Recommendation algorithm based on attributed multiplex heterogeneous network

In the field of deep learning, the processing of large network models on billions or even tens of billions of nodes and numerous edge types is still flawed, and the accuracy of recommendations is greatly compromised when large network embeddings are applied to recommendation systems. To solve the problem of inaccurate recommendations caused by processing deficiencies in large networks, this paper combines the attributed multiplex heterogeneous network with the attention mechanism that introduces the softsign and sigmoid function characteristics and derives a new framework SSN_GATNE-T (S represents the softsign function, SN represents the attention mechanism introduced by the Softsign function, and GATNE-T represents the transductive embeddings learning for attribute multiple heterogeneous networks). The attributed multiplex heterogeneous network can help obtain more user-item information with more attributes. No matter how many nodes and types are included in the model, our model can handle it well, and the improved attention mechanism can help annotations to obtain more useful information via a combination of the two. This can help to mine more potential information to improve the recommendation effect; in addition, the application of the softsign function in the fully connected layer of the model can better reduce the loss of potential user information, which can be used for accurate recommendation by the model. Using the Adam optimizer to optimize the model can not only make our model converge faster, but it is also very helpful for model tuning. The proposed framework SSN_GATNE-T was tested for two different types of datasets, Amazon and YouTube, using three evaluation indices, ROC-AUC (receiver operating characteristic-area under curve), PR-AUC (precision recall-area under curve) and F1 (F1-score), and found that SSN_GATNE-T improved on all three evaluation indices compared to the mainstream recommendation models currently in existence. This not only demonstrates that the framework can deal well with the shortcomings of obtaining accurate interaction information due to the presence of a large number of nodes and edge types of the embedding of large network models, but also demonstrates the effectiveness of addressing the shortcomings of large networks to improve recommendation performance. In addition, the model is also a good solution to the cold start problem.

In the field of deep learning, the processing of large network models on billions or even tens of billions of nodes and numerous edge types is still flawed, and the accuracy of recommendations is greatly compromised when large network embeddings are applied to recommendation systems. To solve the problem of inaccurate recommendations caused by processing deficiencies in large networks, this paper combines the attributed multiplex heterogeneous network with the attention mechanism that introduces the softsign and sigmoid function characteristics and derives a new framework SSN_GATNE-T (S represents the softsign function, SN represents the attention mechanism introduced by the Softsign function, and GATNE-T represents the transductive embeddings learning for attribute multiple heterogeneous networks). The attributed multiplex heterogeneous network can help obtain more user-item information with more attributes. No matter how many nodes and types are included in the model, our model can handle it well, and the improved attention mechanism can help annotations to obtain more useful information via a combination of the two. This can help to mine more potential information to improve the recommendation effect; in addition, the application of the softsign function in the fully connected layer of the model can better reduce the loss of potential user information, which can be used for accurate recommendation by the model. Using the Adam optimizer to optimize the model can not only make our model converge faster, but it is also very helpful for model tuning. The proposed framework SSN_GATNE-T was tested for two different types of datasets, Amazon and YouTube, using three evaluation indices, ROC-AUC (receiver operating characteristic-area under curve), PR-AUC (precision recall-area under curve) and F1 (F1-score), and found that SSN_GATNE-T improved on all three evaluation indices compared to the mainstream recommendation models currently in existence. This not only demonstrates that the framework can deal well with the shortcomings of obtaining accurate interaction information due to the presence of a large number of nodes and edge types of the embedding of large network models, but also demonstrates the effectiveness of addressing the shortcomings of large networks to improve recommendation performance. In addition, the model is also a good solution to the cold start problem.

In the field of deep learning, the processing of large network models on billions or even tens of billions of nodes and numerous edge types is still flawed, and the accuracy of recommendations is greatly compromised when large network embeddings are applied to recommendation systems. To solve the problem of inaccurate recommendations caused by processing deficiencies in large networks, this paper combines the attributed multiplex heterogeneous network with the attention mechanism that introduces the softsign and sigmoid function characteristics and derives a new framework SSN_GATNE-T (S represents the softsign function, SN represents the attention mechanism introduced by the Softsign function, and GATNE-T represents the transductive embeddings learning for attribute multiple heterogeneous networks). The attributed multiplex heterogeneous network can help obtain more user-item information with more attributes. No matter how many nodes and types are included in the model, our model can handle it well, and the improved attention mechanism can help annotations to obtain more useful information via a combination of the two. This can help to mine more potential information to improve the recommendation effect; in addition, the application of the softsign function in the fully connected layer of the model can better reduce the loss of potential user information, which can be used for accurate recommendation by the model. Using the Adam optimizer to optimize the model can not only make our model converge faster, but it is also very helpful for model tuning. The proposed framework SSN_GATNE-T was tested for two different types of datasets, Amazon and YouTube, using three evaluation indices, ROC-AUC (receiver operating characteristic-area under curve), PR-AUC (precision recall-area under curve) and F1 (F1-score), and found that SSN_GATNE-T improved on all three evaluation indices compared to the mainstream recommendation models currently in existence. This not only demonstrates that the framework can deal well with the shortcomings of obtaining accurate interaction information due to the presence of a large number of nodes and edge types of the embedding of large network models, but also demonstrates the effectiveness of addressing the shortcomings of large networks to improve recommendation performance. In addition, the model is also a good solution to the cold start problem.

Background: The accurate classification between malignant and benign breast lesions detected on mammograms is a crucial but difficult challenge for reducing false-positive recall rates and improving the efficacy of breast cancer screening. Objective: This study aims to optimize a new deep transfer learning model by implementing a novel attention mechanism in order to improve the accuracy of breast lesion classification. Methods: ResNet50 is selected as the base model to develop a new deep transfer learning model. To enhance the accuracy of breast lesion classification, we propose adding a convolutional block attention module (CBAM) to the standard ResNet50 model and optimizing a new model for this task. We assembled a large dataset with 4280 mammograms depicting suspicious soft-tissue mass-type lesions. A region of interest (ROI) is extracted from each image based on lesion center. Among them, 2480 and 1800 ROIs depict verified benign and malignant lesions, respectively. The image dataset is randomly split into two subsets with a ratio of 9:1 five times to train and test two ResNet50 models with and without using CBAM. Results: Using the area under ROC curve (AUC) as an evaluation index, the new CBAM-based ResNet50 model yields AUC = 0.866 ± 0.015, which is significantly higher than that obtained by the standard ResNet50 model (AUC = 0.772 ± 0.008) (p < 0.01). Conclusion: This study demonstrates that although deep transfer learning technology attracted broad research interest in medical-imaging informatic fields, adding a new attention mechanism to optimize deep transfer learning models for specific application tasks can play an important role in further improving model performances.

This study was aimed to assess if combining the evaluation of blood glucose level (BGL) and the Triage Revised Trauma Score (T-RTS) may result in a more accurate prediction of the actual clinical outcome, both in general adult population and in elderly patients with trauma. This is a retrospective cohort study, conducted in the emergency department (ED) of an urban teaching hospital, with an average ED admission rate of 75,000 patients per year. Those excluded: known diagnosis of diabetes, age <18 years old, pregnancy, and mild trauma (classified as isolate trauma of upper or lower limb, in absence of exposed fractures). A combined Revised Trauma Score Glucose (RTS-G) score was obtained adding to T-RTS: two for BGL <160mg/dL (8.9mmol/L); one for BGL ≥160mg/dL and < 200mg/dL (11.1mmol/L); and zero for BGL ≥ 200mg/dL. The primary outcome was a composite of patient's death in ED or admission to intensive care unit (ICU). Receiver Operating Characteristic (ROC) curve analysis was used to evaluate the overall performance of T-RTS and of the combined RTS-G score. Among a total of 68,933 traumas, 9,436 patients (4,407 females) were enrolled, aged from 18 to 103 years; 4,288 were aged ≥65 years. A total of 577 (6.1%) met the primary endpoint: 38 patients died in ED (0.4%) and 539 patients were admitted to ICU. The T-RTS and BGL were independently associated to primary endpoint at multivariate analysis. The cumulative RTS-G score was significantly more accurate than T-RTS and reached the best accuracy in elderly patients. In general population, ROC area under curve (AUC) for T-RTS was 0.671 (95% CI, 0.661 - 0.680) compared to RTS-G ROC AUC 0.743 (95% CI, 0.734 - 0.752); P <.001. In patients ≥65 years, T-RTS ROC AUC was 0.671 (95% CI, 0.657 - 0.685) compared to RTS-G ROC AUC 0.780 (95% CI, 0.768 - 0.793); P <.001. Results showed RTS-G could be used effectively at ED triage for the risk stratification for death in ED and ICU admission of trauma patients, and it could reduce under-triage of approximately 20% compared to T-RTS. Comparing ROC AUCs, the combined RTS-G score performs significantly better than T-RTS and gives best results in patients ≥65 years.

We have previously developed a model-based method to compensate for the crosstalk in simultaneous acquisition of Tc-99m stress and Tl-201 rest myocardial perfusion SPECT. The purpose of this study is to evaluate the performance of this method in terms of a defect detection task using a mathematical Channelized Hoteling Observer (CHO) study. In this study we also optimized the iteration number in reconstruction of Tl distributions. A population of male and female NCAT phantoms were used to realistically model variation in patients. Both defect-free and defect-present data of Tc and Tl were simulated using the SimSET/PHG codes and our newly developed and validated angular response function (ARF) method to mimic simultaneous acquisition and separate acquisition. Poisson noise was modeled at clinically realistic levels. The Tl data were reconstructed using filtered backprojection (FBP) and ordered subset expectation maximization (OSEM) algorithms for up to 20 iterations with and without the model-based crosstalk compensation (MBC). CHO methodology and receiver operating characteristic (ROC) analysis were applied to short axis Tl images to obtain ROC curves and area under ROC curves (AUC). The AUC values were compared with those from separately acquired data reconstructed using FBP and OSEM algorithms. The results show that a relatively small number of iteration (/spl sim/3) is optimal for the postreconstruction filtering cutoff frequency of 0.16 pixel/sup -1/ used in this study. Also, the AUC values obtained with the MBC were significantly better than those without crosstalk compensation and those with the FBP reconstruction method applied to the separately acquired data.

目的 基于计算机的胸腔X线影像疾病检测和分类目前存在误诊率高，准确率低的问题。本文在视觉Transformer（vision Transformer，ViT）预训练模型的基础上，通过迁移学习方法，实现胸腔X线影像辅助诊断，提高诊断准确率和效率。方法 选用带有卷积神经网络（convolutional neural network，CNN）的ViT模型，其在超大规模自然图像数据集中进行了预训练；通过微调模型结构，使用预训练的ViT模型参数初始化主干网络，并迁移至胸腔X线影像数据集中再次训练，实现疾病多标签分类。结果 在IU X-Ray数据集中对ViT迁移学习前、后模型平均AUC（area under ROC curve）得分进行对比分析实验。结果表明，预训练ViT模型平均AUC得分为0.774，与不使用迁移学习相比提升了0.208。并针对模型结构和数据预处理进行了消融实验，对ViT中的注意力机制进行可视化，进一步验证了模型有效性。最后使用Chest X-Ray14和CheXpert数据集训练微调后的ViT模型，平均AUC得分为0.839和0.806，与对比方法相比分别有0.014～0.031的提升。结论 与其他方法相比，ViT模型胸腔X线影像的多标签分类精确度更高，且迁移学习可以在降低训练成本的同时提升ViT模型的分类性能和泛化性。消融实验与模型可视化表明，包含CNN结构的ViT模型能重点关注有意义的区域，高效获取胸腔X线影像的视觉特征。;Objective The chest X-ray-relevant screening and diagnostic method is essential for radiology nowadays. Most of chest X-ray images interpretation is still restricted by clinical experience and challenged for misdiagnose and missed diagnoses. To detect and identify one or more potential diseases in images automatically，it is beneficial for improving diagnostic efficiency and accuracy using computer-based technique. Compared to natural images，multiple lesions are challenged to be detected and distinguished accurately in a single image because abnormal areas have a small proportion and complex representations in chest X-ray images. Current convolutional neural network（CNN）based deep learning models have been widely used in the context of medical imaging. The structure of the CNN convolution kernel has sensitive to local detail information，and it is possible to extract richer image features. However，the convolution kernel cannot be used to get global information，and the features-extracted are restricted of redundant information like its relevance of background， muscles，and bones. The model’s performance in multi-label classification tasks are affected to a certain extent. At present，the vision Transformer（ViT）model has achieved its priorities in computer vision-related tasks. The ViT can be used to capture information simultaneously and effectively for multiple regions of the entire image. However，it is required to use large-scale dataset training to achieve good performance. Due to some factors like patient privacy and manual annotate costs，the size of the chest X-ray image data set has been limited. To reduce the model's dependence on data scale and improve the performance of multi-label classification，we develop the CNN-based ViT pre-training model in terms of the transfer learning method for diagnosis-assisted of chest X-ray image and multi-label classification. Method The CNN-based ViT model is pre-trained on a huge scale ground truth dataset，and it is used to obtain the initial parameters of the model. The model structure is fine-tuned according to the features of chest X-ray dataset. A 1×1 convolution layer is used to convert the chest X-ray images channels between 1 to 3. The number of output nodes of the linear layer in the classifier is balanced from 1 000 to the number of chest X-ray classification labels，and the Sigmoid is used as an activation function. The parameters of the backbone network are initialized in terms of the pre-trained ViT model parameters，and it is trained in the chest X-ray dataset after that to complete multi-label classification. The experiment is configured of Python3. 7 and PyTorch1. 8 to construct the model and RTX3090 GPU for training. Stochastic gradient descent（SGD）optimizer，binary cross-entropy（BCE）loss function，an initial learning rate of 1E-3，the cosine annealing learning rate decay are used. For training，each image is scaled to a size of 512×512 pixels，and a 224×224 pixels area and it is then cropped in random as the model input，and data augmentation is performed randomly by some of the flipping，perspective transformation， shearing，translation，zooming，and changing brightness. For testing，the chest X-ray image is scaled to 256×256 pixels and center crop a 224×224 area to input the trained model. Result The experiment is performed on the IU X-Ray，which is a small-scale chest X-ray dataset. This model is evaluated in quantitative using the average of area under ROC curve （AUC）scores across all classification labels. The results show that the average AUC score of the pre-trained ViT model is 0. 774. The accuracy and training efficiency of the non-pre-trained ViT model is dropped significantly. The average AUC score is reached to 0. 566 only，which is 0. 208 lower. In addition，the attention mechanism heat map is generated based on the ViT model，which can strengthen the interpretability of the model. A series of ablation experiments are carried out for data augmentation，model structure，and batch size design. The fine-tuned ViT model is trained on the Chest-Ray14 and CheXpert dataset as well. The average AUC score is reached to 0. 839 and 0. 806，which is optimized by 0. 014 and 0. 031. Conclusion A pre-trained ViT model is used for the multi-label classification of chest X-ray images via transfer learning. The experimental results illustrate that the ViT has its stronger multi-label classification performance in chest Xray images，and its attention mechanism is beneficial for lesions precision-focused like the interior of the chest cavity and the heart. Transfer learning is potential to improve the classification performance and model generalization of the ViT in small-scale datasets，and the training cost is reduced greatly. Ablation experiments demonstrate that the incorporated model of CNN and Transformer has its priority beyond single-structure model. Data enhancement and the batch size cutting can improve the performance of the model，but smaller scale of batch is still interlinked to longer training span. To improve the model's ability，we predict that future research direction can be focused on the extraction for complex disease and highlevel semantic information，such as their small lesions，disease location，and severity.

Cardio-respiratory disorders account for approximately 30% of total mortality in India, with their incidence steadily rising due to factors such as increasing urban pollution, sedentary lifestyles, and delayed clinical diagnosis, as highlighted by the Indian Council of Medical Research. Conventional auscultation using stethoscopes is highly dependent on the clinician’s expertise and experience, which can lead to variability and potential inaccuracies in diagnosis. To address these limitations, this study presents an automated cardio-respiratory sound classification framework designed to enhance early detection and diagnostic consistency through machine learning. The proposed system incorporates a user-friendly Graphical User Interface (GUI) with role-based functionality, allowing administrators to manage model training while enabling end-users to perform real-time predictions. Audio signals are preprocessed using the Librosa library to extract informative acoustic features such as Mel-Frequency Cepstral Coefficients (MFCC), Chroma features, and Mel Spectrogram representations, ensuring comprehensive characterization of both heart and lung sounds. A range of machine learning models, including Quadratic Discriminant Analysis (QDA), Gradient Boosting Classifier (GBC), Naïve Bayes Classifier (NBC), and Logistic Regression Classifier (LRC), are developed and comparatively evaluated. Furthermore, a hybrid deep learning architecture combining a Bi-directional Convolutional Neural Network (BiCNN) with a Bi-directional Gated Recurrent Unit (BiGRU) is introduced to effectively capture both spatial (spectral) and temporal dependencies within the audio data. The system is capable of performing dual classification by identifying both heart sound types and lung sound types simultaneously. Its performance is rigorously assessed using standard evaluation metrics, including accuracy, precision, recall, F1-score, and Receiver Operating Characteristic–Area Under Curve (ROCAUC), demonstrating its potential as a reliable and efficient tool for clinical decision support.

Commonly observed challenges in water quality anomaly detection using Machine Learning (ML) classifiers include unbalanced class distribution and missing data. Classifiers trained on such imbalanced datasets often exhibit biased accuracy, favoring the majority class and neglecting the minority class, while incomplete datasets limit the applicability of more complex models and hinder thorough analysis. This research addresses the handling of incomplete data and class imbalance by proposing a robust framework for an ML-based water quality anomaly detection system using several resampling techniques. A comparative study was conducted on six imputation methods for missing data, including Expectation Maximization (EM) and Multiple Imputation by Chained Equations (MICE), alongside three resampling techniques: Random Under Sampling (RUS), Rapidly Converging Gibbs (RACOG) sampler, and RACOG combined with RUS (RACOG-RUS). These methods were evaluated across three classifiers: Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Naïve Bayes (NB). The models were assessed using stratified 5-fold cross-validation and evaluated based on accuracy, Receiver Operating Characteristic Area Under Curve (ROC-AUC), and F1-score. Further experiments incorporated feature selection methods such as Boruta and Mean Decrease Accuracy (MDA) to optimize performance. Results demonstrate that RF combined with RACOG-RUS and EM achieved the highest F1-score of 0.9954, effectively addressing both class imbalance and missing data. Additionally, computational analysis highlights the efficiency of RF when optimized with appropriate hyperparameters.

Development without any defect is unsubstantial. Timely detection of software defects favors the proper resource utilization saving time, effort and money. With the increasing size and complexity of software, demand for accurate and efficient prediction models is increasing. Recently, search-based techniques (SBTs) have fascinated many researchers for Software Defect Prediction (SDP). The goal of this study is to conduct an empirical evaluation to assess the applicability of SBTs for predicting software defects in object-oriented (OO) softwares. In this study, 16 SBTs are exploited to build defect prediction models for 13 OO software projects. Stable performance measures — GMean, Balance and Receiver Operating Characteristic-Area Under Curve (ROC-AUC) are employed to probe into the predictive capability of developed models, taking into consideration the imbalanced nature of software datasets. Proper measures are taken to handle the stochastic behavior of SBTs. The significance of results is statistically validated using the Friedman test complied with Wilcoxon post hoc analysis. The results confirm that software defects can be detected in the early phases of software development with help of SBTs. This paper identifies the effective subset of SBTs that will aid software practitioners to timely detect the probable software defects, therefore, saving resources and bringing up good quality softwares. Eight SBTs — sUpervised Classification System (UCS), Bioinformatics-oriented hierarchical evolutionary learning (BIOHEL), CHC, Genetic Algorithm-based Classifier System with Adaptive Discretization Intervals (GA_ADI), Genetic Algorithm-based Classifier System with Intervalar Rule (GA_INT), Memetic Pittsburgh Learning Classifier System (MPLCS), Population-Based Incremental Learning (PBIL) and Steady-State Genetic Algorithm for Instance Selection (SGA) are found to be statistically good defect predictors.

Exploring the interrelationships between microbes and disease can help microbiologists make decisions and plan treatments. Predicting new microbe-disease associations currently relies on biological experiments and domain knowledge, which is time-consuming and inefficient. Automated algorithms are used to uncover the intrinsic link between microbes and disease. However, due to data noise and inadequate understanding of relevant biology, the efficient prediction of microbe-disease associations is still crucial. This study develops a multi-view graph augmentation convolutional network (MVGCNMDA) to predict potential disease-associated microbes. First, we use two data augmentation methods, edge perturbation and node dropping, to remove the data noise in the preprocessing stage. Second, we calculate Gaussian interaction profile kernel similarity and cosine similarity. Therefore, the Graph Convolutional Network(GCN) can fully use multi-view features. Then, the multi-view features are fed into the multi-attention block to learn the weights of different features adaptively. Finally, the embedding results are obtained using a Convolutional Neural Network (CNN) combiner, and the matrix completion is used to predict the relationship between potential microbes and diseases. We test our model on the Human microbe-disease Association Database (HMDAD), Disbiome, and the Combined Dataset (Peryton and MicroPhenoDB). The area under PR curve (AUPR), area under ROC curve (AUC), F1 score, and RECALL value are calculated to evaluate the performance of the developed MVGCNMDA. The AUPR is 0.9440, AUC is 0.9428, F1 score is 0.9383, and RECALL value is 0.8858. The experiments show that our model can accurately predict potential microbe-disease associations compared with the state-of-the-art works on the global Leave-One-Out-Cross-Validation (LOOCV) and the fivefold Cross-Validation (fivefold CV). To further verify the effectiveness of the proposed graph data augmentation, we designed five different settings in the ablation study. Furthermore, we present two case studies that validate the prediction of the potential association between microbes and diseases by MVGCNMDA.

In this paper, combining the feature extraction advantages of deep learning technology in image processing, we add the attention mechanism in deep convolutional neural network with a large convolutional kernel as the core of the module. And the multi-scale fusion module is designed to effectively fuse the large convolutional kernel CNN branch and the augmented Transformer branch to form an image segmentation method based on the large convolutional kernel and attention mechanism feature fusion network (TCNet). Setting up the experimental environment, the sample images were subjected to data enhancement operations of vertical flip, horizontal flip, diagonal flip, random shift and random scale change, and the parameters of accuracy (AC), sensitivity (SE), specificity (SP), Jaccard’s coefficient (JA), and Dice’s coefficient (DI) were selected as the evaluation indexes, and the model performance testing was carried out by using several classical datasets. The TCNet model proposed in this paper achieves 90.334% JA, and the indexes of pixel-AC and DI are 93.457% and 91.773%, respectively, and the model has better segmentation performance. In this paper, the TCNet model is designed to achieve the extraction of global context information and local detail information of the image by using the multiple attention mechanism, which improves the image segmentation performance and has a wider scope of application.

Recent years have witnessed the rapid growth of event-based social networks (EBSNs) such as Plancast and DoubanEvent. In these EBSNs, followee recommendation which recommends new users to follow can bring great benefits to both users and service providers. In this paper, we focus on the problem of followee recommendation in EBSNs. However, the sparsity and imbalance of the social relations in EBSNs make this problem very challenging. Therefore, by exploiting the heterogeneous nature of EBSNs, we propose a new method called Heterogenous Network based Followee Recommendation (HNFR) for our problem. In the HNFR method, to relieve the problem of data sparsity, we combine the explicit and latent features captured from both the online social network and the offline event participation network of an EBSN. Moreover, to overcome the problem of data imbalance, we propose a Bayesian optimization framework which adopts pairwise user preference on both the social relations and the events, and aims to optimize the area under ROC curve (AUC). The experiments on real-world data demonstrate the effectiveness of our method.

This paper presents a study on crime classification using two 3D deep learning algorithms, i.e., 3D Convolutional Neural Network and the 3D Residual Network. The Chicago crime dataset, which has 7.29 million records, collected from 2001 to 2020, is used for training the models. The models are evaluated by using F1 score, Area Under Receiver Operator Curve (AUROC), and Area Under Curve - Precision Recall (AUCPR). Furthermore, the effectiveness of spatial grid resolutions on the performance of the models is also evaluated. Results show that the 3D ResNet-18 achieved the best performance with an F1 score of 0.9985, whereas the 3D CNN achieved an F1 score of 0.9979, during training with a spatial resolution of 16 pixels. Furthermore, the 3D ResNet-18 achieved an accuracy of 0.92 and the 3D CNN achieved an accuracy of 0.87 during model testing. In terms of future work, we intend to test these algorithms on multi-label classification and regression crime problems, improve the performance of the 3D CNN by adding RNN layers, and evaluate the implementation of 3D ResNeXt for crime prediction and classification.

Knee osteoarthritis is one of the most common musculoskeletal diseases and is usually diagnosed with medical imaging techniques. Conventionally, case identification using plain radiography is practiced. However, we acknowledge that knee osteoarthritis is a 3D complexity; hence, magnetic resonance imaging will be the ideal modality to reveal the hidden osteoarthritis features from a three-dimensional view. In this work, the feasibility of well-known convolutional neural network (CNN) structures (ResNet, DenseNet, VGG, and AlexNet) to distinguish knees with and without osteoarthritis (OA) is investigated. Using 3D convolutional layers, we demonstrated the potential of 3D convolutional neural networks of 13 different architectures in knee osteoarthritis diagnosis. We used transfer learning by transforming 2D pre-trained weights into 3D as initial weights for the training of the 3D models. The performance of the models was compared and evaluated based on the performance metrics [balanced accuracy, precision, F1 score, and area under receiver operating characteristic (AUC) curve]. This study suggested that transfer learning indeed enhanced the performance of the models, especially for ResNet and DenseNet models. Transfer learning-based models presented promising results, with ResNet34 achieving the best overall accuracy of 0.875 and an F1 score of 0.871. The results also showed that shallow networks yielded better performance than deeper neural networks, demonstrated by ResNet18, DenseNet121, and VGG11 with AUC values of 0.945, 0.914, and 0.928, respectively. This encourages the application of clinical diagnostic aid for knee osteoarthritis using 3DCNN even in limited hardware conditions.