UKSSL: Underlying Knowledge Based Semi-Supervised Learning for Medical Image Classification.

Goal: Deep learning techniques have made significant progress in medical image analysis. However, obtaining ground truth labels for unlabeled medical images is challenging as they often outnumber labeled images. Thus, training a high-performance model with limited labeled data has become a crucial challenge. Methods: This study introduces an underlying knowledge-based semi-supervised framework called UKSSL, consisting of two components: MedCLR extracts feature representations from the unlabeled dataset; UKMLP utilizes the representation and fine-tunes it with the limited labeled dataset to classify the medical images. Results: UKSSL evaluates on the LC25000 and BCCD datasets, using only 50% labeled data. It gets precision, recall, F1-score, and accuracy of 98.9% on LC25000 and 94.3%, 94.5%, 94.3%, and 94.1% on BCCD, respectively. These results outperform other supervised-learning methods using 100% labeled data. Conclusions: The UKSSL can efficiently extract underlying knowledge from the unlabeled dataset and perform better using limited labeled medical images.

Guest Editorial Introduction to the Special Section on Weakly-Supervised Deep Learning and Its Applications.

Researchers in biomedical engineering are increasingly turning to weakly-supervised deep learning (WSDL) techniques [1] to tackle challenges in biomedical data analysis, which often involves noisy, limited, or imprecise expert annotations [2]. WSDL methods have emerged as a solution to alleviate the manual annotation burden for structured biomedical data like signals, images, and videos [3] while enabling deep neural network models to learn from larger-scale datasets at a reduced annotation cost. With the proliferation of advanced deep learning techniques such as generative adversarial networks (GANs), graph neural networks (GNNs) [4], vision transformers (ViTs) [5], and deep reinforcement learning (DRL) models [6], research endeavors are focused on solving WSDL problems and applying these techniques to various biomedical analysis tasks.

MACFNet: Detection of Alzheimer's disease via multiscale attention and cross-enhancement fusion network.

BACKGROUND AND OBJECTIVE: Alzheimer's disease (AD) is a dreaded degenerative disease that results in a profound decline in human cognition and memory. Due to its intricate pathogenesis and the lack of effective therapeutic interventions, early diagnosis plays a paramount role in AD. Recent research based on neuroimaging has shown that the application of deep learning methods by multimodal neural images can effectively detect AD. However, these methods only concatenate and fuse the high-level features extracted from different modalities, ignoring the fusion and interaction of low-level features across modalities. It consequently leads to unsatisfactory classification performance. METHOD: In this paper, we propose a novel multi-scale attention and cross-enhanced fusion network, MACFNet, which enables the interaction of multi-stage low-level features between inputs to learn shared feature representations. We first construct a novel Cross-Enhanced Fusion Module (CEFM), which fuses low-level features from different modalities through a multi-stage cross-structure. In addition, an Efficient Spatial Channel Attention (ECSA) module is proposed, which is able to focus on important AD-related features in images more efficiently and achieve feature enhancement from different modalities through two-stage residual concatenation. Finally, we also propose a multiscale attention guiding block (MSAG) based on dilated convolution, which can obtain rich receptive fields without increasing model parameters and computation, and effectively improve the efficiency of multiscale feature extraction. RESULTS: Experiments on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset demonstrate that our MACFNet has better classification performance than existing multimodal methods, with classification accuracies of 99.59 %, 98.85 %, 99.61 %, and 98.23 % for AD vs. CN, AD vs. MCI, CN vs. MCI and AD vs. CN vs. MCI, respectively, and specificity of 98.92 %, 97.07 %, 99.58 % and 99.04 %, and sensitivity of 99.91 %, 99.89 %, 99.63 % and 97.75 %, respectively. CONCLUSIONS: The proposed MACFNet is a high-accuracy multimodal AD diagnostic framework. Through the cross mechanism and efficient attention, MACFNet can make full use of the low-level features of different modal medical images and effectively pay attention to the local and global information of the images. This work provides a valuable reference for multi-mode AD diagnosis.

Improving iconic memory through contrast detection training with HOA-corrected vision.

Iconic memory and short-term memory are not only crucial for perception and cognition, but also of great importance to mental health. Here, we first showed that both types of memory could be improved by improving limiting processes in visual processing through perceptual learning. Normal adults were trained in a contrast detection task for ten days, with their higher-order aberrations (HOA) corrected in real-time. We found that the training improved not only their contrast sensitivity function (CSF), but also their iconic memory and baseline information maintenance for short-term memory, and the relationship between memory and CSF improvements could be well-predicted by an observer model. These results suggest that training the limiting component of a cognitive task with visual perceptual learning could improve visual cognition. They may also provide an empirical foundation for new therapies to treat people with poor sensory memory.

Dissection of Common Rust Resistance in Tropical Maize Multiparent Population through GWAS and Linkage Studies.

Common rust (CR), caused by Puccina sorghi, is a major foliar disease in maize that leads to quality deterioration and yield losses. To dissect the genetic architecture of CR resistance in maize, this study utilized the susceptible temperate inbred line Ye107 as the male parent crossed with three resistant tropical maize inbred lines (CML312, D39, and Y32) to generate 627 F7 recombinant inbred lines (RILs), with the aim of identifying maize disease-resistant loci and candidate genes for common rust. Phenotypic data showed good segregation between resistance and susceptibility, with varying degrees of resistance observed across different subpopulations. Significant genotype effects and genotype × environment interactions were observed, with heritability ranging from 85.7% to 92.2%. Linkage and genome-wide association analyses across the three environments identified 20 QTLs and 62 significant SNPs. Among these, seven major QTLs explained 66% of the phenotypic variance. Comparison with six SNPs repeatedly identified across different environments revealed overlap between qRUST3-3 and Snp-203,116,453, and Snp-204,202,469. Haplotype analysis indicated two different haplotypes for CR resistance for both the SNPs. Based on LD decay plots, three co-located candidate genes, Zm00001d043536, Zm00001d043566, and Zm00001d043569, were identified within 20 kb upstream and downstream of these two SNPs. Zm00001d043536 regulates hormone regulation, Zm00001d043566 controls stomatal opening and closure, related to trichome, and Zm00001d043569 is associated with plant disease immune responses. Additionally, we performed candidate gene screening for five additional SNPs that were repeatedly detected across different environments, resulting in the identification of five candidate genes. These findings contribute to the development of genetic resources for common rust resistance in maize breeding programs.

ELRL-MD: a deep learning approach for myocarditis diagnosis using cardiac magnetic resonance images with ensemble and reinforcement learning integration.

Objective.Myocarditis poses a significant health risk, often precipitated by viral infections like coronavirus disease, and can lead to fatal cardiac complications. As a less invasive alternative to the standard diagnostic practice of endomyocardial biopsy, which is highly invasive and thus limited to severe cases, cardiac magnetic resonance (CMR) imaging offers a promising solution for detecting myocardial abnormalities.Approach.This study introduces a deep model called ELRL-MD that combines ensemble learning and reinforcement learning (RL) for effective myocarditis diagnosis from CMR images. The model begins with pre-training via the artificial bee colony (ABC) algorithm to enhance the starting point for learning. An array of convolutional neural networks (CNNs) then works in concert to extract and integrate features from CMR images for accurate diagnosis. Leveraging the Z-Alizadeh Sani myocarditis CMR dataset, the model employs RL to navigate the dataset's imbalance by conceptualizing diagnosis as a decision-making process.Main results.ELRL-DM demonstrates remarkable efficacy, surpassing other deep learning, conventional machine learning, and transfer learning models, achieving an F-measure of 88.2% and a geometric mean of 90.6%. Extensive experimentation helped pinpoint the optimal reward function settings and the perfect count of CNNs.Significance.The study addresses the primary technical challenge of inherent data imbalance in CMR imaging datasets and the risk of models converging on local optima due to suboptimal initial weight settings. Further analysis, leaving out ABC and RL components, confirmed their contributions to the model's overall performance, underscoring the effectiveness of addressing these critical technical challenges.

Identification of a novel marker and its associated laccase gene for regulating ear length in tropical and subtropical maize lines.

KEY MESSAGE: This study revealed the identification of a novel gene, Zm00001d042906, that regulates maize ear length by modulating lignin synthesis and reported a molecular marker for selecting maize lines with elongated ears. Maize ear length has garnered considerable attention due to its high correlation with yield. In this study, six maize inbred lines of significant importance in maize breeding were used as parents. The temperate maize inbred line Ye107, characterized by a short ear, was crossed with five tropical or subtropical inbred lines featuring longer ears, creating a multi-parent population displaying significant variations in ear length. Through genome-wide association studies and mutation analysis, the A/G variation at SNP_183573532 on chromosome 3 was identified as an effective site for discriminating long-ear maize. Furthermore, the associated gene Zm00001d042906 was found to correlate with maize ear length. Zm00001d042906 was functionally annotated as a laccase (Lac4), which showed activity and influenced lignin synthesis in the midsection cells of the cob, thereby regulating maize ear length. This study further reports a novel molecular marker and a new gene that can assist maize breeding programs in selecting varieties with elongated ears.

Exploring simple triplet representation learning.

Fully supervised learning methods necessitate a substantial volume of labelled training instances, a process that is typically both labour-intensive and costly. In the realm of medical image analysis, this issue is further amplified, as annotated medical images are considerably more scarce than their unlabelled counterparts. Consequently, leveraging unlabelled images to extract meaningful underlying knowledge presents a formidable challenge in medical image analysis. This paper introduces a simple triple-view unsupervised representation learning model (SimTrip) combined with a triple-view architecture and loss function, aiming to learn meaningful inherent knowledge efficiently from unlabelled data with small batch size. With the meaningful representation extracted from unlabelled data, our model demonstrates exemplary performance across two medical image datasets. It achieves this using only partial labels and outperforms other state-of-the-art methods. The method we present herein offers a novel paradigm for unsupervised representation learning, establishing a baseline that is poised to inspire the development of more intricate SimTrip-based methods across a spectrum of computer vision applications. Code and user guide are released at https://github.com/JerryRollingUp/SimTripSystem, the system also runs at http://43.131.9.159:5000/.

Deep learning-accelerated T2WI: image quality, efficiency, and staging performance against BLADE T2WI for gastric cancer.

PURPOSE: The purpose of our study is to investigate image quality, efficiency, and diagnostic performance of a deep learning-accelerated single-shot breath-hold (DLSB) against BLADE for T2-weighted MR imaging (T2WI) for gastric cancer (GC). METHODS: 112 patients with GCs undergoing gastric MRI were prospectively enrolled between Aug 2022 and Dec 2022. Axial DLSB-T2WI and BLADE-T2WI of stomach were scanned with same spatial resolution. Three radiologists independently evaluated the image qualities using a 5-scale Likert scales (IQS) in terms of lesion delineation, gastric wall boundary conspicuity, and overall image quality. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated in measurable lesions. T staging was conducted based on the results of both sequences for GC patients with gastrectomy. Pairwise comparisons between DLSB-T2WI and BLADE-T2WI were performed using the Wilcoxon signed-rank test, paired t-test, and chi-squared test. Kendall's W, Fleiss' Kappa, and intraclass correlation coefficient values were used to determine inter-reader reliability. RESULTS: Against BLADE, DLSB reduced total acquisition time of T2WI from 495 min (mean 4:42 per patient) to 33.6 min (18 s per patient), with better overall image quality that produced 9.43-fold, 8.00-fold, and 18.31-fold IQS upgrading against BALDE, respectively, in three readers. In 69 measurable lesions, DLSB-T2WI had higher mean SNR and higher CNR than BLADE-T2WI. Among 71 patients with gastrectomy, DLSB-T2WI resulted in comparable accuracy to BLADE-T2WI in staging GCs (P > 0.05). CONCLUSIONS: DLSB-T2WI demonstrated shorter acquisition time, better image quality, and comparable staging accuracy, which could be an alternative to BLADE-T2WI for gastric cancer imaging.

Comparing Business, Innovation, and Platform Ecosystems: A Systematic Review of the Literature.

In recent decades, the term "ecosystem" has garnered substantial attention in scholarly and managerial discourse, featuring prominently in academic and applied contexts. While individual scholars have made significant contributions to the study of various types of ecosystem, there appears to be a research gap marked by a lack of comprehensive synthesis and refinement of findings across diverse ecosystems. This paper systematically addresses this gap through a hybrid methodology, employing bibliometric and content analyses to systematically review the literature from 1993 to 2023. The primary research aim is to critically examine theoretical studies on different ecosystem types, specifically focusing on business, innovation, and platform ecosystems. The methodology of this study involves a content review of the identified literature, combining quantitative bibliometric analyses to differentiate patterns and content analysis for in-depth exploration. The core findings center on refining and summarizing the definitions of business, innovation, and platform ecosystems, shedding light on both commonalities and distinctions. Notably, the research unveils shared characteristics such as openness and diversity across these ecosystems while highlighting significant differences in terms of participants and objectives. Furthermore, the paper delves into the interconnections within these three ecosystem types, offering insights into their dynamics and paving the way for discussions on future research directions. This comprehensive examination not only advances our understanding of business, innovation, and platform ecosystems but also lays the groundwork for future scholarly inquiries in this dynamic and evolving field.

Flexible Grid Graphene Electrothermal Films for Real-Time Monitoring Applications.

Flexible electrothermal films are crucial for protecting equipment and systems in cold weather, such as ice blockages in natural gas pipelines and icing on aircraft wings. Therefore, a flexible electric heater is one of the essential devices in industrial operations. One of the main challenges is to develop flexible electrothermal films with low operating voltage, high steady-state temperature, and good mechanical stability. In this study, a flexible electrothermal film based on graphene-patterned structures was manufactured by combining the laser induction method and the transfer printing process. The grid structure design provides accurate real-time monitoring for the application of electrothermal films and shows potential in solving problems related to deicing and clearing ice blockages in pipelines. The flexible electrothermal film can reach a high heating temperature of 165 °C at 15 V and exhibits sufficient heating stability. By employing a simple and efficient method to create a flexible, high-performance electrothermal film, we provide a reliable solution for deicing and monitoring applications.

Asiatic acid and madecassic acid cause cardiotoxicity via inflammation and production of excessive reactive oxygen species in zebrafish.

Centella asiatica (L.) Urban is a famous Chinese traditional medicine, which is widely used for treating various chronic inflammatory diseases. Although there are reports that Centella total glycosides exhibit heart-protective properties, our previous experiment showed that it has cardiac toxic effects in zebrafish. The components of Centella total glycosides are complex, so we recommend further research to determine their key components and mechanisms. In this study, sample quantification was done using liquid chromatography-tandem mass spectrometry. The cardiotoxicity of Centella total glycosides, asiaticoside, madecassoside, asiatic acid, and madecassic acid was evaluated using zebrafish and cell models. The zebrafish oxidative stress model and myocarditis model were used to explore further the mechanisms through which cardiotoxicity is achieved. Asiatic acid and madecassic acid caused zebrafish cardiotoxicity and H9C2 cell death. However, no toxicity effects were observed for asiaticoside and madecassoside in zebrafish, until the solution was saturated. The results from the cell model study showed that asiatic acid and madecassic acid changed the expression of apoptosis-related genes in myocardial cells. In the zebrafish model, high concentrations of these components raised the levels of induced systemic inflammation, neutrophils gathered in the heart, and oxidative stress injury. Asiatic acid and madecassic acid are the main components causing cardiotoxicity in zebrafish. This may be due to enhanced inflammation and reactive oxygen species injury, which causes myocardial cell apoptosis, which further leads to cardiac toxicity.

Prognostic Evaluation for Hepatocellular Carcinoma with Portal Vein Tumor Thrombus Patients Treated with Transarterial Chemoembolization Plus Molecular Targeted Therapies-Development and Validation of the ABPS Score.

RATIONALE AND OBJECTIVES: Transarterial chemoembolization (TACE) plus molecular targeted therapies has emerged as the main approach for treating hepatocellular carcinoma (HCC) with portal vein tumor thrombus (PVTT). A robust model for outcome prediction and risk stratification of recommended TACE plus molecular targeted therapies candidates is lacking. We aimed to develop an easy-to-use tool specifically for these patients. METHODS: A retrospective analysis was conducted on 384 patients with HCC and PVTT who underwent TACE plus molecular targeted therapies at 16 different institutions. We developed and validated a new prognostic score which called ABPS score. Additionally, an external validation was performed on data from 200 patients enrolled in a prospective cohort study. RESULTS: The ABPS score (ranging from 0 to 3 scores), which involves only Albumin-bilirubin (ALBI, grade 1: 0 score; grade 2: 1 score), PVTT(I-II type: 0 score; III-IV type: 1 score), and systemic-immune inflammation index (SII,<550 × 1012: 0 score; ≥550 × 1012: 1 score). Patients were categorized into three risk groups based on their ABPS score: ABPS-A, B, and C (scored 0, 1-2, and 3, respectively). The concordance index (C-index) of the ABPS scoring system was calculated to be 0.802, significantly outperforming the HAP score (0.758), 6-12 (0.712), Up to 7 (0.683), and ALBI (0.595) scoring systems (all P < 0.05). These research findings were further validated in the external validation cohorts. CONCLUSION: The ABPS score demonstrated a strong association with survival outcomes and radiological response in patients undergoing TACE plus molecular targeted therapy for HCC with PVTT. The ABPS scoring system could serve as a valuable tool to guide treatment selection for these patients.

Improved neural network with multi-task learning for Alzheimer's disease classification.

Alzheimer's disease(AD) poses a significant challenge due to its widespread prevalence and the lack of effective treatments, highlighting the urgent need for early detection. This research introduces an enhanced neural network, named ADnet, which is based on the VGG16 model, to detect Alzheimer's disease using two-dimensional MRI slices. ADNet incorporates several key improvements: it replaces traditional convolution with depthwise separable convolution to reduce model parameters, replaces the ReLU activation function with ELU to address potential issues with exploding gradients, and integrates the SE(Squeeze-and-Excitation) module to enhance feature extraction efficiency. In addition to the primary task of MRI feature extraction, ADnet is simultaneously trained on two auxiliary tasks: clinical dementia score regression and mental state score regression. Experimental results demonstrate that compared to the baseline VGG16, ADNet achieves a 4.18% accuracy improvement for AD vs. CN classification and a 6% improvement for MCI vs. CN classification. These findings highlight the effectiveness of ADnet in classifying Alzheimer's disease, providing crucial support for early diagnosis and intervention by medical professionals. The proposed enhancements represent advancements in neural network architecture and training strategies for improved AD classification.

Performance improvement of atherosclerosis risk assessment based on feature interaction.

BACKGROUND AND OBJECTIVE: Cardiovascular disease is a leading cause of mortality and premature death. Early intervention in asymptomatic individuals through risk assessment can reduce the incidence of disease. Atherosclerosis is a major cause of cardiovascular disease and early detection can effectively prevent and treat it. In this study, we used real patient data to evaluate the risk of atherosclerosis, assisting doctors in diagnosis and reducing the incidence of cardiovascular disease. METHODS: We proposed a multi-stage atherosclerosis risk assessment model that includes three main stages: (i) SMOTE and decorrelation weighting algorithm technology were added to the causal stability middle layer to address class imbalance in the dataset and reduce the impact of feature-induced dataset distribution shifts on model differences. (ii) The feature interaction layer considered possible feature interactions and classified features by different categories. By adding more effective feature information, the accuracy and generalizability of the model were improved. (iii) In the integrated model layer, we chose LightGBM as the decision tree integration model for risk assessment because it has higher accuracy and robustness compared to other machine learning algorithms. RESULTS: The final model used a dataset containing 21 original features and 17 interaction features, achieving excellent performance under a 10-fold cross-validation strategy. The macro accuracy reached 93.86%, macro precision was 94.82%, macro recall was 93.52%, and macro F1 score was as high as 93.37%. These indicators demonstrate the accuracy and robustness of the model in atherosclerosis risk assessment. CONCLUSION: The model provides strong support for the prevention and diagnosis of cardiovascular disease. Through atherosclerosis risk assessment, the model can help doctors develop personalized prevention and treatment plans, which is of great significance for the prevention and treatment of cardiovascular disease.

A self-supervised network for image denoising and watermark removal.

In image watermark removal, popular methods depend on given reference non-watermark images in a supervised way to remove watermarks. However, reference non-watermark images are difficult to be obtained in the real world. At the same time, they often suffer from the influence of noise when captured by digital devices. To resolve these issues, in this paper, we present a self-supervised network for image denoising and watermark removal (SSNet). SSNet uses a parallel network in a self-supervised learning way to remove noise and watermarks. Specifically, each sub-network contains two sub-blocks. The upper sub-network uses the first sub-block to remove noise, according to noise-to-noise. Then, the second sub-block in the upper sub-network is used to remove watermarks, according to the distributions of watermarks. To prevent the loss of important information, the lower sub-network is used to simultaneously learn noise and watermarks in a self-supervised learning way. Moreover, two sub-networks interact via attention to extract more complementary salient information. The proposed method does not depend on paired images to learn a blind denoising and watermark removal model, which is very meaningful for real applications. Also, it is more effective than the popular image watermark removal methods in public datasets. Codes can be found at https://github.com/hellloxiaotian/SSNet.

Laryngeal mask airway combined with bronchial blocker achieved 1-lung ventilation in a patient with bilateral vocal cord paralysis: A case report.

INTRODUCTION: One-lung ventilation (OLV) is a commonly used technique to facilitate surgical visualization during thoracic surgical procedures. Double-lumen endotracheal tubes and one-lumen tracheal tube combined with bronchial blocker might lead to intubation-related laryngeal injury. PATIENT CONCERNS: In the perioperative period, how to avoid further damage to the vocal cord while achieving OLV during operation is challenging work. DIAGNOSIS: She was diagnosed with systemic lupus erythematosus, bilateral vocal cord paralysis, and lung tumor. INTERVENTIONS: We used a combination of a laryngeal mask airway with bronchial blocker to avoid further damage to the vocal cord when achieving OLV. OUTCOMES: At 1-month follow-up, she had fully recovered without obvious abnormalities. CONCLUSION: When OLV was required for patients with bilateral vocal cord paralysis, a combination of a laryngeal mask airway with bronchial blocker was considered a better choice.

Genome-Wide Association Study and Prediction of Tassel Weight of Tropical Maize Germplasm in Multi-Parent Population.

Tassel weight (TW) is a crucial agronomic trait that significantly affects pollen supply and grain yield development in maize breeding. To improve maize yield and develop new varieties, a comprehensive understanding of the genetic mechanisms underlying tassel weight is essential. In this study, tropical maize inbred lines, namely CML312, CML373, CML444, and YML46, were selected as female parents and crossed with the elite maize inbred line Ye107, which served as the common male parent, to develop a multi-parent population comprising four F8 recombinant inbred line (RIL) subpopulations. Using 6616 high-quality single nucleotide polymorphism (SNP) markers, we conducted genome-wide association analysis (GWAS) and genomic selection (GS) on 642 F8 RILs in four subpopulations across three different environments. Through GWAS, we identified 16 SNPs that were significantly associated with TW, encompassing two stable loci expressed across multiple environments. Furthermore, within the candidate regions of these SNPs, we discovered four novel candidate genes related to TW, namely Zm00001d044362, Zm00001d011048, Zm00001d011049, and Zm00001d031173 distributed on chromosomes 1, 3, and 8, which have not been previously reported. These genes are involved in processes such as signal transduction, growth and development, protein splicing, and pollen development, all of which play crucial roles in inflorescence meristem development, directly affecting TW. The co-localized SNP, S8_137379725, on chromosome 8 was situated within a 16.569 kb long terminal repeat retrotransposon (LTR-RT), located 22.819 kb upstream and 26.428 kb downstream of the candidate genes (Zm00001d011048 and Zm00001d011049). When comparing three distinct GS models, the BayesB model demonstrated the highest accuracy in predicting TW. This study establishes the theoretical foundation for future research into the genetic mechanisms underlying maize TW and the efficient breeding of high-yielding varieties with desired tassel weight through GS.

Utilization of the corticomedullary difference in magnetic resonance imaging-derived apparent diffusion coefficient for noninvasive assessment of chronic kidney disease in type 2 diabetes.

BACKGROUNDS: Accumulating data demonstrated that the cortico-medullary difference in apparent diffusion coefficient (ΔADC) of diffusion-weighted magnetic resonance imaging (DWI) was a better correlation with kidney fibrosis, tubular atrophy progression, and a predictor of kidney function evolution in chronic kidney disease (CKD). OBJECTIVES: We aimed to assess the value of ΔADC in evaluating disease severity, differential diagnosis, and the prognostic risk stratification for patients with type 2 diabetes (T2D) and CKD. METHODS: Total 119 patients with T2D and CKD who underwent renal MRI were prospectively enrolled. Of them, 89 patients had performed kidney biopsy for pathological examination, including 38 patients with biopsy-proven diabetic kidney disease (DKD) and 51 patients with biopsy-proven non-diabetic kidney disease (NDKD) and Mix (DKD + NDKD). Clinicopathological characteristics were compared according to different ΔADC levels. Moreover, univariate and multivariate-linear regression analyses were performed to explore whether ΔADC was independently associated with estimated glomerular filtration rate (eGFR) and urinary albumin creatinine ratio (UACR). The diagnostic performance of ΔADC for discriminating DKD from NDKD + Mix was evaluated by receiver operating characteristic (ROC) analysis. In addition, an individual's 2- or 5-year risk probability of progressing to end-stage kidney disease (ESKD) was calculated by the kidney failure risk equation (KFRE). The effect of ΔADC on prognostic risk stratification was assessed. Additionally, net reclassification improvement (NRI) was used to evaluate the model performance. RESULTS: All enrolled patients had a median ΔADC level of 86 (IQR 28, 155) × 10-6 mm2/s. ΔADC significantly decreased across the increasing staging of CKD (P < 0.001). Moreover, those with pathological-confirmed DKD has a significantly lower level of ΔADC than those with NDKD and Mix (P < 0.001). It showed that ΔADC was independently associated with eGFR (ß = 1.058, 95% CI = [1.002,1.118], P = 0.042) and UACR (ß = -3.862, 95% CI = [-7.360, -0.365], P = 0.031) at multivariate linear regression analyses. Besides, ΔADC achieved an AUC of 0.707 (71% sensitivity and 75% specificity) and AUC of 0.823 (94% sensitivity and 67% specificity) for discriminating DKD from NDKD + Mix and higher ESKD risk categories (≥50% at 5 years; ≥10% at 2 years) from lower risk categories (<50% at 5 years; <10% at 2 years). Accordingly, the optimal cutoff value of ΔADC for higher ESKD risk categories was 66 × 10-6 mm2/s, and the group with the low-cutoff level of ΔADC group was associated with 1.232 -fold (95% CI 1.086, 1.398) likelihood of higher ESKD risk categories as compared to the high-cutoff level of ΔADC group in the fully-adjusted model. Reclassification analyses confirmed that the final adjusted model improved NRI. CONCLUSIONS: ΔADC was strongly associated with eGFR and UACR in patients with T2D and CKD. More importantly, baseline ΔADC was predictive of higher ESKD risk, independently of significant clinical confounding. Specifically, ΔADC <78 × 10-6 mm2/s and <66 × 10-6 mm2/s would help to identify T2D patients with the diagnosis of DKD and higher ESKD risk categories, respectively.