Self-supervised deep learning of gene-gene interactions for improved gene expression recovery.

Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool to gain biological insights at the cellular level. However, due to technical limitations of the existing sequencing technologies, low gene expression values are often omitted, leading to inaccurate gene counts. Existing methods, including advanced deep learning techniques, struggle to reliably impute gene expressions due to a lack of mechanisms that explicitly consider the underlying biological knowledge of the system. In reality, it has long been recognized that gene-gene interactions may serve as reflective indicators of underlying biology processes, presenting discriminative signatures of the cells. A genomic data analysis framework that is capable of leveraging the underlying gene-gene interactions is thus highly desirable and could allow for more reliable identification of distinctive patterns of the genomic data through extraction and integration of intricate biological characteristics of the genomic data. Here we tackle the problem in two steps to exploit the gene-gene interactions of the system. We first reposition the genes into a 2D grid such that their spatial configuration reflects their interactive relationships. To alleviate the need for labeled ground truth gene expression datasets, a self-supervised 2D convolutional neural network is employed to extract the contextual features of the interactions from the spatially configured genes and impute the omitted values. Extensive experiments with both simulated and experimental scRNA-seq datasets are carried out to demonstrate the superior performance of the proposed strategy against the existing imputation methods.

MicroSegNet: A deep learning approach for prostate segmentation on micro-ultrasound images.

Micro-ultrasound (micro-US) is a novel 29-MHz ultrasound technique that provides 3-4 times higher resolution than traditional ultrasound, potentially enabling low-cost, accurate diagnosis of prostate cancer. Accurate prostate segmentation is crucial for prostate volume measurement, cancer diagnosis, prostate biopsy, and treatment planning. However, prostate segmentation on micro-US is challenging due to artifacts and indistinct borders between the prostate, bladder, and urethra in the midline. This paper presents MicroSegNet, a multi-scale annotation-guided transformer UNet model designed specifically to tackle these challenges. During the training process, MicroSegNet focuses more on regions that are hard to segment (hard regions), characterized by discrepancies between expert and non-expert annotations. We achieve this by proposing an annotation-guided binary cross entropy (AG-BCE) loss that assigns a larger weight to prediction errors in hard regions and a lower weight to prediction errors in easy regions. The AG-BCE loss was seamlessly integrated into the training process through the utilization of multi-scale deep supervision, enabling MicroSegNet to capture global contextual dependencies and local information at various scales. We trained our model using micro-US images from 55 patients, followed by evaluation on 20 patients. Our MicroSegNet model achieved a Dice coefficient of 0.939 and a Hausdorff distance of 2.02 mm, outperforming several state-of-the-art segmentation methods, as well as three human annotators with different experience levels. Our code is publicly available at https://github.com/mirthAI/MicroSegNet and our dataset is publicly available at https://zenodo.org/records/10475293.

p-Cresol Sulfate Is a Sensitive Urinary Marker of Fecal Microbiota Transplantation and Antibiotics Treatments in Human Patients and Mouse Models.

Fecal microbiota transplantation (FMT) has emerged as a highly effective therapy for recurrent Clostridioides difficile infection (rCDI) and also a potential therapy for other diseases associated with dysbiotic gut microbiota. Monitoring metabolic changes in biofluids and excreta is a noninvasive approach to identify the biomarkers of microbial recolonization and to understand the metabolic influences of FMT on the host. In this study, the pre-FMT and post FMT urine samples from 11 rCDI patients were compared through metabolomic analyses for FMT-induced metabolic changes. The results showed that p-cresol sulfate in urine, a microbial metabolite of tyrosine, was rapidly elevated by FMT and much more responsive than other microbial metabolites of aromatic amino acids (AAAs). Because patients were treated with vancomycin prior to FMT, the influence of vancomycin on the microbial metabolism of AAAs was examined in a mouse feeding trial, in which the decreases in p-cresol sulfate, phenylacetylglycine, and indoxyl sulfate in urine were accompanied with significant increases in their AAA precursors in feces. The inhibitory effects of antibiotics and the recovering effects of FMT on the microbial metabolism of AAAs were further validated in a mouse model of FMT. Overall, urinary p-cresol sulfate may function as a sensitive and convenient therapeutic indicator on the effectiveness of antibiotics and FMT for the desired manipulation of gut microbiota in human patients.

Migration experiences and reported commercial and non-commercial sexual behaviors among newly diagnosed HIV infections in China: a cross-sectional study.

BACKGROUND: Migration is known to influence human health. China has a high migration rate and a significant number of people who are HIV-positive, but little is known about how these factors intersect in sexual risk behaviors. OBJECTIVE: This study aimed to explore sexual risk behaviors between migrants and non-migrants among newly diagnosed HIV infections, and assess the changes of sexual risk behaviors with length of stay in the current city of migrants. METHODS: A cross-sectional questionnaire was conducted among people newly diagnosed with HIV from July 2018 to December 2020 who lived in Zhejiang Province. In the study, sexual risk behaviors included having multiple sexual partners and unprotected sexual behaviors (in commercial sexual behaviors, non-commercial sexual behaviors, heterosexual behaviors, and homosexual behaviors). Binary logistic regression models were employed to explore the influencing factors of sexual risk behaviors, measured by multiple sexual partners and unprotected sexual partners. RESULTS: A total of 836 people newly diagnosed with HIV/AIDS were incorporated in the study and 65.31% (546) were migrants. The percentages of non-commercial sexual behaviors among migrants were statistically higher than those of non-migrants. Commercial heterosexual behavior was higher among non-migrants compared with migrants. The proportion of study participants having unprotected sexual behaviors and multiple sexual partners with commercial/non-commercial partners was both higher among migrants compared with non-migrants. Among migrants, the likelihood of sexual risk behaviors in both commercial and non-commercial sex increased in the first 3 years and reduced after 10 years. Compared with non-migrants, migrants were statistically associated with multiple sexual partners [P = .007, odds ratio (OR) = 1.942]. However, migrants did not exhibit a significant difference in unprotected sexual behaviors compared with non-migrants. In addition, migrants aged between 18 and 45 years who relocated to the current city in the past 2-3 years tended to have multiple sexual partners (P < .05). CONCLUSIONS: People newly diagnosed with HIV engaged in different sexual risk behaviors among migrants and non-migrants and more attention should be paid to migrants. For non-migrants, it is urgent to promote the prevention of commercial sexual behaviors. For migrants, prevention of non-commercial sexual behaviors and universal access to health care especially for new arrivals who migrated to the current city for 2-3 years are needed. Moreover, sexual health education and early HIV diagnosis are necessary for the entire population.

The Vulnerability of International Floating Populations to Sexually Transmitted Infections: A Qualitative Study.

With the rapid development of the global economy, along with globalisation, the health of international floating populations (especially their sexual health) has become a problem that cannot be ignored. This study explored the potential vulnerability of international floating populations to sexually transmitted infections (STIs) from the points of view of society, religion, culture, migration, community environment, and personal behaviours. In-depth exploratory interviews with 51 members of the international floating population living in China were conducted in June and July 2022. A qualitative thematic analysis methodology was used to analyse the content of these interviews. We found that a conservative culture orientated around religion leads to a lack of sex education, resulting in insufficient personal knowledge as well as a lack of the motivation and awareness required to encourage condom use during sexual contact. Additionally, both geographical isolation and reduced social supervision have expanded personal space, which has led to social isolation and marginalisation, in addition to challenges for coping with STI risk. These factors have increased the possibility of individuals engaging in risky behaviours.

Patient-Specific Auto-segmentation on Daily kVCT Images for Adaptive Radiation Therapy.

PURPOSE: This study explored deep-learning-based patient-specific auto-segmentation using transfer learning on daily RefleXion kilovoltage computed tomography (kVCT) images to facilitate adaptive radiation therapy, based on data from the first group of patients treated with the innovative RefleXion system. METHODS AND MATERIALS: For head and neck (HaN) and pelvic cancers, a deep convolutional segmentation network was initially trained on a population data set that contained 67 and 56 patient cases, respectively. Then the pretrained population network was adapted to the specific RefleXion patient by fine-tuning the network weights with a transfer learning method. For each of the 6 collected RefleXion HaN cases and 4 pelvic cases, initial planning computed tomography (CT) scans and 5 to 26 sets of daily kVCT images were used for the patient-specific learning and evaluation separately. The performance of the patient-specific network was compared with the population network and the clinical rigid registration method and evaluated by the Dice similarity coefficient (DSC) with manual contours being the reference. The corresponding dosimetric effects resulting from different auto-segmentation and registration methods were also investigated. RESULTS: The proposed patient-specific network achieved mean DSC results of 0.88 for 3 HaN organs at risk (OARs) of interest and 0.90 for 8 pelvic target and OARs, outperforming the population network (0.70 and 0.63) and the registration method (0.72 and 0.72). The DSC of the patient-specific network gradually increased with the increment of longitudinal training cases and approached saturation with more than 6 training cases. Compared with using the registration contour, the target and OAR mean doses and dose-volume histograms obtained using the patient-specific auto-segmentation were closer to the results using the manual contour. CONCLUSIONS: Auto-segmentation of RefleXion kVCT images based on the patient-specific transfer learning could achieve higher accuracy, outperforming a common population network and clinical registration-based method. This approach shows promise in improving dose evaluation accuracy in RefleXion adaptive radiation therapy.

Label-Efficient Self-Supervised Federated Learning for Tackling Data Heterogeneity in Medical Imaging.

The collection and curation of large-scale medical datasets from multiple institutions is essential for training accurate deep learning models, but privacy concerns often hinder data sharing. Federated learning (FL) is a promising solution that enables privacy-preserving collaborative learning among different institutions, but it generally suffers from performance deterioration due to heterogeneous data distributions and a lack of quality labeled data. In this paper, we present a robust and label-efficient self-supervised FL framework for medical image analysis. Our method introduces a novel Transformer-based self-supervised pre-training paradigm that pre-trains models directly on decentralized target task datasets using masked image modeling, to facilitate more robust representation learning on heterogeneous data and effective knowledge transfer to downstream models. Extensive empirical results on simulated and real-world medical imaging non-IID federated datasets show that masked image modeling with Transformers significantly improves the robustness of models against various degrees of data heterogeneity. Notably, under severe data heterogeneity, our method, without relying on any additional pre-training data, achieves an improvement of 5.06%, 1.53% and 4.58% in test accuracy on retinal, dermatology and chest X-ray classification compared to the supervised baseline with ImageNet pre-training. In addition, we show that our federated self-supervised pre-training methods yield models that generalize better to out-of-distribution data and perform more effectively when fine-tuning with limited labeled data, compared to existing FL algorithms. The code is available at https://github.com/rui-yan/SSL-FL.

Organ-specific fuel rewiring in acute and chronic hypoxia redistributes glucose and fatty acid metabolism.

Oxygen deprivation can be detrimental. However, chronic hypoxia is also associated with decreased incidence of metabolic syndrome and cardiovascular disease in high-altitude populations. Previously, hypoxic fuel rewiring has primarily been studied in immortalized cells. Here, we describe how systemic hypoxia rewires fuel metabolism to optimize whole-body adaptation. Acclimatization to hypoxia coincided with dramatically lower blood glucose and adiposity. Using in vivo fuel uptake and flux measurements, we found that organs partitioned fuels differently during hypoxia adaption. Acutely, most organs increased glucose uptake and suppressed aerobic glucose oxidation, consistent with previous in vitro investigations. In contrast, brown adipose tissue and skeletal muscle became "glucose savers," suppressing glucose uptake by 3-5-fold. Interestingly, chronic hypoxia produced distinct patterns: the heart relied increasingly on glucose oxidation, and unexpectedly, the brain, kidney, and liver increased fatty acid uptake and oxidation. Hypoxia-induced metabolic plasticity carries therapeutic implications for chronic metabolic diseases and acute hypoxic injuries.

Vulnerability to HIV Infection Among International Immigrants in China: Cross-sectional Web-Based Survey.

BACKGROUND: The rising number of migrants worldwide, including in China given its recent rapid economic development, poses a challenge for the public health system to prevent infectious diseases, including sexually transmitted infections (STIs) caused by risky sexual behaviors. OBJECTIVE: The aim of this study was to explore the risky sexual behaviors of international immigrants living in China to provide evidence for establishment of a localized public health service system. METHODS: Risky sexual behaviors were divided into multiple sexual partners and unprotected sexual behaviors. Basic characteristics, sexual knowledge, and behaviors of international immigrants were summarized with descriptive statistics. Multivariate logistic regression analyses were used to identify factors associated with risky sexual behaviors, and the associations of demographic characteristics and risk behaviors with HIV testing and intention to test for HIV. RESULTS: In total, 1433 international immigrants were included in the study, 61.76% (n=885) of whom had never heard of STIs, and the mean HIV knowledge score was 5.42 (SD 2.138). Overall, 8.23% (118/1433) of the participants had been diagnosed with an STI. Among the 1433 international immigrants, 292 indicated that they never use a condom for homosexual sex, followed by sex with a stable partner (n=252), commercial sex (n=236), group sex (n=175), and casual sex (n=137). In addition, 119 of the international immigrants had more than three sex partners. Individuals aged 31-40 years were more likely to have multiple sexual partners (adjusted odds ratio [AOR] 2.364, 95% CI 1.149-4.862). Married participants were more likely to have unprotected sexual behaviors (AOR 3.096, 95% CI -1.705 to 5.620), whereas Asians were less likely to have multiple sexual partners (AOR 0.446, 95% CI 0.328-0.607) and unprotected sexual behaviors (AOR 0.328, 95% CI 0.219-0.492). Women were more likely to have taken an HIV test than men (AOR 1.413, 95% CI 1.085-1.841). Those who were married (AOR 0.577, 95% CI 0.372-0.894), with an annual disposable income >150,000 yuan (~US $22,000; AOR 0.661, 95% CI 0.439-0.995), considered it impossible to become infected with HIV (AOR 0.564, 95% CI 0.327-0.972), and of Asian ethnicity (AOR 0.330, 95% CI 0.261-0.417) were less likely to have an HIV test. People who had multiple sexual partners were more likely to have taken an HIV test (AOR 2.041, 95% CI 1.442-2.890) and had greater intention to test for HIV (AOR 1.651, 95% CI 1.208-2.258). CONCLUSIONS: International immigrants in China exhibit risky sexual behaviors, especially those aged over 30 years. In addition, the level of HIV-related knowledge is generally low. Therefore, health interventions such as targeted, tailored programming including education and testing are urgently needed to prevent new HIV infections and transmission among international immigrants and the local population.

Small-Object Sensitive Segmentation Using Across Feature Map Attention.

Semantic segmentation is an important step in understanding the scene for many practical applications such as autonomous driving. Although Deep Convolutional Neural Networks-based methods have significantly improved segmentation accuracy, small/thin objects remain challenging to segment due to convolutional and pooling operations that result in information loss, especially for small objects. This article presents a novel attention-based method called Across Feature Map Attention (AFMA) to address this challenge. It quantifies the inner-relationship between small and large objects belonging to the same category by utilizing the different feature levels of the original image. The AFMA could compensate for the loss of high-level feature information of small objects and improve the small/thin object segmentation. Our method can be used as an efficient plug-in for a wide range of existing architectures and produces much more interpretable feature representation than former studies. Extensive experiments on eight widely used segmentation methods and other existing small-object segmentation models on CamVid and Cityscapes demonstrate that our method substantially and consistently improves the segmentation of small/thin objects.

Opioid-induced microbial dysbiosis disrupts irinotecan (CPT-11) metabolism and increases gastrointestinal toxicity in a murine model.

BACKGROUND AND PURPOSE: Opioids are commonly used for the management of cancer-associated pain and chemotherapy-induced diarrhoea. The chemotherapeutic irinotecan (CPT-11) causes severe gastrointestinal (GI) toxicity due to deconjugation of inactive metabolite SN-38 glucuronide (SN-38G) by bacterial ß-glucuronidases to the active 7-ethyl-10-hydroxycamptothecin (SN-38). Opioids are known to cause gut microbial dysbiosis, this study evaluated whether CPT-11 anti-tumour efficacy and GI toxicity are exacerbated by opioid co-administration. EXPERIMENTAL APPROACH: Eight-week-old C57BL/6 male mice were co-administration with CPT-11 ± opioid. 16S rRNA sequencing was used for gut microbiome analysis. LC-MS analyses of plasma and intestinal extracts were performed to investigate the pharmacokinetic profile of CPT-11. Histological analysis and quantitative real-time polymerase chain reaction were used to determine the severity of intestinal tissue damage. Human liver microsome In vitro assay was performed to confirm the effects of opioids on CPT-11 metabolism. KEY RESULTS: Gut microbiome analysis showed that morphine treatment induced enrichment of ß-glucuronidase-producing bacteria in the intestines of CPT-11-treated mice, resulting in SN-38 accumulation and exacerbation of GI toxicity in the small intestine. Oral administration of both antibiotics and glucuronidase inhibitor protected mice against GI toxicity induced with CPT-11 and morphine co-administration, implicating a microbiome-dependent mechanism. Additionally, morphine and loperamide decreased the plasma concentration of SN-38 and compromised CPT-11 anti-tumour efficacy, this seemed to be microbiome independent. CONCLUSION AND IMPLICATIONS: Gut microbiota play a significant role in opioid and chemotherapeutic agent drug-drug interactions. Inhibition of gut microbial glucuronidase may also prevent adverse GI effects of CPT-11 in patients on opioids.

HIV Detection and Delayed Diagnosis: A Time Series Analysis in China.

BACKGROUND: Insufficient HIV detection and late presentation to antiretroviral therapy (ART) pose significant public health challenges. In China, universal access to HIV testing is available now. Under this background, we aim to analyze the trends of HIV detection and the prevalence of delayed HIV diagnosis (DHD) in order to provide evidence for HIV prevention and treatment in China. METHODS: Data of HIV tests in Hangzhou city between 2007 and 2018 were collected from the Chinese National HIV/AIDS Comprehensive Response Information Management System (CRIMS). Descriptive statistics were used to describe the characteristics of HIV testing and detection and the prevalence of DHD among newly diagnosed HIV cases. Non-parametric tests were employed to examine the prevalence of DHD among newly diagnosed HIV cases. Moreover, logistic regression models were employed to explore the influencing factors of DHD. RESULTS: Testing rates doubled from 14.1% in 2007-2010 to 28.2% in 2016-2018. The total positive rate of HIV tests was 5.3 per 10,000. Preoperative testing was the predominant pathway for HIV tests, accounting for 41.9%, followed by testing for health screening, maternal examination and other patients, accounting for 18.4%, 13.2% and 11.8%, respectively. Meanwhile, the predominant pathway for HIV case detection was also preoperative testing, accounting for 29.1%, followed by testing for other patients, testing at STD clinics and VCT, with the proportions of 18.8%, 15.8% and 13.6%, respectively. MSM (men who have sex with men) contact was the main transmission route, accounting for 55.3%, followed by heterosexual contact, accounting for 41.6%. Overall, DHD occurred in 29.0% of the newly diagnosed cases, and this rate had not improved over the years. A higher prevalence of DHD was found in those diagnosed through a pre-test for receiving blood/products [OR (95%CI): 5.42(2.95-9.97)], detection of other patients [OR (95%CI): 2.08(1.64-2.63)], preoperative testing [OR (95%CI): 1.83(1.44-2.32)] and spouse or sexual partner testing in positive person [OR (95%CI): 1.93(1.34-2.78)] compared with those diagnosed at a VCT clinic. Heterosexuals [OR (95%CI): 1.20(1.06-1.36)] had a higher risk of DHD than MSM. Diagnosis at a CDC [OR (95%CI): 0.68(0.55-0.83)] and community health centers [OR (95%CI): 0.54(0.39-0.75)] had a lower risk of DHD than in hospitals. Older age, males, being single/divorced or widowed and floating population were also associated with DHD. CONCLUSIONS: In China, DHD had not improved in the last 10 years, although HIV testing had been expanded. Therefore, it is important for continued efforts to promote early diagnosis of HIV to prevent transmission, morbidity and early mortality in HIV infection.

Multiple Instance Neuroimage Transformer.

For the first time, we propose using a multiple instance learning based convolution-free transformer model, called Multiple Instance Neuroimage Transformer (MINiT), for the classification of T1-weighted (T1w) MRIs. We first present several variants of transformer models adopted for neuroimages. These models extract non-overlapping 3D blocks from the input volume and perform multi-headed self-attention on a sequence of their linear projections. MINiT, on the other hand, treats each of the non-overlapping 3D blocks of the input MRI as its own instance, splitting it further into non-overlapping 3D patches, on which multi-headed self-attention is computed. As a proof-of-concept, we evaluate the efficacy of our model by training it to identify sex from T1w-MRIs of two public datasets: Adolescent Brain Cognitive Development (ABCD) and the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA). The learned attention maps highlight voxels contributing to identifying sex differences in brain morphometry. The code is available at https://github.com/singlaayush/MINIT.

Primary Health Institutions and Service Quality in China: Implications for Health Policy.

BACKGROUND: To protect and improve the health of populations, the important role of primary health institutions has been strengthened through a series of health policies, especially the implementation of a national hierarchical diagnosis and treatment system. In this light, we aim to evaluate the development of primary health institutions between 2013, before the implementation of the hierarchical diagnosis and treatment system, and 2020 as well as people's perception of the quality of primary healthcare services. METHOD: The national-level data (e.g., the numbers of primary health institutions, personnel, beds, visits, and hospitalizations) regarding primary health institutions were collected from the Health Statistics Yearbook, and the perceptions of the quality of primary healthcare services were collected by a web-based questionnaire survey using an internationally recognized assessment tool (i.e., PCAT-AE). In total, 10,850 persons were surveyed, and 10,419 participants were incorporated into the final analysis after removing invalid questionnaires. A descriptive statistical analysis (i.e., frequency and percentage) was used to analyze the national-level characteristics of primary health institutions and people's perceptions of the quality of primary healthcare services. Moreover, a logistic regression model was used to analyze the factors influencing the perceptions of the quality of primary healthcare services. RESULTS: From the macro perspective, the number of primary health institutions, beds, and personnel per 10 thousand residents slightly increased from 2013 to 2020, especially in the eastern and central areas. However, the average number of visits and the hospitalization rate in primary health institutions showed a decrease, especially in central and eastern areas. Among participants, 92.2% (9606/10,419) of them had previously sought healthcare services in primary health institutions, and most were seeking general outpatient services (57.06-63.45%), followed by medicine purchasing (16.49-21.51%), physical examinations (9.91-11.49%), preventive health services (5.11-6.48%), and hospitalization services (3.17-5.67%). The total perception scores on the quality of primary healthcare services reported by the participants were 26.19 and 27.00 for rural and urban areas, respectively, which accounted for 65.5% and 67.5% of the total score, respectively, and 26.62, 26.86, and 25.89 for the eastern, central, and western areas, respectively, with percentages of 66.6%, 67.2%, and 64.7%. The perception score on the quality among people contracted with a family doctor (29.83, 74.58%) was much higher than those who were not (25.25, 63.13%), and the difference was statistically significant (p < 0.001). Moreover, people who were female, married, had higher incomes, and were diagnosed with various diseases had better perceptions of the primary healthcare services compared to their counterparts (p < 0.05). CONCLUSION: Improvements were seen for primary health institutions, especially in terms of hardware resources such as beds and personnel. However, the service utilization in primary health institutions did not improve between 2013 and 2020. The perception score on the quality of primary healthcare was moderate to low in rural and urban as well as eastern, central, and western areas, but it was significantly higher among people contracted with a family doctor than those who were not. Therefore, it is important for policy makers to take or adjust measures focusing on quality improvement and increasing the service utilization in primary health institutions with good first contact, accessibility, continuity, comprehensiveness, and coordination, such as raising the enrollment rate of family doctors and promoting the provision of high-quality services.

External Attention Assisted Multi-Phase Splenic Vascular Injury Segmentation With Limited Data.

The spleen is one of the most commonly injured solid organs in blunt abdominal trauma. The development of automatic segmentation systems from multi-phase CT for splenic vascular injury can augment severity grading for improving clinical decision support and outcome prediction. However, accurate segmentation of splenic vascular injury is challenging for the following reasons: 1) Splenic vascular injury can be highly variant in shape, texture, size, and overall appearance; and 2) Data acquisition is a complex and expensive procedure that requires intensive efforts from both data scientists and radiologists, which makes large-scale well-annotated datasets hard to acquire in general. In light of these challenges, we hereby design a novel framework for multi-phase splenic vascular injury segmentation, especially with limited data. On the one hand, we propose to leverage external data to mine pseudo splenic masks as the spatial attention, dubbed external attention, for guiding the segmentation of splenic vascular injury. On the other hand, we develop a synthetic phase augmentation module, which builds upon generative adversarial networks, for populating the internal data by fully leveraging the relation between different phases. By jointly enforcing external attention and populating internal data representation during training, our proposed method outperforms other competing methods and substantially improves the popular DeepLab-v3+ baseline by more than 7% in terms of average DSC, which confirms its effectiveness.

Utilization of, satisfaction toward, and challenges for Internet-based healthcare services provided by primary health institutions: Evidence from China.

Background: The rapid development of "Internet plus healthcare" in China has provided new ways for the innovative development of primary healthcare. In addition, a series of favorable policies have been issued to promote Internet-based healthcare services in primary health institutions. Objective: The aim of this study was to describe the utilization of, satisfaction toward, and challenges faced by Internet-based healthcare services provided by primary health institutions in China. Methods: A self-designed structured questionnaire was employed to collect related data in January 2022 through Credamo. The questionnaire mainly included sociodemographic characteristics, health-related information, utilization of, satisfaction toward, and challenges faced by Internet-based healthcare services provided by primary health institutions. Descriptive analysis was used to describe the sociodemographic characteristics, utilization, satisfaction, and challenges by subgroups. The Wilcoxon rank-sum test was carried out to examine the differences in satisfaction with Internet-based healthcare services between participants who ever received these services and those who did not. A multiple logistic regression model was also used to examine the factors influencing the utilization of Internet-based healthcare services provided by primary health institutions. Results: A total of 10,600 residents were included in the final analysis, of whom 5,754 (54.3%) were women. Overall, 51.3% (5,434) of the total participants ever used Internet-based healthcare services provided by primary health institutions. Among those who used Internet-based healthcare services, the most widely used services were procedure-related consultation services (63.7%). The satisfaction among those who ever used it was significantly higher than that among those who did not (84.7 vs. 45.4%; p-value < 0.001). One of the biggest challenges (69.3%) expressed by the residents was that it was difficult for the elderly to use Internet-based services, followed by community doctors with low capacity of providing primary care online (49.0%) and residents were worried about the information security and privacy protection (48.5%). Younger people, people with lower education levels, and people with chronic diseases were significantly more likely to use Internet-based healthcare services provided by primary health institutions (P < 0.05). Conclusion: Among 10,600 residents surveyed in China in 2022, more than half of the people used Internet-based healthcare services provided by primary health institutions, and most of them were satisfied, although subgroups significant differences existed. The most common use was procedure-related (e.g., online registration and result query), and several challenges of using Internet-based healthcare services exist (e.g., information safety and usage among elderly people). Therefore, it is important to further improve Internet-based primary healthcare services based on the population perception of achieving healthy China in 2030.

Rethinking Architecture Design for Tackling Data Heterogeneity in Federated Learning.

Federated learning is an emerging research paradigm enabling collaborative training of machine learning models among different organizations while keeping data private at each institution. Despite recent progress, there remain fundamental challenges such as the lack of convergence and the potential for catastrophic forgetting across real-world heterogeneous devices. In this paper, we demonstrate that self-attention-based architectures (e.g., Transformers) are more robust to distribution shifts and hence improve federated learning over heterogeneous data. Concretely, we conduct the first rigorous empirical investigation of different neural architectures across a range of federated algorithms, real-world benchmarks, and heterogeneous data splits. Our experiments show that simply replacing convolutional networks with Transformers can greatly reduce catastrophic forgetting of previous devices, accelerate convergence, and reach a better global model, especially when dealing with heterogeneous data. We release our code and pretrained models to encourage future exploration in robust architectures as an alternative to current research efforts on the optimization front.

Learning Inductive Attention Guidance for Partially Supervised Pancreatic Ductal Adenocarcinoma Prediction.

Pancreatic ductal adenocarcinoma (PDAC) is the third most common cause of cancer death in the United States. Predicting tumors like PDACs (including both classification and segmentation) from medical images by deep learning is becoming a growing trend, but usually a large number of annotated data are required for training, which is very labor-intensive and time-consuming. In this paper, we consider a partially supervised setting, where cheap image-level annotations are provided for all the training data, and the costly per-voxel annotations are only available for a subset of them. We propose an Inductive Attention Guidance Network (IAG-Net) to jointly learn a global image-level classifier for normal/PDAC classification and a local voxel-level classifier for semi-supervised PDAC segmentation. We instantiate both the global and the local classifiers by multiple instance learning (MIL), where the attention guidance, indicating roughly where the PDAC regions are, is the key to bridging them: For global MIL based normal/PDAC classification, attention serves as a weight for each instance (voxel) during MIL pooling, which eliminates the distraction from the background; For local MIL based semi-supervised PDAC segmentation, the attention guidance is inductive, which not only provides bag-level pseudo-labels to training data without per-voxel annotations for MIL training, but also acts as a proxy of an instance-level classifier. Experimental results show that our IAG-Net boosts PDAC segmentation accuracy by more than 5% compared with the state-of-the-arts.

Added value of deep learning-based liver parenchymal CT volumetry for predicting major arterial injury after blunt hepatic trauma: a decision tree analysis.

PURPOSE: In patients presenting with blunt hepatic injury (BHI), the utility of CT for triage to hepatic angiography remains uncertain since simple binary assessment of contrast extravasation (CE) as being present or absent has only modest accuracy for major arterial injury on digital subtraction angiography (DSA). American Association for the Surgery of Trauma (AAST) liver injury grading is coarse and subjective, with limited diagnostic utility in this setting. Volumetric measurements of hepatic injury burden could improve prediction. We hypothesized that in a cohort of patients that underwent catheter-directed hepatic angiography following admission trauma CT, a deep learning quantitative visualization method that calculates % liver parenchymal disruption (the LPD index, or LPDI) would add value to CE assessment for prediction of major hepatic arterial injury (MHAI). METHODS: This retrospective study included adult patients with BHI between 1/1/2008 and 5/1/2017 from two institutions that underwent admission trauma CT prior to hepatic angiography (n = 73). Presence (n = 41) or absence (n = 32) of MHAI (pseudoaneurysm, AVF, or active contrast extravasation on DSA) served as the outcome. Voxelwise measurements of liver laceration were derived using an existing multiscale deep learning algorithm trained on manually labeled data using cross-validation with a 75-25% split in four unseen folds. Liver volume was derived using a pre-trained whole liver segmentation algorithm. LPDI was automatically calculated for each patient by determining the percentage of liver involved by laceration. Classification and regression tree (CART) analyses were performed using a combination of automated LPDI measurements and either manually segmented CE volumes, or CE as a binary sign. Performance metrics for the decision rules were compared for significant differences with binary CE alone (the current standard of care for predicting MHAI), and the AAST grade. RESULTS: 36% of patients (n = 26) had contrast extravasation on CT. Median [Q1-Q3] automated LPDI was 4.0% [1.0-12.1%]. 41/73 (56%) of patients had MHAI. A decision tree based on auto-LPDI and volumetric CE measurements (CEvol) had the highest accuracy (0.84, 95% CI 0.73-0.91) with significant improvement over binary CE assessment (0.68, 95% CI 0.57-0.79; p = 0.01). AAST grades at different cut-offs performed poorly for predicting MHAI, with accuracies ranging from 0.44-0.63. Decision tree analysis suggests an auto-LPDI cut-off of ≥ 12% for minimizing false negative CT exams when CE is absent or diminutive. CONCLUSION: Current CT imaging paradigms are coarse, subjective, and limited for predicting which BHIs are most likely to benefit from AE. LPDI, automated using deep learning methods, may improve objective personalized triage of BHI patients to angiography at the point of care.

Adversarial Metric Attack and Defense for Person Re-Identification.

Person re-identification (re-ID) has attracted much attention recently due to its great importance in video surveillance. In general, distance metrics used to identify two person images are expected to be robust under various appearance changes. However, our work observes the extreme vulnerability of existing distance metrics to adversarial examples, generated by simply adding human-imperceptible perturbations to person images. Hence, the security danger is dramatically increased when deploying commercial re-ID systems in video surveillance. Although adversarial examples have been extensively applied for classification analysis, it is rarely studied in metric analysis like person re-identification. The most likely reason is the natural gap between the training and testing of re-ID networks, that is, the predictions of a re-ID network cannot be directly used during testing without an effective metric. In this work, we bridge the gap by proposing Adversarial Metric Attack, a parallel methodology to adversarial classification attacks. Comprehensive experiments clearly reveal the adversarial effects in re-ID systems. Meanwhile, we also present an early attempt of training a metric-preserving network, thereby defending the metric against adversarial attacks. At last, by benchmarking various adversarial settings, we expect that our work can facilitate the development of adversarial attack and defense in metric-based applications.