Precise Grading Research Articles

An eXplainable Deep Learning Model for Multi-Modal MRI Grading of IDH-Mutant Astrocytomas

IDH-Mutant-astrocytomas are malignant brain glioma tumors. They are graded as lower (grades II and III) or higher (grade IV) according to their rate of growth and molecular features. Conventional IDH grading requires pathology examination, which is invasive, costly, and time-consuming. Non-invasive MRI methods are widely used but require a high level of interpretation for tumor grading. Recent advances in deep learning have shown potential to aid and improve the accuracy and efficiency of tumor classification. This study proposes a novel non-invasive approach for IDH-Mutant-astrocytomas grading using Light-weight Attention Network (LAN) that integrates multi-modal MRI data, including both conventional and advanced MR modalities. The study addresses inter-modality heterogeneity using Principal Component Analysis (PCA) while minimizing computational complexity. LAN uses a 3D Convolutional Neural Network (CNN) and a volumetric attention mechanism to extract and classify local and global tumor patterns. This is followed by an eXplainable AI approach that use SHapley Additive exPlanations (SHAP) values to understand the model decisions. LAN model outperforms other pre-trained state-of-the-art models, with an overall accuracy of 0.84. The advanced modalities Arterial Spin Labeling (ASL) and the Diffusion-weighted imaging(DWI) SHAP score contributes significantly in improving tumor heterogeneity visualization for aggressive grade II along with conventional MRIs. This work demonstrates the feasibility of integrating explainable deep learning with multi-modal MRI data for precise and comprehensible early glioma grading, potentially leading to better clinical decision-making.

Precise grading of non-muscle invasive bladder cancer with multi-scale pyramidal CNN

The grading of non-muscle invasive bladder cancer (NMIBC) continues to face challenges due to subjective interpretations, which affect the assessment of its severity. To address this challenge, we are developing an innovative artificial intelligence (AI) system aimed at objectively grading NMIBC. This system uses a novel convolutional neural network (CNN) architecture called the multi-scale pyramidal pretrained CNN to analyze both local and global pathology markers extracted from digital pathology images. The proposed CNN structure takes as input three levels of patches, ranging from small patches (e.g., 128×128\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$128 \ imes 128$$\\end{document}) to the largest size patches (512×512\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$512 \ imes 512$$\\end{document}). These levels are then fused by random forest (RF) to estimate the severity grade of NMIBC. The optimal patch sizes and other model hyperparameters are determined using a grid search algorithm. For each patch size, the proposed system has been trained on 32K patches (comprising 16K low-grade and 16K high-grade samples) and subsequently tested on 8K patches (consisting of 4K low-grade and 4K high-grade samples), all annotated by two pathologists. Incorporating light and efficient processing, defining new benchmarks in the application of AI to histopathology, the ShuffleNet-based AI system achieved notable metrics on the testing data, including 94.25% ± 0.70% accuracy, 94.47% ± 0.93% sensitivity, 94.03% ± 0.95% specificity, and a 94.29% ± 0.70% F1-score. These results highlight its superior performance over traditional models like ResNet-18. The proposed system’s robustness in accurately grading pathology demonstrates its potential as an advanced AI tool for diagnosing human diseases in the domain of digital pathology.

Machine learning-driven intelligent tire wear detection system

Traditional methods detect wear by interpreting mathematical models and tire characteristics; however, these methods struggle to accurately reflect the actual rolling condition of the tire. In this study, we propose a machine learning-based tire wear detection module that can provide accurate results under tire test rig conditions. To develop this module, we designed three key components: integrated acceleration and PVDF sensors within the tire to capture vibration and deformation data; signal preprocessing algorithms to highlight multi-source signal differences under varying wear conditions; and deep learning algorithms to achieve precise tire wear grade identification. Experimental results demonstrate that, under different tire pressures, loads, speeds, and wear levels, the system can accurately identify tire wear grades with 99.99% accuracy by combining data from both sensors.

Assessing water quality environmental grades using hyperspectral images and a deep learning model: A case study in Jiangsu, China

Water quality assessment is essential for effective environmental management, yet traditional methods such as chemical sampling are often labor-intensive and inefficient for large-scale, continual monitoring. This study addresses these limitations by leveraging hyperspectral images (HSIs) analysis and introducing a capsule network (CapsNet) model enhanced with a multidimensional integration attention (MDIA) mechanism. The model is specifically designed to integrate both channel and spatial information, enabling precise water quality grade assessment by detecting subtle features within HSIs data. To validate the performance of the model, spectral data from 5 water quality regions are collected and processed via a UAV-carried spectrometer, with 4503 water quality data samples. Rigorous classification experiments demonstrated that the model achieves 98.73 % accuracy, with an average improvement of 4.89 % compared with the other models. This approach significantly improves decision support systems for water resource management, facilitating the sustainable use of water resources.

What's new in neuropathology 2024: CNS WHO 5th edition updates.

The fifth edition of the World Health Organization (WHO) Classification of Central Nervous System (CNS) Tumors was released in 2021, just five years following the updated fourth edition. Advanced molecular testing such as next-generation sequencing, RNA fusion analysis, and DNA methylation profiling has led to more precise grading and classification of pre-existing tumor types as well as the recognition of new ones. Herein, we outline the major updates of the 2021 WHO Classification of CNS tumors, with emphasis on the expanded molecular characterization of CNS tumors.

Prediction of Mortality and Outcome of Various Trauma Scores in Polytrauma Patients.

Background In developing nations, the primary cause of death is trauma, and the prevalence of trauma is increasing as more vehicles are driven. Numerous trauma scoring systems have been created in order to predict the mortality rate and patients with trauma's prognosis. The purpose of the current study was to assess the prognostic ability of various trauma scoring systems for patients' mortality and morbidity in cases involving chest and abdominal injuries, as they are common in the surgery department. Methodology At Ranchi, Jharkhand's Rajendra Institute of Medical Sciences, a prospective observational study was conducted from June 2021 to September 2022. All patients who met the requirements for inclusion were older than 18 and reported chest and abdominal trauma totaling 204. Before any essential therapies, primary care and resuscitation, including airway maintenance, breathing, circulation, and hemorrhage control, were established. A comprehensive clinical evaluation was done based on each patient's needs. Radiological evaluation included chest X-ray and ultrasonography (USG) for chest trauma, whereas USG (FAST) and CT scans were for abdominal trauma. Trauma scores, such as the Revised Trauma Score (RTS), the Trauma Revised Injury Severity Score (TRISS), the New Injury Severity Score (NISS), and the Injury Severity Score (ISS), were computed and examined using IBM Corp. Released 2011. IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp. Results Of the 204 patients, 14.7% were female and 85.3% were male. The age range of 21-30 years old accounted for the largest percentage of cases (28%). 50 percent of injuries were caused by automobile accidents, while 25% were the result of assaults. At 83.8% versus 16.2%, blunt injuries were substantially more common than penetrating ones. In comparison to the chest, the abdomen was more frequently involved. The study's findings demonstrated that, while every trauma scoring was statistically significant in predicting mortality, the New Injury Severity Score (NISS), as well as the Trauma Revised Injury Severity Score (TRISS), became the most effective in predicting mortality (p < 0.0001). Conclusion According to the results, the most precise trauma grading method for chest and abdominal injuries is the Trauma Revised Injury Severity Score (TRISS), even though all other trauma scoring systems are useful in predicting patient outcomes. Surgeons using these metrics to predict outcomes and make well-informed treatment decisions can benefit greatly.

2D Path Following Control of Motor Graders

Optimal ground conditions in mining are crucial for operational efficiency and safety. Motor graders play a crucial role in achieving precise grading of passages; however, unlike many other mining vehicles, they have not yet been commercially automated. The redundant kinematics of motor graders, including articulation, front axle steering, and blade operations, pose significant challenges for automation. This research explores path following with a 2D kinematic model of motor graders and uses simulation to evaluate and compare the performance of these controllers in terms of step response and operational effectiveness. To address kinematic redundancy, two methods are proposed: Rear Offset Control, which defines the articulation angle using a lateral offset between the front and rear axles, and Single Track Control, which utilizes steering redundancy to adapt existing automation approaches to articulated vehicles. Simulation tests were conducted using two controllers: a Feedback Linearized PD controller (FBL+PD) and a Feedback Linearized Model Predictive Controller (FBL+MPC). Both were tuned for optimal performance on a step input path, with performance assessed based on Root Mean Square Error (RMSE) and control effort. This work introduces two methodologies—Rear Offset Control and Single Track Control—for autonomous motor grader operation. These were implemented with FBL+PD and FBL+MPC controllers on a 2D kinematic model. Future research will focus on dynamic modeling, implementing a Non-Linear Model-Predictive Controller, refining tuning methodologies for Feedback Linearized systems, and integrating these approaches with live vehicles.

Machine Learning-Based Tomato Fruit Shape Classification System.

Fruit shape significantly impacts the quality and commercial value of tomatoes (Solanum lycopersicum L.). Precise grading is essential to elucidate the genetic basis of fruit shape in breeding programs, cultivar descriptions, and variety registration. Despite this, fruit shape classification is still primarily based on subjective visual inspection, leading to time-consuming and labor-intensive processes prone to human error. This study presents a novel approach incorporating machine learning techniques to establish a robust fruit shape classification system. We trained and evaluated seven supervised machine learning algorithms by leveraging a public dataset derived from the Tomato Analyzer tool and considering the current four classification systems as label variables. Subsequently, based on class-specific metrics, we derived a novel classification framework comprising seven discernible shape classes. The results demonstrate the superiority of the Support Vector Machine model in terms of its accuracy, surpassing human classifiers across all classification systems. The new classification system achieved the highest accuracy, averaging 88%, and maintained a similar performance when validated with an independent dataset. Positioned as a common standard, this system contributes to standardizing tomato fruit shape classification, enhancing accuracy, and promoting consensus among researchers. Its implementation will serve as a valuable tool for overcoming bias in visual classification, thereby fostering a deeper understanding of consumer preferences and facilitating genetic studies on fruit shape morphometry.

Artificial intelligence-assisted grading for tear trough deformity

BackgroundVarious classification systems for tear trough deformity (TTD) have been published, but their complexity can pose challenges in clinical use, especially for less experienced surgeons. It is believed that artificial intelligence (AI) technology can address some of these challenges by reducing inadvertent errors and improving the accuracy of medical practice. In this study, we aimed to establish a reliable and precise digital image grading model for TTD using smartphone-based photography, enhanced by AI deep learning technology. This model is designed to aid and guide surgeons, particularly those who are less experienced or from younger generations, during clinical examinations and in making decisions about further surgical interventions. Materials and MethodsA total of 504 patients and 983 photos were included in the study. We adopted the Barton’s grading system for TTD. All photos were taken with the same smartphone and processed and analyzed using the medical AI assistant (MAIATM) software. The photos were then randomly divided into two groups to establish training and testing models. ResultsThe confusion matrix for the training model demonstrated a sensitivity of 56%, specificity of 87.3%, an F1 score of 0.57, and an area under the curve (AUROC) of 0.85. For the testing group, the sensitivity was 49.3%, specificity was 85%, with an F1 score of 0.49, and an AUROC of 0.83. Representative heatmaps were also generated. ConclusionOur study is the first to demonstrate that tear trough deformities can be easily categorized using a built-in smartphone camera in conjunction with an AI deep learning program. This approach can reduce errors during clinical patient evaluations, particularly for less experienced practitioners.

Exploring the Histopathological Landscape of Urinary Bladder Diseases: A Tertiary Care Center Study.

Introduction Urinary bladder lesions encompass a wide spectrum, from benign inflammatory conditions to malignant neoplasms, presenting diagnostic and therapeutic challenges. Urothelial carcinoma predominates among bladder malignancies, exhibiting diverse clinical presentations and prognoses. Objective This study aimed to delineate the histopathological spectrum of urinary bladder lesions and correlate demographic profiles, clinical features, and cystoscopic findings with various bladder lesions. Methods This prospective descriptive observational study spanned 24 months at a tertiary care center, involving 65 cases of urinary bladder biopsies, including transurethral resection of bladder tumors, cystoscopic biopsies, and cystectomy specimens. The histopathological examination followed the WHO 2022 classification of urinary bladder tumors and the American Joint Committee on Cancer eighth edition staging. Clinical data, including age, gender, cystoscopic findings, and presenting symptoms, were correlated with histopathological diagnoses to explore the spectrum of bladder lesions. Results Neoplastic lesions predominated, constituting 92.3% of cases, with urothelial carcinoma comprising 83.33% of these cases. Among neoplastic lesions, invasive high-grade urothelial carcinoma (36.7%) and non-invasive low-grade papillary urothelial neoplasm (20.0%) were the most frequently observed subtypes. Non-neoplastic lesions accounted for 7.7%, including various forms of cystitis. Hematuria was the predominant presenting symptom (81.5%), while cystoscopic examinations revealed that most lesions were situated in the lateral bladder wall. High-grade urothelial carcinomas were mostly associated with muscularis propria invasion. Conclusion This study underscores the critical role of histopathological examination in diagnosing and managing urinary bladder diseases and distinguishing between non-neoplastic and neoplastic lesions. Urothelial carcinoma, prevalent among older age groups, often demonstrated muscle invasion indicative of high-grade tumors. Including the muscle layer in cystoscopic biopsies is crucial for an accurate diagnosis. Conversely, though less common, non-neoplastic conditions encompass various forms of cystitis. These findings highlight the importance of precise diagnostic tools such as cystoscopy and histopathological examination for the early detection and management of bladder neoplasms. Histopathological assessment offers essential prognostic guidance, aids in precise staging and grading, and directs tailored treatment strategies.

P-256 AI-enhanced precision: unveiling the future of blastocyst quality assessment

Abstract Study question Can artificial intelligence (AI) effectively assess various regions of embryos to estimate the quality of human embryos captured through an optical microscope? Summary answer Artificial Intelligence excels in analyzing diverse embryonic regions based on morphology, ensuring precise grade predictions, assessing human embryos via optical microscopes. What is known already In recent research, the integration of AI and computer vision has marked significant strides in embryo selection, providing precise predictions of clinical pregnancy based on images of human embryos, particularly at Day 5. This study extends its findings by incorporating techniques that exhibit resilience to variations in camera and microscope types. Notably, the AI algorithms underwent rigorous testing on double-blind datasets, comprising optical microscopic images sourced from Optical microscope systems across diverse regions such as India, Europe, the UK, the US, and Australia. These results signify a promising advancement in the standardization and applicability of AI-assisted embryo assessment methodologies. Study design, size, duration Approximately 11824 static 2D images of Day 5 blastocysts with related pregnancy, preimplantation genetic testing for aneuploidy (PGT-A) outcomes, demographic and clinic geographical location information have been collected. Images were divided into three groups: training, validation and blind test sets. An AI model was trained on 3473 embryo images, validated on 1184 and tested on further blind set of 1253 images from a separate clinics and demographsic. Gardner based grading system was used. Participants/materials, setting, methods Dataset included optical-microscope images from fertility patients across 25 IVF laboratories in six countries was utilized to train, validate, and test the AI. Dataset was graded by 9 embryologists and final grade was decided based on max-voting. The identity of participants was kept secret and did not associate with used data. Assessment encompassed accuracy, distribution, and robustness concerning camera/microscope type. The dataset spanned various embryo development stages, excluding frozen, degenerated, non-blastocyst embryos. Main results and the role of chance The study distinguishes itself by incorporating data from diverse environments, enhancing generalizability. Leveraging transformer-based attention models ensures meticulous focus on regions relevant to the Gardner grading system, ensuring accurate predictions akin to human judgment and minimizing errors prone to subjectivity. AI’s predictive accuracy on blind dataset for expansion is 82.18%, for ICM is 76.53%, and for trophectoderm is 77.53%, surpassing previous studies. The model demonstrates specificity 79.16%, sensitivity of 85.95% and accuracy of 82.44% in estimating embryo quality. Direct application on embryoscope images is hindered by differing morphology, excluding the embryo part. However, employing segmentation and applying models yields comparable performance with approx. 2.87% drop in accuracy. These findings suggest that the proposed method, with AI trained on a globally diverse dataset, establishes a generalizable AI robust to variations in camera type and focal settings. Limitations, reasons for caution Data collected for this solution has been limited to mainly static images. The study’s focus on good-quality Day 5 blastocyst embryos might limit generalizability to other stages. The dataset’s global diversity may not fully capture demographic nuances. Ongoing AI advancements may necessitate continual updates for relevance and accuracy. Wider implications of the findings Broader implications of the findings suggests transformative potential for AI in embryo quality assessment. The methodology’s success in diverse environments and robustness to variations in camera types underscore its applicability on global scale. This could lead to standardized, automated, and precise embryo evaluations, offering significant advancements in assisted reproductive technology. Trial registration number not applicable

DFCAFNet: Dual-feature co-attentive fusion network for diabetic retinopathy grading

Diabetic retinopathy (DR) grading is a complicated task characterized by inter-class variations among several categories, the subtle detection of tiny lesions, and uneven data distribution. To achieve precise DR grading, extracting distinct features that can effectively capture minor visual variations in lesions is necessary. The constraints of conventional convolutional neural networks (CNNs) are evident in their difficulty in detecting tiny lesions, which is further enhanced by their biased focus due to an uneven distribution of data. Furthermore, comparable color and texture attributes among different classes add to the complexity of achieving accurate grading results. This paper presents a new multi-scale attention fusion network (MS-AFNet) incorporating multi-scale contextual information. By successfully combining local and global features, the network improves the diagnostic accuracy of the model in handling inter-class variances. A concurrent attention module (CAM) with multiple attention blocks enhances the ability to distinguish information by capturing complex features within the lesions through numerous attention mechanisms. This novel approach greatly improves the model’s capacity to distinguish nuanced features in minor lesions, resulting in more precise diagnostic outcomes. A dual-feature co-attentive fusion network (DFCAFNet) is proposed by combining CAM and MS-AFNet, which capture spatial patterns and highlight inter-channel interactions to detect complex patterns and enhance accuracy. An extensive assessment of two datasets on DR demonstrates the exceptional accuracy and computational efficiency of DFCAFNet.

An Empirical Study on Cataract Multiclass Grading Assessment with Slit Lamp Bio-microscope Images Using Neural Network Models

Cataract, an age-related eye disease, poses a significant ophthalmological public health challenge in both developed and developing nations. Tailoring treatment or surgery plans helps accurately categorise the cataract's developmental stage. Precise cataract grading helps in diagnosing cataracts and subsequently scheduling surgical intervention. In this project endeavour, a solution is presented to automate the cataract grading process utilizing slit lamp bio-microscope data sets acquired through smartphones. This innovation is particularly valuable for novice practitioners and non-specialist doctors/experts who may struggle with proficiently interpreting cataract progression, leading to potential misdiagnoses. To address this challenge, a Neural Network model is harnessed to automatically predict the grade of cataracts. The study employs multi-class image classification models, including the Convolutional neural network (CNN) model, the Efficient Net B0 model, and the ResNet50 model, for this purpose. Notably, the ResNet50 model outperforms the other models in terms of accuracy and prediction capability for the provided data set. Achieving an accuracy rate of 0.8611, the ResNet50 model demonstrates superior performance in classifying cataract grades, after augmenting the data set with 544 images. This performance comparison establishes the ResNet50 model as the most robust choice among the considered models and data sets.

A non-invasive AI-based system for precise grading of anosmia in COVID-19 using neuroimaging

COVID-19 (Coronavirus), an acute respiratory disorder, is caused by SARS-CoV-2 (coronavirus severe acute respiratory syndrome). The high prevalence of COVID-19 infection has drawn attention to a frequent illness symptom: olfactory and gustatory dysfunction. The primary purpose of this manuscript is to create a Computer-Assisted Diagnostic (CAD) system to determine whether a COVID-19 patient has normal, mild, or severe anosmia. To achieve this goal, we used fluid-attenuated inversion recovery (FLAIR) Magnetic Resonance Imaging (FLAIR-MRI) and Diffusion Tensor Imaging (DTI) to extract the appearance, morphological, and diffusivity markers from the olfactory nerve. The proposed system begins with the identification of the olfactory nerve, which is performed by a skilled expert or radiologist. It then proceeds to carry out the subsequent primary steps: (i) extract appearance markers (i.e., 1st and 2nd order markers), morphology/shape markers (i.e., spherical harmonics), and diffusivity markers (i.e., Fractional Anisotropy (FA) & Mean Diffusivity (MD)), (ii) apply markers fusion based on the integrated markers, and (iii) determine the decision and corresponding performance metrics based on the most-promising classifier. The current study is unusual in that it ensemble bags the learned and fine-tuned ML classifiers and diagnoses olfactory bulb (OB) anosmia using majority voting. In the 5-fold approach, it achieved an accuracy of 94.1%, a balanced accuracy (BAC) of 92.18%, precision of 91.6%, recall of 90.61%, specificity of 93.75%, F1 score of 89.82%, and Intersection over Union (IoU) of 82.62%. In the 10-fold approach, stacking continued to demonstrate impressive results with an accuracy of 94.43%, BAC of 93.0%, precision of 92.03%, recall of 91.39%, specificity of 94.61%, F1 score of 91.23%, and IoU of 84.56%. In the leave-one-subject-out (LOSO) approach, the model continues to exhibit notable outcomes, achieving an accuracy of 91.6%, BAC of 90.27%, precision of 88.55%, recall of 87.96%, specificity of 92.59%, F1 score of 87.94%, and IoU of 78.69%. These results indicate that stacking and majority voting are crucial components of the CAD system, contributing significantly to the overall performance improvements. The proposed technology can help doctors assess which patients need more intensive clinical care.

Diagnostic Challenges in Aortic Stenosis.

Aortic stenosis (AS) is the most prevalent degenerative valvular disease in western countries. Transthoracic echocardiography (TTE) is considered, nowadays, to be the main imaging technique for the work-up of AS due to high availability, safety, low cost, and excellent capacity to evaluate aortic valve (AV) morphology and function. Despite the diagnosis of AS being considered straightforward for a very long time, based on high gradients and reduced aortic valve area (AVA), many patients with AS represent a real dilemma for cardiologist. On the one hand, the acoustic window may be inadequate and the TTE limited in some cases. On the other hand, a growing body of evidence shows that patients with low gradients (due to systolic dysfunction, concentric hypertrophy or coexistence of another valve disease such as mitral stenosis or regurgitation) may develop severe AS (low-flow low-gradient severe AS) with a similar or even worse prognosis. The use of complementary imaging techniques such as transesophageal echocardiography (TEE), multidetector computed tomography (MDTC), or cardiac magnetic resonance (CMR) plays a key role in such scenarios. The aim of this review is to summarize the diagnostic challenges associated with patients with AS and the advantages of a comprehensive multimodality cardiac imaging (MCI) approach to reach a precise grading of the disease, a crucial factor to warrant an adequate management of patients.

MediDRNet: Tackling category imbalance in diabetic retinopathy classification with dual-branch learning and prototypical contrastive learning

Background and Objective:The classification of diabetic retinopathy (DR) aims to utilize the implicit information in images for early diagnosis, to prevent and mitigate the further worsening of the condition. However, existing methods are often limited by the need to operate within large, annotated datasets to show significant advantages. Additionally, the number of samples for different categories within the dataset needs to be evenly distributed, because the characteristic of sample imbalance distribution can lead to an excessive focus on high-frequency disease categories, while neglecting the less common but equally important disease categories. Therefore, there is an urgent need to develop a new classification method that can effectively alleviate the issue of sample distribution imbalance, thereby enhancing the accuracy of diabetic retinopathy classification. Methods:In this work, we propose MediDRNet, a dual-branch network model based on prototypical contrastive learning. This model adopts prototype contrastive learning, creating prototypes for different levels of lesions, ensuring they represent the core features of each lesion level. It classifies by comparing the similarity between data points and their category prototypes. Our dual-branch network structure effectively resolves the issue of category imbalance and improves classification accuracy by emphasizing subtle differences in retinal lesions. Moreover, our approach combines a dual-branch network with specific lesion-level prototypes for core feature representation and incorporates the convolutional block attention module for enhanced lesion feature identification. Results:Our experiments using both the Kaggle and UWF classification datasets have demonstrated that MediDRNet exhibits exceptional performance compared to other advanced models in the industry, especially on the UWF DR classification dataset where it achieved state-of-the-art performance across all metrics. On the Kaggle DR classification dataset, it achieved the highest average classification accuracy (0.6327) and Macro-F1 score (0.6361). Particularly in the classification tasks for minority categories of diabetic retinopathy on the Kaggle dataset (Grades 1, 2, 3, and 4), the model reached high classification accuracies of 58.08%, 55.32%, 69.73%, and 90.21%, respectively. In the ablation study, the MediDRNet model proved to be more effective in feature extraction from diabetic retinal fundus images compared to other feature extraction methods. Conclusions:This study employed prototype contrastive learning and bidirectional branch learning strategies, successfully constructing a grading system for diabetic retinopathy lesions within imbalanced diabetic retinopathy datasets. Through a dual-branch network, the feature learning branch effectively facilitated a smooth transition of features from the grading network to the classification learning branch, accurately identifying minority sample categories. This method not only effectively resolved the issue of sample imbalance but also provided strong support for the precise grading and early diagnosis of diabetic retinopathy in clinical applications, showcasing exceptional performance in handling complex diabetic retinopathy datasets. Moreover, this research significantly improved the efficiency of prevention and management of disease progression in diabetic retinopathy patients within medical practice. We encourage the use and modification of our code, which is publicly accessible on GitHub: https://github.com/ReinforceLove/MediDRNet.

Construction of diagnostic models for the progression of hepatocellular carcinoma using machine learning.

Liver cancer is one of the most prevalent forms of cancer worldwide. A significant proportion of patients with hepatocellular carcinoma (HCC) are diagnosed at advanced stages, leading to unfavorable treatment outcomes. Generally, the development of HCC occurs in distinct stages. However, the diagnostic and intervention markers for each stage remain unclear. Therefore, there is an urgent need to explore precise grading methods for HCC. Machine learning has emerged as an effective technique for studying precise tumor diagnosis. In this research, we employed random forest and LightGBM machine learning algorithms for the first time to construct diagnostic models for HCC at various stages of progression. We categorized 118 samples from GSE114564 into three groups: normal liver, precancerous lesion (including chronic hepatitis, liver cirrhosis, dysplastic nodule), and HCC (including early stage HCC and advanced HCC). The LightGBM model exhibited outstanding performance (accuracy = 0.96, precision = 0.96, recall = 0.96, F1-score = 0.95). Similarly, the random forest model also demonstrated good performance (accuracy = 0.83, precision = 0.83, recall = 0.83, F1-score = 0.83). When the progression of HCC was categorized into the most refined six stages: normal liver, chronic hepatitis, liver cirrhosis, dysplastic nodule, early stage HCC, and advanced HCC, the diagnostic model still exhibited high efficacy. Among them, the LightGBM model exhibited good performance (accuracy = 0.71, precision = 0.71, recall = 0.71, F1-score = 0.72). Also, performance of the LightGBM model was superior to that of the random forest model. Overall, we have constructed a diagnostic model for the progression of HCC and identified potential diagnostic characteristic gene for the progression of HCC.

Methods for Constructing a Refined Early-Warning Model for Rainstorm-Induced Waterlogging in Historic and Cultural Districts

Against the backdrop of increasingly severe global climate change, the risk of rainstorm-induced waterlogging has become the primary threat to the safety of historic and cultural districts worldwide. This paper focuses on the historic and cultural districts of Beijing, China, and explores techniques and methods for identifying extreme rainstorm warnings in cultural heritage areas. Refined warning and forecasting have become important non-engineering measures to enhance these districts’ waterlogging prevention control and emergency management capabilities. This paper constructs a rainstorm-induced waterlogging risk warning model tailored for Beijing’s historical and cultural districts. This model system encompasses three sets of models: a building waterlogging early-warning model, a road waterlogging early-warning model, and a public evacuation early-warning model. During the construction of the model, the core concepts and determination methods of “1 h rainfall intensity water logging index” and “the waterlogging risk index in historical and cultural districts” were proposed. The construction and application of the three models take into full account the correlation between rainfall intensity and rainwater accumulation, while incorporating the characteristics of flood resilience in buildings, roads, and the society in districts. This allows for a precise grading of warning levels, leading to the formulation of corresponding warning response measures. Empirical tests have shown that the construction method proposed in this paper is reliable. The innovative results not only provide a new perspective and method for the early-warning of rainstorm-induced waterlogging, but also offer scientific support for emergency planning and response in historical and cultural districts.

Visual interpretable MRI fine grading of meniscus injury for intelligent assisted diagnosis and treatment

Meniscal injury represents a common type of knee injury, accounting for over 50% of all knee injuries. The clinical diagnosis and treatment of meniscal injury heavily rely on magnetic resonance imaging (MRI). However, accurately diagnosing the meniscus from a comprehensive knee MRI is challenging due to its limited and weak signal, significantly impeding the precise grading of meniscal injuries. In this study, a visual interpretable fine grading (VIFG) diagnosis model has been developed to facilitate intelligent and quantified grading of meniscal injuries. Leveraging a multilevel transfer learning framework, it extracts comprehensive features and incorporates an attributional attention module to precisely locate the injured positions. Moreover, the attention-enhancing feedback module effectively concentrates on and distinguishes regions with similar grades of injury. The proposed method underwent validation on FastMRI_Knee and Xijing_Knee dataset, achieving mean grading accuracies of 0.8631 and 0.8502, surpassing the state-of-the-art grading methods notably in error-prone Grade 1 and Grade 2 cases. Additionally, the visually interpretable heatmaps generated by VIFG provide accurate depictions of actual or potential meniscus injury areas beyond human visual capability. Building upon this, a novel fine grading criterion was introduced for subtypes of meniscal injury, further classifying Grade 2 into 2a, 2b, and 2c, aligning with the anatomical knowledge of meniscal blood supply. It can provide enhanced injury-specific details, facilitating the development of more precise surgical strategies. The efficacy of this subtype classification was evidenced in 20 arthroscopic cases, underscoring the potential enhancement brought by intelligent-assisted diagnosis and treatment for meniscal injuries.

A hybrid classification model with radiomics and CNN for high and low grading of prostate cancer Gleason score on mp-MRI

Prostate cancer is a prevalent malignancy among men, and the precise diagnosis and grading of prostate cancer have become key research areas in contemporary medicine. Magnetic resonance imaging is an ideal non-invasive method for the detection and diagnosis of prostate cancer, offering superior spatial resolution compared with invasive biopsy procedures. We developed a hybrid model based on radiomics and deep learning, which performs binary classification on prostate cancer cases at the Gleason score ≤ 3 + 4 and ≥ 4 + 3 cases by utilizing T2-weighted imaging and the apparent diffusion coefficient. In the radiomics model, we incorporated a combined region of interest that included a 5 mm margin around the tumor and employed a mirror comparison method for feature selection. In the deep learning model, we utilized a convolutional neural network framework enhanced by the CPCB and SE-Net modules to improve the discrimination ability of our model. A retrospective study was conducted on 650 cases from two hospitals, and our hybrid model achieved superior performance in terms of accuracy, recall, and F1 score compared with the individual models, with respective accuracy values of 0.907, 0.944, and 0.918.