Rethinking Human-AI Collaboration in Complex Medical Decision Making: A Case Study in Sepsis Diagnosis.

Today's AI systems for medical decision support often succeed on benchmark datasets in research papers but fail in real-world deployment. This work focuses on the decision making of sepsis, an acute life-threatening systematic infection that requires an early diagnosis with high uncertainty from the clinician. Our aim is to explore the design requirements for AI systems that can support clinical experts in making better decisions for the early diagnosis of sepsis. The study begins with a formative study investigating why clinical experts abandon an existing AI-powered Sepsis predictive module in their electrical health record (EHR) system. We argue that a human-centered AI system needs to support human experts in the intermediate stages of a medical decision-making process (e.g., generating hypotheses or gathering data), instead of focusing only on the final decision. Therefore, we build SepsisLab based on a state-of-the-art AI algorithm and extend it to predict the future projection of sepsis development, visualize the prediction uncertainty, and propose actionable suggestions (i.e., which additional laboratory tests can be collected) to reduce such uncertainty. Through heuristic evaluation with six clinicians using our prototype system, we demonstrate that SepsisLab enables a promising human-AI collaboration paradigm for the future of AI-assisted sepsis diagnosis and other high-stakes medical decision making.

Establishment and verification of a prognostic model based on coagulation and fibrinolysis-related genes in hepatocellular carcinoma.

BACKGROUND: Studies have shown that coagulation and fibrinolysis (CFR) are correlated with Hepatocellular carcinoma (HCC) progression and prognosis. We aim to build a model based on CFR-correlated genes for risk assessment and prediction of HCC patient. METHODS: HCC samples were selected from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases respectively. The Molecular Signatures Database (MSigDB) was used to select the CFR genes. RiskScore model were established by single sample gene set enrichment analysis (ssGSEA), weighted correlation network analysis (WGCNA), multivariate Cox regression analysis, LASSO regression analysis. RESULTS: PCDH17, PGF, PDE2A, FAM110D, FSCN1, FBLN5 were selected as the key genes and designed a RiskScore model. Those key genes were Differential expressions in HCC cell and patients. Overexpression PDE2A inhibited HCC cell migration and invasion. The higher the RiskScore, the lower the probability of survival. The model has high AUC values in the first, third and fifth year prediction curves, indicating that the model has strong prediction performance. The difference analysis of clinicopathological features found that a great proportion of high clinicopathological grade samples showed higher RiskScore. RiskScore were positively correlated with immune scores and TIDE scores. High levels of immune checkpoints and immunomodulators were observed in high RiskScore group. High RiskScore groups may benefit greatly from taking traditional chemotherapy drugs. CONCLUSIONS: We screened CFR related genes to design a RiskScore model, which could accurately evaluate the prognosis and survival status of HCC patients, providing certain value for optimizing the clinical treatment of cancer in the future.

Geographic differences in pharmacotherapy patterns and outcomes of acute ischemic stroke in China.

BACKGROUND: Vast economic and healthcare status discrepancies exist among regions in China, contributing to different treatment patterns. This study was aimed to investigate the current status of pharmacotherapy for acute ischemic stroke (AIS) and outcomes in China and explore the geographic variation in stroke care. METHODS: This study was a multicenter prospective registry study, which collected the data of patients with AIS from 80 hospitals in 46 cities in 2015-2017 across China. Poor functional outcome defined as a modified Rankin Scale score of 3-6 was assessed at 3 and 12 months. Multivariate logistic regression was used. RESULTS: Among 9973 eligible patients, the number of receiving intravenous thrombolysis (IVT), antiplatelet agents, anticoagulants, statin and human urinary kallidinogenase was 429 (4.3%), 9363 (93.9%), 1063 (10.7%), 6828 (74.7%) and 5112 (51.2%), respectively. Multivariable analysis showed IVT use in northeastern was significantly more frequent than in eastern region (OR = 3.17, 95% CI, 2.53-3.99), while the antiplatelets agents use were less frequent (OR = 0.46, 95%CI: 0.38-0.57). The proportions of poor outcomes at 3 and 12 months were 20.7% and 15.8%, respectively. Multivariate analysis showed AIS patients from northeastern and central region had significantly lower risk of poor outcome at month 3 and 12 than those from eastern region (all P < 0.05). CONCLUSIONS: There was a low IVT use and a high antiplatelet agent and statin use for AIS in China. The pharmacotherapy and prognosis of AIS had variation by geographic region. TRIAL REGISTRATION: This study was registered with ClinicalTrials.gov (NCT02470624).

Trametinib ameliorates aging-associated gut pathology in Drosophila females by reducing Pol III activity in intestinal stem cells.

Pharmacological therapies are promising interventions to slow down aging and reduce multimorbidity in the elderly. Studies in animal models are the first step toward translation of candidate molecules into human therapies, as they aim to elucidate the molecular pathways, cellular mechanisms, and tissue pathologies involved in the anti-aging effects. Trametinib, an allosteric inhibitor of MEK within the Ras/MAPK (Ras/Mitogen-Activated Protein Kinase) pathway and currently used as an anti-cancer treatment, emerged as a geroprotector candidate because it extended lifespan in the fruit fly Drosophila melanogaster. Here, we confirm that trametinib consistently and robustly extends female lifespan, and reduces intestinal stem cell (ISC) proliferation, tumor formation, tissue dysplasia, and barrier disruption in guts in aged flies. In contrast, pro-longevity effects of trametinib are weak and inconsistent in males, and it does not influence gut homeostasis. Inhibition of the Ras/MAPK pathway specifically in ISCs is sufficient to partially recapitulate the effects of trametinib. Moreover, in ISCs, trametinib decreases the activity of the RNA polymerase III (Pol III), a conserved enzyme synthesizing transfer RNAs and other short, non-coding RNAs, and whose inhibition also extends lifespan and reduces gut pathology. Finally, we show that the pro-longevity effect of trametinib in ISCs is partially mediated by Maf1, a repressor of Pol III, suggesting a life-limiting Ras/MAPK-Maf1-Pol III axis in these cells. The mechanism of action described in this work paves the way for further studies on the anti-aging effects of trametinib in mammals and shows its potential for clinical application in humans.

New insight into PAH4 induced hepatotoxicity and the dose-response assessment in rats model.

PAH4 (sum of benzo[a]pyrene, chrysene, benz[a]anthracene and benzo[b]fluoranthene) has been proposed as better marker than benzo[a]pyrene to assess total PAHs exposure in foodstuffs. However, the toxicological behaviors of PAH4 combined exposure remain unclear. This study aimed to investigate PAH4 toxicity effects with non-targeted metabolomics approach and evaluate the external and internal dose-response relationships based on benchmark dose (BMD) analysis. Male Sprague-Dawley rats were treated by gavage with vehicle (corn oil) or four doses of PAH4 (10, 50, 250, 1000 µg/kg·bw) for consecutive 30 days. After the final dose, the liver, blood and urine samples of rats were subsequently collected for testing. The concentrations of urinary mono-hydroxylated PAHs metabolites (OH-PAHs) including 3-hydroxybenzo[a]pyrene (3-OHB[a]P), 3-hydroxychrysene (3-OHCHR) and 3-hydroxybenz[a]anthracene (3-OHB[a]A) were determined to reflect internal PAH4 exposure. Our results showed PAH4 exposure increased relative liver weight and serum aspartate aminotransferase (AST) activity and caused hepatocyte swelling and degeneration, implying hepatotoxicity induced by PAH4. Serum metabolomics suggested PAH4 exposure perturbed lipid metabolism through upregulating the expression of glycerolipids metabolites, which was evidenced by markedly increased serum triglyceride (TG) level and hepatic TG content. Additionally, urinary OH-PAHs concentrations presented strong positive correlations with the external dose, indicating they were able to reflect PAH4 exposure. Furthermore, PAH4 exposure led to a dose-response increase of hepatic TG content, based on which the 95% lower confidence value of BMDs for external and internal doses were estimated as 5.45 µg/kg·bw and 0.11 µmol/mol·Cr, respectively. In conclusion, this study suggested PAH4 exposure could induce hepatotoxicity and lipid metabolism disorder, evaluating the involved dose-response relationships and providing a basis for the risk assessment of PAHs.

Hippo pathway in intestinal diseases: focusing on ferroptosis.

The incidence of intestinal diseases, such as inflammatory bowel disease, gastric cancer, and colorectal cancer, has steadily increased over the past decades. The Hippo pathway is involved in cell proliferation, tissue and organ damage, energy metabolism, tumor formation, and other physiologic processes. Ferroptosis is a form of programmed cell death characterized by the accumulation of iron and lipid peroxides. The Hippo pathway and ferroptosis are associated with various intestinal diseases; however, the crosstalk between them is unclear. This review elaborates on the current research on the Hippo pathway and ferroptosis in the context of intestinal diseases. We summarized the connection between the Hippo pathway and ferroptosis to elucidate the underlying mechanism by which these pathways influence intestinal diseases. We speculate that a mutual regulatory mechanism exists between the Hippo pathway and ferroptosis and these two pathways interact in several ways to regulate intestinal diseases.

Stimuli-Responsive Protein Hydrogels: Their Design, Properties, and Biomedical Applications.

Protein-based hydrogels are considered ideal biomaterials due to their high biocompatibility, diverse structure, and their improved bioactivity and biodegradability. However, it remains challenging to mimic the native extracellular matrices that can dynamically respond to environmental stimuli. The combination of stimuli-responsive functionalities with engineered protein hydrogels has facilitated the development of new smart hydrogels with tunable biomechanics and biological properties that are triggered by cyto-compatible stimuli. This review summarizes the recent advancements of responsive hydrogels prepared from engineered proteins and integrated with physical, chemical or biological responsive moieties. We underscore the design principles and fabrication approaches of responsive protein hydrogels, and their biomedical applications in disease treatment, drug delivery, and tissue engineering are briefly discussed. Finally, the current challenges and future perspectives in this field are highlighted.

Integrated Tandem Electrochemical-chemical-electrochemical Coupling of Biomass and Nitrate to Sustainable Alanine.

Alanine is widely employed for synthesizing polymers, pharmaceuticals, and agrochemicals. Electrocatalytic coupling of biomass molecules and waste nitrate is attractive for the nitrate removal and alanine production under ambient conditions. However, the reaction efficiency is relatively low due to the activation of the stable substrates, and the coupling of two reactive intermediates remains challenging. Herein, we realize the integrated tandem electrochemical-chemical-electochemical synthesis of alanine from the biomass-derived pyruvic acid (PA) and waste nitrate (NO3 - ) catalyzed by PdCu nano-bead-wires (PdCu NBWs). The overall reaction pathway is demonstrated as a multiple-step catalytic cascade process via coupling the reactive intermediates NH2 OH and PA on the catalyst surface. Interestingly, in this integrated tandem electrochemical-chemical-electrochemical catalytic cascade process, Cu facilitates the electrochemical reduction of nitrate to NH2 OH intermediates, which chemically couple with PA to form the pyruvic oxime, and Pd promotes the electrochemical reduction of pyruvic oxime to the desirable alanine. This work provides a green strategy to convert waste NO3 - to wealth and enriches the substrate scope of renewable biomass feedstocks to produce high-value amino acids.

Effect of weekend versus weekday admission on the mortality of acute ischemic stroke patients in China: an analysis of data from the Chinese acute ischemic stroke treatment outcome registry.

Background: Due to disparities in medical resources in rural and urban areas as well as in different geographic regions in China, the effect of weekend versus weekday admission on the outcomes of acute ischemic stroke (AIS) patients is unknown. Our aim was to investigate whether the outcomes of AIS patients differ according to the day of admission in China. Methods: The data were extracted from the Chinese Acute Ischemic Stroke Treatment Outcome Registry (CASTOR), a multicenter prospective study database of patients diagnosed with AIS. The chi-square test (χ2) and logistic regression were used to assess mortality for weekday and weekend admissions among AIS patients stratified by rural or urban status and geographic region (including the eastern, northeastern, central, and western regions). Results: In total, 9,256 patients were included in this study. Of these patients, 57.2% were classified as urban, and 42.8% were classified as rural. A total of 6,760 (73%) patients were admitted on weekdays, and 2,496 (27%) were admitted on weekends. There was no significant difference in the mortality rate among patients admitted on weekends compared with those admitted on weekdays in urban (7.5% versus 7.4%) or rural areas (8.8% versus 8.1%; p > 0.05). The mortality rate was the highest among patients admitted on weekends and weekdays (11.6% versus 10.3%) in the northeastern area, without statistical significance before and after adjusting for the patients' background characteristics (p > 0.05). In addition, regression analysis revealed that the mortality of patients admitted on weekdays was more likely to be influenced by regional subgroup, hospital level and intravenous thrombolysis than that of patients admitted on weekends. Conclusion: The weekend effect was not observed in the mortality of patients with AIS regardless of rural-urban status or geographic region in China.

Targeted co-delivery of PD-L1 monoclonal antibody and sorafenib to circulating tumor cells via platelet-functionalized nanocarriers.

BACKGROUND: Circulating tumor cells (CTCs) can adsorb and activate platelets to form a microthrombus protective barrier around them, so that therapeutic drugs and immune cells cannot effectively kill CTCs. The platelet membrane (PM) bionic carrying drug system has the powerful ability of immune escape, and can circulate in the blood for a long time. MATERIALS AND METHODS: we developed platelet membrane coated nanoparticles (PM HMSNs) to improve the precise delivery of drugs to tumor sites and to achieve more effective immunotherapy combined with chemotherapy strategy. RESULTS: Successfully prepared aPD-L1-PM-SO@HMSNs particles, whose diameter is 95-130 nm and presenting the same surface protein as PM. Laser confocal microscopy and flow cytometry experimental results showed that the fluorescence intensity of aPD-L1-PM-SO@HMSNs was greater than SO@HMSNs that are not coated by PM. Biodistribution studies in H22 tumor-bearing mice showed that due to the combined action of the active targeting effect and the EPR effect, the high accumulation of aPD-L1-PM-SO@HMSNs in the local tumor was more effective in inhibiting tumor growth than other groups of therapeutic agents. CONCLUSION: Platelet membrane biomimetic nanoparticles have a good targeted therapeutic effect, which can effectively avoid immune clearance and have little side effects. It provides a new direction and theoretical basis for further research on targeted therapy of CTCs in liver cancer.

Correlating the Valence State with the Adsorption Behavior of a Cu-Based Electrocatalyst for Furfural Oxidation with Anodic Hydrogen Production Reaction.

The low-potential furfural oxidation reaction (FFOR) on a Cu-based electrocatalyst can produce H2 at the anode, thereby providing a bipolar H2 production system with an ultralow cell voltage. However, the intrinsic activity and stability of the Cu-based electrocatalyst for the FFOR remain unsatisfactory for practical applications. This study investigates the correlation between the valence state and the adsorption behavior of the Cu-based electrocatalyst in furfural oxidation. Cu0 is the adsorption site with low intrinsic activity. Cu+ , which exists in the form of Cu(OH)ads in alkaline electrolytes, has no adsorption ability but can improve the performance of Cu0 by promoting the adsorption of FF. Moreover, a mixed-valence Cu-based electrocatalyst (MV Cu) with high intrinsic activity and stability is prepared electrochemically. With the MV Cu catalyst, the assembled dual-side H2 production electrolyzer has a low electricity requirement of only 0.24 kWh mH2 -3 at an ultralow cell voltage of 0.3 V, and it exhibits sufficient stability. This study not only correlates the valence state with the adsorption behavior of the Cu-based electrocatalyst for the low-potential FFOR with anodic H2 production but also reveals the mechanism of deactivation to provide design principles for Cu-based electrocatalysts with satisfactory stability.

Promoted hydrogen and acetaldehyde production from alcohol dehydrogenation enabled by electrochemical hydrogen pumps.

The dehydrogenation reaction of bioderived ethanol is of particular interest for the synthesis of fuels and value-added chemicals. However, this reaction historically suffered from high energy consumption (>260 °C or >0.8 V) and low efficiency. Herein, the efficient conversion of alcohol to hydrogen and aldehyde is achieved by integrating the thermal dehydrogenation reaction with electrochemical hydrogen transfer at low temperature (120 °C) and low voltage (0.06 V), utilizing a bifunctional catalyst (Ru/C) with both thermal-catalytic and electrocatalytic activities. Specifically, the coupled electrochemical hydrogen separation procedure can serve as electrochemical hydrogen pumps, which effectively promote the equilibrium of ethanol dehydrogenation toward hydrogen and acetaldehyde production and simultaneously purifies hydrogen at the cathode. By utilizing this strategy, we achieved boosted hydrogen and acetaldehyde yields of 1,020 mmol g-1 h-1 and 1,185 mmol g-1 h-1, respectively, which are threefold higher than the exclusive ethanol thermal dehydrogenation. This work opens up a prospective route for the high-efficiency production of hydrogen and acetaldehyde via coupled thermal-electrocatalysis.

Transcranial Doppler Ultrasonography detection on cerebrovascular flow for evaluating neonatal hypoxic-ischemic encephalopathy modeling.

Objective: This study aimed to investigate the feasibility of Transcranial Doppler Ultrasonography (TCD) in evaluating neonatal hypoxic-ischemic encephalopathy (NHIE) modeling through monitoring the alteration of cerebrovascular flow in neonatal hypoxic-ischemic (HI) rats. Methods: Postnatal 7-day-old Sprague Dawley (SD) rats were divided into the control group, HI group, and hypoxia (H) group. TCD was applied to assess the changes of cerebral blood vessels, cerebrovascular flow velocity, and heart rate (HR) in sagittal and coronal sections at 1, 2, 3, and 7 days after the operation. For accuracy, cerebral infarct of rats was examined by 2,3,5-Triphenyl tetrazolium chloride (TTC) staining and Nissl staining to simultaneously verify the establishment of NHIE modeling. Results: Coronal and sagittal TCD scans revealed obvious alteration of cerebrovascular flow in main cerebral vessels. Obvious cerebrovascular back-flow was observed in anterior cerebral artery (ACA), basilar artery (BA), middle cerebral artery (MCA) of HI rats, along with accelerated cerebrovascular flows in the left internal carotid artery (ICA-L) and BA, decreased flows in right internal carotid artery (ICA-R) relative to those in the H and control groups. The alterations of cerebral blood flows in neonatal HI rats indicated successful ligation of right common carotid artery. Besides, TTC staining further validated the cerebral infarct was indeed caused due to ligation-induced insufficient blood supply. Damage to nervous tissues was also revealed by Nissl staining. Conclusion: Cerebral blood flow assessment by TCD in neonatal HI rats contributed to cerebrovascular abnormalities observed in a real-time and non-invasive way. The present study elicits the potentials to utilize TCD as an effective means for monitoring the progression of injury as well as NHIE modeling. The abnormal appearance of cerebral blood flow is also beneficial to the early warning and effective detection in clinical practice.

Ageing-associated changes in transcriptional elongation influence longevity.

Physiological homeostasis becomes compromised during ageing, as a result of impairment of cellular processes, including transcription and RNA splicing1-4. However, the molecular mechanisms leading to the loss of transcriptional fidelity are so far elusive, as are ways of preventing it. Here we profiled and analysed genome-wide, ageing-related changes in transcriptional processes across different organisms: nematodes, fruitflies, mice, rats and humans. The average transcriptional elongation speed (RNA polymerase II speed) increased with age in all five species. Along with these changes in elongation speed, we observed changes in splicing, including a reduction of unspliced transcripts and the formation of more circular RNAs. Two lifespan-extending interventions, dietary restriction and lowered insulin-IGF signalling, both reversed most of these ageing-related changes. Genetic variants in RNA polymerase II that reduced its speed in worms5 and flies6 increased their lifespan. Similarly, reducing the speed of RNA polymerase II by overexpressing histone components, to counter age-associated changes in nucleosome positioning, also extended lifespan in flies and the division potential of human cells. Our findings uncover fundamental molecular mechanisms underlying animal ageing and lifespan-extending interventions, and point to possible preventive measures.

Effect of Dl-3-n-butylphthalide on mitochondrial Cox7c in models of cerebral ischemia/reperfusion injury.

Several studies have demonstrated the protective effect of dl-3-n-Butylphthalide (NBP) against cerebral ischemia, which may be related to the attenuation of mitochondrial dysfunction. However, the specific mechanism and targets of NBP in cerebral ischemia/reperfusion remains unclear. In this study, we used a chemical proteomics approach to search for targets of NBP and identified cytochrome C oxidase 7c (Cox7c) as a key interacting target of NBP. Our findings indicated that NBP inhibits mitochondrial apoptosis and reactive oxygen species (ROS) release and increases ATP production through upregulation of Cox7c. Subsequently, mitochondrial respiratory capacity was improved and the HIF-1α/VEGF pathway was upregulated, which contributed to the maintenance of mitochondrial membrane potential and blood brain barrier integrity and promoting angiogenesis. Therefore, our findings provided a novel insight into the mechanisms underlying the neuroprotective effects of NBP, and also proposed for the first time that Cox7c exerts a critical role by protecting mitochondrial function.

The First-Water-Layer Evolution at the Graphene/Water Interface under Different Electro-Modulated Hydrophilic Conditions Observed by Suspended/Supported Field-Effect-Device Architectures.

Interfacial water molecules affect carrier transportation within graphene and related applications. Without proper tools, however, most of the previous works focus on simulation modeling rather than experimental validation. To overcome this obstacle, a series of graphene field-effect transistors (GFETs) with suspended (substrate-free, SF) and supported (oxide-supported, OS) configurations are developed to investigate the graphene-water interface under different hydrophilic conditions. With deionized water environments, in our experiments, the electrical transportation behaviors of the graphene mainly originate from the evolution of the interfacial water-molecule arrangement. Also, these current-voltage behaviors can be used to elucidate the first-water layer at the graphene-water interface. For SF-GFET, our experimental results show positive hysteresis in electrical transportation. These imply highly ordered interfacial water molecules with a separated-ionic distributed structure. For OS-GFET, on the contrary, the negative hysteresis shows the formation of the hydrogen-bond interaction between the interfacial water layer and the SiO2 substrate under the graphene. This interaction further promotes current conduction through the graphene/water interface. In addition, the net current-voltage relationship also indicates the energy required to change the orientation of the first-layer water molecules during electro-potential change. Therefore, our work gives an insight into graphene-water interfacial evolution with field-effect modulation. Furthermore, this experimental architecture also paves the way for investigating 2D solid-liquid interfacial features.

Microcirculation No-Reflow Phenomenon after Acute Ischemic Stroke.

BACKGROUND: The no-reflow phenomenon refers to a failure to restore normal cerebral microcirculation despite brain large artery recanalization after acute ischemic stroke, which was observed over 50 years ago. SUMMARY: Different mechanisms contributing to no-reflow extend across the endovascular, vascular wall, and extravascular factors. There are some clinical tools to evaluate cerebral microvascular hemodynamics and represent biomarkers of the no-reflow phenomenon. As substantial experimental and clinical data showed that clinical outcome was better correlated with reperfusion status rather than recanalization in patients with ischemic stroke, how to address the no-reflow phenomenon is critical. But effective treatments for restoring cerebral microcirculation have not been well established until now, so there is an urgent need for novel therapeutic perspectives to improve outcomes after recanalization therapies. CONCLUSION: Here, we review the occurrence of the no-reflow phenomenon after ischemic stroke and discuss its impact, detection method, and therapeutic strategies on the course of ischemic stroke, from basic science to clinical findings.

Improving Biomedical Question Answering by Data Augmentation and Model Weighting.

Biomedical Question Answering aims to extract an answer to the given question from a biomedical context. Due to the strong professionalism of specific domain, it's more difficult to build large-scale datasets for specific domain question answering. Existing methods are limited by the lack of training data, and the performance is not as good as in open-domain settings, especially degrading when facing to the adversarial sample. We try to resolve the above issues. First, effective data augmentation strategies are adopted to improve the model training, including slide window, summarization and round-trip translation. Second, we propose a model weighting strategy for the final answer prediction in biomedical domain, which combines the advantage of two models, open-domain model QANet and BioBERT pre-trained in biomedical domain data. Finally, we give adversarial training to reinforce the robustness of the model. The public biomedical dataset collected from PubMed provided by BioASQ challenge is used to evaluate our approach. The results show that the model performance has been improved significantly compared to the single model and other models participated in BioASQ challenge. It can learn richer semantic expression from data augmentation and adversarial samples, which is beneficial to solve more complex question answering problems in biomedical domain.

Modality-Specific Segmentation Network for Lung Tumor Segmentation in PET-CT Images.

Lung tumor segmentation in PET-CT images plays an important role to assist physicians in clinical application to accurately diagnose and treat lung cancer. However, it is still a challenging task in medical image processing field. Due to respiration and movement, the lung tumor varies largely in PET images and CT images. Even the two images are almost simultaneously collected and registered, the shape and size of lung tumors in PET-CT images are different from each other. To address these issues, a modality-specific segmentation network (MoSNet) is proposed for lung tumor segmentation in PET-CT images. MoSNet can simultaneously segment the modality-specific lung tumor in PET images and CT images. MoSNet learns a modality-specific representation to describe the inconsistency between PET images and CT images and a modality-fused representation to encode the common feature of lung tumor in PET images and CT images. An adversarial method is proposed to minimize an approximate modality discrepancy through an adversarial objective with respect to a modality discriminator and reserve modality-common representation. This improves the representation power of the network for modality-specific lung tumor segmentation in PET images and CT images. The novelty of MoSNet is its ability to produce a modality-specific map that explicitly quantifies the modality-specific weights for the features in each modality. To demonstrate the superiority of our method, MoSNet is validated in 126 PET-CT images with NSCLC. Experimental results show that MoSNet outperforms state-of-the-art lung tumor segmentation methods.