For surgically treated patients (n = 45), ADCmax was an independent prognostic parameter for both 5-year OS and 5-year PFS.

Conclusions

ADCmean, ADCmin, and ADCmax are independent prognostic factors for NECCs. ADC analysis could be useful in predicting the survival outcomes in patients with NECCs.

Derivation and Preclinical Characterization of CYT-303, a Novel NKp46-NK Cell Engager Targeting GPC3

Glypican-3 (GPC3) is an oncofetal antigen that is highly expressed in multiple solid tumors, including hepatocellular carcinoma, and is barely expressed in adult normal tissues except the placenta. NKp46 activation receptor is expressed in all-natural killer (NK) cells, including tumor-infiltrating NK cells. FLEX-NK^TM is a platform for the production of tetravalent multifunctional antibody NK cell engagers (NKE). CYT-303 was designed using the FLEX-NK scaffold, incorporating a novel humanized NKp46 binder that does not induce NKp46 internalization and a humanized GPC3 binder that targets the membrane-proximal lobe to mediate NK cell-redirected killing of HCC tumors. CYT-303 shows sub-nanomolar binding affinities to both GPC3 and NKp46. CYT-303 was highly potent and effective in mediating NK cell-redirected cytotoxicity against multiple HCC tumor cell lines and tumor spheroids. More interestingly, it can reverse the dysfunction induced in NK cells following repeated rounds of serial killing of tumors. It also mediated antibody-dependent cellular phagocytosis (ADCP) and complement-dependent cytotoxicity against GPC3-expressing HCC tumors. In vivo, CYT-303 showed no toxicity or cytokine release in cynomolgus monkeys up to the highest dose (60 mg/kg), administered weekly by intravenous infusion for 28 days. These results demonstrate the potential of CYT-303 to be a safe and effective therapy against HCC.