Spatial and geometric learning for classification of breast tumors from multi-center ultrasound images: a hybrid learning approach.

BACKGROUND: Breast cancer is the most common cancer among women, and ultrasound is a usual tool for early screening. Nowadays, deep learning technique is applied as an auxiliary tool to provide the predictive results for doctors to decide whether to make further examinations or treatments. This study aimed to develop a hybrid learning approach for breast ultrasound classification by extracting more potential features from local and multi-center ultrasound data. METHODS: We proposed a hybrid learning approach to classify the breast tumors into benign and malignant. Three multi-center datasets (BUSI, BUS, OASBUD) were used to pretrain a model by federated learning, then every dataset was fine-tuned at local. The proposed model consisted of a convolutional neural network (CNN) and a graph neural network (GNN), aiming to extract features from images at a spatial level and from graphs at a geometric level. The input images are small-sized and free from pixel-level labels, and the input graphs are generated automatically in an unsupervised manner, which saves the costs of labor and memory space. RESULTS: The classification AUCROC of our proposed method is 0.911, 0.871 and 0.767 for BUSI, BUS and OASBUD. The balanced accuracy is 87.6%, 85.2% and 61.4% respectively. The results show that our method outperforms conventional methods. CONCLUSIONS: Our hybrid approach can learn the inter-feature among multi-center data and the intra-feature of local data. It shows potential in aiding doctors for breast tumor classification in ultrasound at an early stage.

Development and evaluation of machine learning models and nomogram for the prediction of severe acute pancreatitis.

BACKGROUND AND AIM: Severe acute pancreatitis (SAP) in patients progresses rapidly and can cause multiple organ failures associated with high mortality. We aimed to train a machine learning (ML) model and establish a nomogram that could identify SAP, early in the course of acute pancreatitis (AP). METHODS: In this retrospective study, 631 patients with AP were enrolled in the training cohort. For predicting SAP early, five supervised ML models were employed, such as random forest (RF), K-nearest neighbors (KNN), and naive Bayes (NB), which were evaluated by accuracy (ACC) and the areas under the receiver operating characteristic curve (AUC). The nomogram was established, and the predictive ability was assessed by the calibration curve and AUC. They were externally validated by an independent cohort of 109 patients with AP. RESULTS: In the training cohort, the AUC of RF, KNN, and NB models were 0.969, 0.954, and 0.951, respectively, while the AUC of the Bedside Index for Severity in Acute Pancreatitis (BISAP), Ranson and Glasgow scores were only 0.796, 0.847, and 0.837, respectively. In the validation cohort, the RF model also showed the highest AUC, which was 0.961. The AUC for the nomogram was 0.888 and 0.955 in the training and validation cohort, respectively. CONCLUSIONS: Our findings suggested that the RF model exhibited the best predictive performance, and the nomogram provided a visual scoring model for clinical practice. Our models may serve as practical tools for facilitating personalized treatment options and improving clinical outcomes through pre-treatment stratification of patients with AP.

MTANS: Multi-Scale Mean Teacher Combined Adversarial Network with Shape-Aware Embedding for Semi-Supervised Brain Lesion Segmentation.

The annotation of brain lesion images is a key step in clinical diagnosis and treatment of a wide spectrum of brain diseases. In recent years, segmentation methods based on deep learning have gained unprecedented popularity, leveraging a large amount of data with high-quality voxel-level annotations. However, due to the limited time clinicians can provide for the cumbersome task of manual image segmentation, semi-supervised medical image segmentation methods present an alternative solution as they require only a few labeled samples for training. In this paper, we propose a novel semi-supervised segmentation framework that combines improved mean teacher and adversarial network. Specifically, our framework consists of (i) a student model and a teacher model for segmenting the target and generating the signed distance maps of object surfaces, and (ii) a discriminator network for extracting hierarchical features and distinguishing the signed distance maps of labeled and unlabeled data. Besides, based on two different adversarial learning processes, a multi-scale feature consistency loss derived from the student and teacher models is proposed, and a shape-aware embedding scheme is integrated into our framework. We evaluated the proposed method on the public brain lesion datasets from ISBI 2015, ISLES 2015, and BRATS 2018 for the multiple sclerosis lesion, ischemic stroke lesion, and brain tumor segmentation respectively. Experiments demonstrate that our method can effectively leverage unlabeled data while outperforming the supervised baseline and other state-of-the-art semi-supervised methods trained with the same labeled data. The proposed framework is suitable for joint training of limited labeled data and additional unlabeled data, which is expected to reduce the effort of obtaining annotated images.

Fabrication of Sub-Micrometer-Sized MoS2 Thin-Film Transistor by Phase Mode AFM Lithography.

The phase mode atomic force microscopy (AFM) lithography and monolayer lift-off process are combined to fabricate electronics based on 2D materials (2DMs), which remove the need for pre-fabricating markers and increase the accuracy of the overlay and alignment. The promising phase mode of AFM lithography eliminates the drawbacks of the conventional force mode such as the over-cut, under-cut, debris effect, and severe tip wear. The planar size of MoS2 thin-film transistors is shrunken down to sub-micrometer by the proposed method, and the fabricated devices demonstrate n-type characteristics. It offers a more flexible and easier way to fabricate prototypes of sub-micrometer-sized 2DMs based devices, and gives the opportunity to explore the size effect on the performance of 2DMs devices.

The co-design of interface sensing and tailoring of ultra-thin film with ultrasonic vibration-assisted AFM system.

Ultra-thin films (e.g., graphene, MoS2, and black phosphorus) have shown amazing performance in a variety of applications. The tailoring or machining of these ultra-thin films is often the preliminary step to manufacturing them into functional devices. Atomic force microscopy (AFM) is a flexible, high-efficiency and low-cost tailoring or machining tool with the advantages of high resolution and precision. However, the current AFM-based tailoring methods are often set up as an open loop regarding the machined depth and state. Thus, because of a lack of real-time feedback, an inappropriate applied force leads to over-cutting or under-cutting, which limits the performance of the manufactured devices. In this study, we propose a real-time tailoring and sensing method based on an ultrasonic vibration-assisted (USV-assisted) AFM system to solve the above problems. With the proposed method, the machined depth and state can be sensed in real time by detecting the phase value of the vibrating cantilever. To characterize and gain insight into the phase responses of the cantilever to the machined depth and sample material, a theoretical dynamic model of a cantilever-film vibrating system is introduced to model the machining process, and a sensing theory of machined depth and state is developed based on a USV-assisted AFM system. The experimental results verify the feasibility and effectiveness of the proposed method, which in turn lay the foundation for a closed-loop tailoring control strategy for ultra-thin films.

A rapid and automated relocation method of an AFM probe for high-resolution imaging.

The atomic force microscope (AFM) is one of the most powerful tools for high-resolution imaging and high-precision positioning for nanomanipulation. The selection of the scanning area of the AFM depends on the use of the optical microscope. However, the resolution of an optical microscope is generally no larger than 200 nm owing to wavelength limitations of visible light. Taking into consideration the two determinants of relocation-relative angular rotation and positional offset between the AFM probe and nano target-it is therefore extremely challenging to precisely relocate the AFM probe to the initial scan/manipulation area for the same nano target after the AFM probe has been replaced, or after the sample has been moved. In this paper, we investigate a rapid automated relocation method for the nano target of an AFM using a coordinate transformation. The relocation process is both simple and rapid; moreover, multiple nano targets can be relocated by only identifying a pair of reference points. It possesses a centimeter-scale location range and nano-scale precision. The main advantages of this method are that it overcomes the limitations associated with the resolution of optical microscopes, and that it is label-free on the target areas, which means that it does not require the use of special artificial markers on the target sample areas. Relocation experiments using nanospheres, DNA, SWCNTs, and nano patterns amply demonstrate the practicality and efficiency of the proposed method, which provides technical support for mass nanomanipulation and detection based on AFM for multiple nano targets that are widely distributed in a large area.

Experimental study and modeling of atomic-scale friction in zigzag and armchair lattice orientations of MoS2.

Physical properties of two-dimensional materials, such as graphene, black phosphorus, molybdenum disulfide (MoS2) and tungsten disulfide, exhibit significant dependence on their lattice orientations, especially for zigzag and armchair lattice orientations. Understanding of the atomic probe motion on surfaces with different orientations helps in the study of anisotropic materials. Unfortunately, there is no comprehensive model that can describe the probe motion mechanism. In this paper, we report a tribological study of MoS2 in zigzag and armchair orientations. We observed a characteristic power spectrum and friction force values. To explain our results, we developed a modified, two-dimensional, stick-slip Tomlinson model that allows simulation of the probe motion on MoS2 surfaces by combining the motion in the Mo layer and S layer. Our model fits well with the experimental data and provides a theoretical basis for tribological studies of two-dimensional materials.

Advance on combination therapy strategies based on biomedical nanotechnology induced ferroptosis for cancer therapeutics.

Globally, cancer is a serious health problem. It is unfortunate that current anti-cancer strategies are insufficiently specific and damage the normal tissues. There's urgent need for development of new anti-cancer strategies. More recently, increasing attention has been paid to the new application of ferroptosis and nano materials in cancer research. Ferroptosis, a condition characterized by excessive reactive oxygen species-induced lipid peroxidation, as a new programmed cell death mode, exists in the process of a number of diseases, including cancers, neurodegenerative disease, cerebral hemorrhage, liver disease, and renal failure. There is growing evidence that inducing ferroptosis has proven to be an effective strategy against a variety of chemo-resistant cancer cells. Nano-drug delivery system based on nanotechnology provides a highly promising platform with the benefits of precise control of drug release and reduced toxicity and side effects. This paper reviews the latest advances of combination therapy strategies based on biomedical nanotechnology induced ferroptosis for cancer therapeutics. Given the new chances and challenges in this emerging area, we need more attention to the combination of nanotechnology and ferroptosis in the treatment of cancer in the future.

A survey of label-noise deep learning for medical image analysis.

Several factors are associated with the success of deep learning. One of the most important reasons is the availability of large-scale datasets with clean annotations. However, obtaining datasets with accurate labels in the medical imaging domain is challenging. The reliability and consistency of medical labeling are some of these issues, and low-quality annotations with label noise usually exist. Because noisy labels reduce the generalization performance of deep neural networks, learning with noisy labels is becoming an essential task in medical image analysis. Literature on this topic has expanded in terms of volume and scope. However, no recent surveys have collected and organized this knowledge, impeding the ability of researchers and practitioners to utilize it. In this work, we presented an up-to-date survey of label-noise learning for medical image domain. We reviewed extensive literature, illustrated some typical methods, and showed unified taxonomies in terms of methodological differences. Subsequently, we conducted the methodological comparison and demonstrated the corresponding advantages and disadvantages. Finally, we discussed new research directions based on the characteristics of medical images. Our survey aims to provide researchers and practitioners with a solid understanding of existing medical label-noise learning, such as the main algorithms developed over the past few years, which could help them investigate new methods to combat with the negative effects of label noise.

An important diagnostic marker of acute myocardial infarction patients: Plasma miRNA133 levels.

The objective of this study was to explore changes in miRNA133 levels as a basis for clinical diagnostic markers in patients with acute myocardial infarction (AMI). A total of 100 chest pain patient cases admitted to a hospital from June 2021 to December 2022 were used. The study involved the selection of 50 patients: 25 patients with unstable undetermined heart pain and 25 healthy subjects were included in the control group of 50 patients with non-AMI patients. Meanwhile, 50 patients with AMI were designated as the experimental group. Changes in miRNA133 levels in patients' plasma were analyzed for expression using quantitative fluorescence analysis. When the serum TPI, plasma NT-ProBNP, glycosylated hemoglobin, and plasma D-dimer index values were compared between the control and experimental groups, there was a statistically significant difference (P < .05). mi-RNA-133 had a mean plasma level value of 2.60 ±â€…1.01, the mean level value of mi-RNA-133 in patients with non-AMI was 1.34 ±â€…1.18, and the patients in the AMI group showed significantly high values of the mean plasma level of mi-RNA-133. The relative expression level value of cTnl in patients with AMI was 10.84 ±â€…12.64. Of the specificity and sensitivity diagnostics, mi-RNA-133 had the best diagnostic effect. The area under mi-RNA-133 in the regression curve was 95.4%, the specificity of the whole combination of indicators was 89.4% and the sensitivity was 100%. Finally, the correlation between mi-RNA-133 and white blood cell count (WBC) and TG was statistically significant (P < .05). In conclusion, changes in the level of mi-RNA-133 may be an important marker for diagnosing the status of patients with AMI, while a faster and more accurate method will emerge along with the improvement of the detection technology, and at the same time, due to the variability of the study cases and other limitations, further research will be carried out subsequently.

Motion-artifact-augmented pseudo-label network for semi-supervised brain tumor segmentation.

MRI image segmentation is widely used in clinical practice as a prerequisite and a key for diagnosing brain tumors. The quest for an accurate automated segmentation method for brain tumor images, aiming to ease clinical doctors' workload, has gained significant attention as a research focal point. Despite the success of fully supervised methods in brain tumor segmentation, challenges remain. Due to the high cost involved in annotating medical images, the dataset available for training fully supervised methods is very limited. Additionally, medical images are prone to noise and motion artifacts, negatively impacting quality. In this work, we propose MAPSS, a motion-artifact-augmented pseudo-label network for semi-supervised segmentation. Our method combines motion artifact data augmentation with the pseudo-label semi-supervised training framework. We conduct several experiments under different semi-supervised settings on a publicly available dataset BraTS2020 for brain tumor segmentation. The experimental results show that MAPSS achieves accurate brain tumor segmentation with only a small amount of labeled data and maintains robustness in motion-artifact-influenced images. We also assess the generalization performance of MAPSS using the Left Atrium dataset. Our algorithm is of great significance for assisting doctors in formulating treatment plans and improving treatment quality.

An AFM-Based Model-Fitting-Free Viscoelasticity Characterization Method for Accurate Grading of Primary Prostate Tumor.

Viscoelasticity is a crucial property of cells, which plays an important role in label-free cell characterization. This paper reports a model-fitting-free viscoelasticity calculation method, correcting the effects of frequency, surface adhesion and liquid resistance on AFM force-distance (FD) curves. As demonstrated by quantifying the viscosity and elastic modulus of PC-3 cells, this method shows high self-consistency and little dependence on experimental parameters such as loading frequency, and loading mode (Force-volume vs. PeakForce Tapping). The rapid calculating speed of less than 1ms per curve without the need for a model fitting process is another advantage. Furthermore, this method was utilized to characterize the viscoelastic properties of primary clinical prostate cells from 38 patients. The results demonstrate that the reported characterization method a comparable performance with the Gleason Score system in grading prostate cancer cells, This method achieves a high average accuracy of 97.6% in distinguishing low-risk prostate tumors (BPH and GS6) from higher-risk (GS7-GS10) prostate tumors and a high average accuracy of 93.3% in distinguishing BPH from prostate cancer.

Understanding secondary particles in a regional site of Yangtze River Delta: Insights from mass spectrometric measurement.

Submicron particulate matter (PM1) poses significant risks to health risks and global climate. In this study, secondary organic aerosols (SOA) and inorganic compositions were examined for their physicochemical characteristics and evolution using high-resolution aerosol instruments in Changzhou over one-month period. The results showed that transport accompanied by regional static conditions leaded to the occurrence of heavy pollution. In addition, regional generation and local emissions also leaded to the occurrence of light and moderate pollution during the observation period in Changzhou. Organic aerosols (OA) and nitrate (NO3-) accounted for 45 % and 23 % of PM1, respectively. The increase in PM1 was dominated by the contribution of NO3- and OA. SOA was dominance in OA (63 % with 40 % MO-OOA), which was higher than primary organic aerosols (POA). Besides, photochemical reactions and the high oxidizing nature of the urban atmosphere promoted the production of OA, especially MO-OOA in Changzhou. Our results highlight that secondary particles contribute significantly to PM pollution in Changzhou, underlining the importance of controlling emissions of gaseous precursors, especially under high oxidation conditions.

Bionic Scotopic Adaptation Transistors for Nighttime Low Illumination Imaging.

Human vision excels in perceiving nighttime low illumination due to biological feedforward adaptation. Replicating this ability in biomimetic vision using solid-state devices has been highly sought after. However, emulating scotopic adaptation, entailing a confluence of efficient photoexcitation and dynamic carrier modulation, presents formidable challenges. Here, we demonstrate a low-power and bionic scotopic adaptation transistor by coupling a light-absorption layer and an electron-trapping layer at the bottom of the semiconducting channel, enabling simultaneous achievement of efficient generation of free photocarriers and adaptive carrier accumulation within a single device. This innovation empowers our transistor to exhibit sensitivity-potentiated characteristics after adaptation, detecting scotopic-level illumination (0.001 lx) with exceptional photosensitivity up to 103 at low voltages below 2 V. Moreover, we have successfully replicated diverse scotopic vision functions, encompassing time-dependent visual threshold enhancement, light intensity-dependent adaptation index, imaging contrast enhancement for nighttime low illumination imaging, opening an opportunity for artificial night vision.

Effective and mechanistic insights into shale gas wastewater reverse osmosis concentrate treatment using ozonation-biological activated carbon process.

Reverse osmosis (RO) plays a pivotal role in shale gas wastewater resource utilization. However, managing the reverse osmosis concentrate (ROC) characterized by high salinity and increased concentrations of organic matter is challenging. In this study, we aimed to elucidate the enhancement effects and mechanisms of pre-ozonation on organic matter removal efficacy in ROC using a biological activated carbon (BAC) system. Our findings revealed that during the stable operation phase, the ozonation (O3 and O3/granular activated carbon)-BAC system removes 43.6-72.2 % of dissolved organic carbon, achieving a 4-7 fold increase in efficiency compared with that in the BAC system alone. Through dynamic analysis of influent and effluent water quality, biofilm performance, and microbial community structure, succession, and function prediction, we elucidated the following primary enhancement mechanisms: 1) pre-ozonation significantly enhances the biodegradability of ROC by 4.5-6 times and diminishes the organic load on the BAC system; 2) pre-ozonation facilitates the selective enrichment of microbes capable of degrading organic compounds in the BAC system, thereby enhancing the biodegradation capacity and stability of the microbial community; and 3) pre-ozonation accelerates the regeneration rate of the granular activated carbon adsorption sites. Collectively, our findings provide valuable insights into treating ROC through pre-oxidation combined with biotreatment.

Breaking Fundamental Limitation of Flow-Induced Anisotropic Growth for Large-Scale and Fast Printing of Organic Single-Crystal Films.

Advanced organic electronic technologies have put forward a pressing demand for cost-effective and high-throughput fabrication of organic single-crystal films (OSCFs). However, solution-printed OSCFs are typically plagued by the existence of abundant structural defects, which pose a formidable challenge to achieving large-scale and high-performance organic electronics. Here, it is elucidated that these structural defects are mainly originated from printing flow-induced anisotropic growth, an important factor that is overlooked for too long. In light of this, a surfactant-additive printing method is proposed to effectively overcome the anisotropic growth, enabling the deposition of uniform OSCFs over the wafer scale at a high speed of 1.2 mm s-1 at room temperature. The resulting OSCF exhibits appealing performance with a high average mobility up to 10.7 cm2 V-1 s-1, which is one of the highest values for flexible organic field-effect transistor arrays. Moreover, large-scale OSCF-based flexible logic circuits, which can be bent without degradation to a radius as small as 4.0 mm and over 1000 cycles are realized. The work provides profound insights into breaking the limitation of flow-induced anisotropic growth and opens new avenues for printing large-scale organic single-crystal electronics.

Identification of key biomarkers based on the proliferation of secondary hyperparathyroidism by bioinformatics analysis and machine learning.

Objective: Secondary hyperparathyroidism (SHPT) is a frequent complication of chronic kidney disease (CKD) associated with morbidity and mortality. This study aims to identify potential biomarkers that may be used to predict the progression of SHPT and to elucidate the molecular mechanisms of SHPT pathogenesis at the transcriptome level. Methods: We analyzed differentially expressed genes (DEGs) between diffuse and nodular parathyroid hyperplasia of SHPT patients from the GSE75886 dataset, and then verified DEG levels with the GSE83421 data file of primary hyperparathyroidism (PHPT) patients. Candidate gene sets were selected by machine learning screens of differential genes and immune cell infiltration was explored with the CIBERSORT algorithm. RcisTarget was used to predict transcription factors, and Cytoscape was used to construct a lncRNA-miRNA-mRNA network to identify possible molecular mechanisms. Immunohistochemistry (IHC) staining and quantitative real-time polymerase chain reaction (qRT-PCR) were used to verify the expression of screened genes in parathyroid tissues of SHPT patients and animal models. Results: A total of 614 DEGs in GSE75886 were obtained as candidate gene sets for further analysis. Five key genes (USP12, CIDEA, PCOLCE2, CAPZA1, and ACCN2) had significant expression differences between groups and were screened with the best ranking in the machine learning process. These genes were shown to be closely related to immune cell infiltration levels and play important roles in the immune microenvironment. Transcription factor ZBTB6 was identified as the master regulator, alongside multiple other transcription factors. Combined with qPCR and IHC assay of hyperplastic parathyroid tissues from SHPT patients and rats confirm differential expression of USP12, CIDEA, PCOLCE2, CAPZA1, and ACCN2, suggesting that they may play important roles in the proliferation and progression of SHPT. Conclusion: USP12, CIDEA, PCOLCE2, CAPZA1, and ACCN2 have great potential both as biomarkers and as therapeutic targets in the proliferation of SHPT. These findings suggest novel potential targets and future directions for SHPT research.

Emotional involvement matters, too: Associations among parental involvement, time management and academic engagement vary with Youth's developmental phase.

BACKGROUND: Prior studies have emphasized the importance of parents' educational involvement (a type of cognitive involvement) to academic engagement, although little is known about emotional involvement. AIMS: This study investigated whether and how different facets of involvement (cognitive vs. emotional, paternal vs. maternal) are differentially related to academic engagement and whether and how the associations among parental involvement, time management and academic engagement vary by adolescents' developmental phases. SAMPLES: The participants of this large national survey were students in elementary, middle and high school across different regions of mainland China. A total of 2687 adolescents (52.7% females, Mage = 14.07 ± 2.47) participated in this study. METHODS: Structural equation models and multigroup analysis were conducted. RESULTS: We found that the total effect of paternal and maternal emotional involvement on academic engagement was positive in elementary-, middle- and high school students, and an indirect effect of time management underlying the above paths was found in all three groups. In contrast, the positive effect of maternal cognitive involvement on academic engagement as well as the indirect effects underlying the above pathways was established only in high school students. CONCLUSIONS: The findings highlight the necessity of parents' emotional involvement and the consideration of adolescent developmental characteristics in the design of interventions.

ErbB4 promotes M2 activation of macrophages in idiopathic pulmonary fibrosis.

Idiopathic pulmonary fibrosis (IPF) is the most common and fatal diffuse fibrotic lung disease accompanied by macrophage M2 activation. ErbB4 is involved in and affects the process of inflammation. In this study, we determined that the mRNA level and protein expression of ErbB4 and M2 cytokine members were increased in the serum of IPF patients. In mouse alveolar macrophage MH-S cells, after knocking down ErbB4 by siRNA, the mRNA level and protein expression of M2 activator induced by interleukin (IL)-4 were decreased compared with the control group. Activating by ErbB4 agonist neuromodulatory protein (NRG)-1, IL-4-induced M2 program was promoted. Mechanistically, treated with NRG-1 in MH-S cells, the phosphorylation level of Akt did not change, while the phosphorylation level of ERK increased. Using SCH772984 to inhibit ERK pathway, the increasing IL-4-induced M2 activation by NRG-1 was inhibited, and the high level of M2 activator protein expression and mRNA expression was restored. Collectively, our data support that ErbB4 and M2 programs are implicated in IPF, and ErbB4 participates in the regulation of M2 activation induced by IL-4 through the ERK pathway.

Efficient inactivation of harmful algae K. mikimotoi by a novel algicidal bacterium via a rare direct contact pathway: Performances and mechanisms.

Harmful algal blooms (HABs) caused by Karenia mikimotoi have posed great threats to marine ecosystems, and algal inactivation by symbiotic bacteria has been recognized as environmental benign methods for controlling HABs. However, the identified algicidal bacteria for K. mikimotoi is limited and exclusively based on indirect algicidal pathways, which may cause secondary pollution due to releasing toxic algicidal agents. In this study, a novel strain of algicidal bacteria Tenacibaculum sp. GD3 was isolated from the phycosphere of K. mikimotoi. The bacterial strain GD3 could achieve 92.6 % of inactivation efficiency against K. mikimotoi within 8 h of co-culturing period, which outperformed those in existing literatures reported so far. The algicidal mechanisms were revealed to be a rare direct cell-to-cell contact pathway, and the GD3 could grow by utilizing metabolites from K. mikimotoi, exhibiting excellent bacterial adaptability in the phycosphere. Cell morphology changes were monitored by live cell imaging system combined with SEM and TEM observations, which showed that the GD3 was first attached to the algal cell membrane, followed by lipid peroxidation and lysis of membrane protein. Oxidative stress responses were induced as reveled by up-regulation of intracellular ROSs and antioxidant enzyme activity. Photosynthetic parameters including rETRmax, Fv/Fm, YII and NPQ were reduced, and expression of functional genes involved in decomposition of chlorophyll and cell wall was significantly suppressed. Moreover, the intracellular release profile and acute toxicity assessment indicated that the GD3 could also detoxify the K. mikimotoi cultures and the released biomolecules would not cause adverse effect to marine environment. This study not only provides a novel algicidal bacterium against K. mikimotoi via a rare direct mode, but also helps to better understand the algicidal mechanisms at physiological and genetic level, thus moving forward the areas of HABs control by microbiological strategies.