Diagnostic Model Research Articles

Risk factors of non-arteritic anterior ischaemic optic neuropathy and central retinal artery occlusion.

To investigate the difference in risk factors between non-arteritic anterior ischaemic optic neuropathy (NAION) and central retinal artery occlusion (CRAO) and develop a predictive diagnostic nomogram. The study included 37 patients with monocular NAION, 20 with monocular CRAO, and 24 with hypertension. Gender, age, and systemic diseases were recorded. Blood routine, lipids, hemorheology, carotid and brachial artery doppler ultrasound, and echocardiography were collected. The optic disc area, cup area, and cup-to-disc ratio (C/D) of the unaffected eye in the NAION and CRAO group and the right eye in the hypertension group were measured. The carotid artery intimal medial thickness (C-IMT) of the affected side of the CRAO group was thicker (P=0.039) and its flow-mediated dilation (FMD) was lower (P=0.049) than the NAION group. Compared with hypertension patients, NAION patients had higher whole blood reduced viscosity low-shear (WBRV-L) and erythrocyte aggregation index (EAI; P=0.045, 0.037), and CRAO patients had higher index of rigidity of erythrocyte (IR) and erythrocyte deformation index (EDI; P=0.004, 0.001). The optic cup and the C/D of the NAION group were smaller than the other two groups (P<0.0001). The diagnostic prediction model showed high diagnostic specificity (83.7%) and sensitivity (85.6%), which was highly related to hypertension, the C-IMT of the affected side, FMD, platelet (PLT), EAI, and C/D. CRAO patients show thicker C-IMT and worse endothelial function than NAION. NAION and CRAO may be related to abnormal hemorheology. A small cup and small C/D may be involved in NAION. The diagnostic nomogram can be used to preliminarily identify NAION and CRAO.

ESG guidance and artificial intelligence support for power systems analytics in the energy industry.

In order to increase the precision and effectiveness of power system analysis and fault diagnosis, this study aims to assess the power systems in the energy sector while utilizing artificial intelligence (AI) and environmental social governance (ESG). First, the ESG framework is presented in this study to fully account for the effects of the power system on the environment, society, and governance. Second, to coordinate the operation of various components and guarantee the balance and security of the power system, the CNN-BiLSTM power load demand forecasting model is built by merging convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM). Lastly, the particle swarm optimization (PSO) algorithm is used to introduce and optimize the deep belief network (DBN), and a power grid fault diagnostic model is implemented using the PSO technique and DBN. The model's performance is assessed through experimentation. The outcomes demonstrate how the CNN-BiLSTM algorithm significantly increases forecasting accuracy while overcoming the drawback of just having one dimension of power load data. The values of 0.054, 0.076, and 0.102, respectively, are the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). Effective processing of large-scale nonlinear data is achieved in the area of power grid fault diagnosis, resulting in prediction accuracy of 96.22% and prediction time of only 129.94s. This is clearly better than other algorithms and increases fault prediction efficiency and accuracy. Consequently, the model presented in this study not only produces impressive results in fault diagnosis and load demand forecasting, but also advances the field of power system analysis in the energy industry and offers a significant amount of support for the sustainable and intelligent growth of the energy industry.

Deciphering smooth muscle cell heterogeneity in atherosclerotic plaques and constructing model: a multi-omics approach with focus on KLF15/IGFBP4 axis.

Ruptured atherosclerotic plaques often precipitate severe ischemic events, such as stroke and myocardial infarction. Unraveling the intricate molecular mechanisms governing vascular smooth muscle cell (VSMC) behavior in plaque stabilization remains a formidable challenge. In this study, we leveraged single-cell and transcriptomic datasets from atherosclerotic plaques retrieved from the gene expression omnibus (GEO) database. Employing a combination of single-cell population differential analysis, weighted gene co-expression network analysis (WGCNA), and transcriptome differential analysis techniques, we identified specific genes steering the transformation of VSMCs in atherosclerotic plaques. Diagnostic models were developed and validated through gene intersection, utilizing the least absolute shrinkage and selection operator (LASSO) and random forest (RF) methods. Nomograms for plaque assessment were constructed. Tissue localization and expression validation were performed on specimens from animal models, utilizing immunofluorescence co-localization, western blot, and reverse-transcription quantitative-polymerase chain reaction (RT-qPCR). Various online databases were harnessed to predict transcription factors (TFs) and their interacting compounds, with determination of the cell-specific localization of TF expression using single-cell data. Following rigorous quality control procedures, we obtained a total of 40,953 cells, with 6,261 representing VSMCs. The VSMC population was subsequently clustered into 5 distinct subpopulations. Analyzing inter-subpopulation cellular communication, we focused on the SMC2 and SMC5 subpopulations. Single-cell subpopulation and WGCNA analyses revealed significant module enrichments, notably in collagen-containing extracellular matrix and cell-substrate junctions. Insulin-like growth factor binding protein 4 (IGFBP4), apolipoprotein E (APOE), and cathepsin C (CTSC) were identified as potential diagnostic markers for early and advanced plaques. Notably, gene expression pattern analysis suggested that IGFBP4 might serve as a protective gene, a hypothesis validated through tissue localization and expression analysis. Finally, we predicted TFs capable of binding to IGFBP4, with Krüppel-like family 15 (KLF15) emerging as a prominent candidate showing relative specificity within smooth muscle cells. Predictions about compounds associated with affecting KLF15 expression were also made. Our study established a plaque diagnostic and assessment model and analyzed the molecular interaction mechanisms of smooth muscle cells within plaques. Further analysis revealed that the transcription factor KLF15 may regulate the biological behaviors of smooth muscle cells through the KLF15/IGFBP4 axis, thereby influencing the stability of advanced plaques via modulation of the PI3K-AKT signaling pathway. This could potentially serve as a target for plaque stability assessment and therapy, thus driving advancements in the management and treatment of atherosclerotic plaques.

Longitudinal Artificial Intelligence-Based Deep Learning Models for Diagnosis and Prediction of the future occurrence of Polyneuropathy in Diabetes and prediabetes

The objective of this study was to develop artificial intelligence-based deep learning models and assess their potential utility and accuracy in diagnosing and predicting the future occurrence of diabetic distal sensorimotor polyneuropathy (DSPN) among individuals with type 2 diabetes mellitus (T2DM) and prediabetes. In 394 patients (T2DM=300, Prediabetes=94), we developed a DSPN diagnostic and predictive model using Random Forest (RF)-based variable selection techniques, specifically incorporating the combined capabilities of the Clinical Toronto Neuropathy Score (TCNS) and nerve conduction study (NCS) to identify relevant variables. These important variables were then integrated into a deep learning framework comprising Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. To evaluate temporal predictive efficacy, patients were assessed at enrollment and one-year follow-up. RF-based variable selection identified key factors for diagnosing DSPN. Numbness scores, sensory test results (vibration), reflexes (knee, ankle), sural nerve attributes (sensory nerve action potential [SNAP] amplitude, nerve conduction velocity [NCV], latency), and peroneal/tibial motor NCV were candidate variables at baseline and over one year. Tibial compound motor action potential amplitudes were used for initial diagnosis, and ulnar SNAP amplitude for subsequent diagnoses. CNNs and LSTMs achieved impressive AUC values of 0.98 for DSPN diagnosis prediction, and 0.93 and 0.89 respectively for predicting the future occurrence of DSPN. RF techniques combined with two deep learning algorithms exhibited outstanding performance in diagnosing and predicting the future occurrence of DSPN. These algorithms have the potential to serve as surrogate measures, aiding clinicians in accurate diagnosis and future prediction of DSPN.

Integrated analysis reveals critical cisplatin-resistance regulators E2F7 contributed to tumor progression and metastasis in lung adenocarcinoma.

Drug resistance poses a significant challenge in cancer treatment, particularly as a leading cause of therapy failure. Cisplatin, the primary drug for lung adenocarcinoma (LUAD) chemotherapy, shows effective treatment outcomes. However, the development of resistance against cisplatin is a major obstacle. Therefore, identifying genes resistant to cisplatin and adopting personalized treatment could significantly improve patient outcomes. By examining transcriptome data of cisplatin-resistant LUAD cells from the GEO database, 181 genes associated with cisplatin resistance were identified. Using univariate regression analysis, random forest and multivariate regression analyses, two prognostic genes, E2F7 and FAM83A, were identified. This study developed a prognostic model utilizing E2F7 and FAM83A as key indicators. The Cell Counting Kit 8 assay, Transwell assay, and flow cytometry were used to detect the effects of E2F7 on the proliferation, migration, invasiveness and apoptosis of A549/PC9 cells. Western blotting was used to determine the effect of E2F7 on AKT/mTOR signaling pathway. This study has pinpointed two crucial genes associated with cisplatin resistance, E2F7 and FAM83A, and developed a comprehensive model to assist in the diagnosis, prognosis, and evaluation of relapse risk in LUAD. Analysis revealed that patients at higher risk, according to these genetic markers, had elevated levels of immune checkpoints (PD-L1 and PD-L2). The prognostic and diagnosis values of E2F7 and FAM83A were further confirmed in clinical data. Furthermore, inhibiting E2F7 in lung cancer cells markedly reduced their proliferation, migration, invasion, and increased apoptosis. In vivo experiments corroborated these findings, showing reduced tumor growth and lung metastasis upon E2F7 suppression in lung cancer models. Our study affirms the prognostic value of a model based on two DEGs, offering a reliable method for predicting the success of tumor immunotherapy in patients with LUAD. The diagnostic and predictive model based on these genes demonstrates excellent performance. In vitro, reducing E2F7 levels shows antitumor effects by blocking LUAD growth and progression. Further investigation into the molecular mechanisms has highlighted E2F7's effect on the AKT/mTOR signaling pathway, underscoring its therapeutic potential. In the era of personalized medicine, this DEG-based model promises to guide clinical practice.

Bronchopulmonary dysplasia: analysis and validation of ferroptosis-related diagnostic biomarkers and immune cell infiltration features.

Early and precise diagnosis of bronchopulmonary dysplasia (BPD) is essential to improve the prognosis of preterm infants with BPD. Studying ferroptosis-related genes for diagnostic markers of BPD was the objective of this study. Using the GEO database and the FerrDb database, we obtained the GSE32472 dataset and screened the ferroptosis-related differentially expressed mRNAs (FRDE-mRNAs). By using Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG), possible biological functions and pathways were identified for FRDE-mRNAs. Three machine learning algorithms (LASSO, SVM-RFE, Random Forest) were used to recognize hub genes, as well as CIBERSORT for exploring the immune landscape of BPD and controls. Functional predictions for hub genes were made using single-gene gene set enrichment analysis (GSEA). Twenty three FRDE-mRNAs were obtained and were mainly involved in autophagy, fatty acid metabolism and ferroptosis. The four hub genes (LPIN1, ACADSB, WIPI1 and SLC7A11) screened were utilized to construct a diagnostic nomogram. The receiver operating characteristic (ROC) curves and calibration curves demonstrateld that the nomogram exhibited good predictive performance. Eight types of immune cell markers differed significantly between BPD and controls. We developed a diagnostic model for BPD, which could facilitate the early diagnosis and timely intervention of BPD. The role of ferroptosis in bronchopulmonary dysplasia is rarely reported. The ferroptosis-related genes (LPIN1, ACADSB, WIPI1 and SLC7A11) we identified could serve as early diagnostic biomarkers for BPD. Immune cell infiltration features in BPD and signaling pathways associated with marker genes give new insight into the disease process and provide a basis for further research.

Energy wavelet signal processed ECG and standard 12 lead ECG: Diagnosis of early diastolic dysfunction.

Left ventricular (LV) diastolic dysfunction (LVDD) is the result of impaired LV relaxation and identifies those at risk of developing heart failure. Echocardiography has been used as the gold standard to identify early LVDD. The signal processed electrocardiogram (hsECG) has demonstrated effectiveness to detect early LVDD. Whether or not the standard 12‑lead electrocardiogram (ECG) can accurately predict early LVDD is not known. A standard 12‑lead ECG including signal processing (hsECG) was performed in 569 patients. Patients with atrial fibrillation, bundle branch block, pre-excitation, left ventricular hypertrophy or known cardiovascular disease were excluded, leaving 464 examinations for analysis. Early LVDD was diagnosed by established methods using echocardiography. Repolarization abnormalities (T wave discordance) in V1, V6, I and aVL and the hsECG were compared to the echocardiographic findings to establish diagnostic accuracy. A total of 84 (18.1%) patients were diagnosed with early LVDD. A combination of a borderline or abnormal finding on the hsECG produced the best diagnostic model (sensitivity 84.5%, specificity 47.9%). The best performing ECG lead was V1 with a sensitivity of 38.1% and specificity of 92.1%. Regression analysis demonstrated increasing age and V1 to be predictive of LVDD. The hsECG displayed reasonable ability to detect early LVDD. Other than V1, repolarization abnormalities on the standard 12‑lead ECG did not. While lead V1 showed promise in detecting LVDD, whether this or any other simple ECG variable can predict future LVDD would be of further interest.

Updates from the International League Against Epilepsy Classification of Epilepsy (2017) and Focal Cortical Dysplasias (2022): Imaging Phenotype and Genetic Characterization.

The International League Against Epilepsy (ILAE) is an organization of 120 national chapters providing the most widely accepted and updated guidelines on epilepsy. In 2022, the ILAE Task Force revised the prior (2011) classification of focal cortical dysplasias to incorporate and update clinicopathologic and genetic information, with the aim to provide an objective classification scheme. New molecular-genetic information has led to the concept of "integrated diagnosis" on the same lines as brain tumors, with a multilayered diagnostic model providing a phenotype-genotype integration. Major changes in the new update were made to type II focal cortical dysplasias, apart from identification of new entities, such as mild malformations of cortical development and cortical malformation with oligodendroglial hyperplasia. No major changes were made to type I and III focal cortical dysplasias, given the lack of significant new genetic information. This review provides the latest update on changes to the classification of focal cortical dysplasias with discussion about the new entities. The ILAE in 2017 updated the classification of seizure and epilepsy with 3 levels of diagnosis, including seizure type, epilepsy type, and epilepsy syndrome, which are also briefly discussed here.

A fault diagnosis framework based on heterogeneous ensemble learning for air conditioning chiller with unbalanced samples

Big data-based air conditioning fault diagnosis research has developed rapidly in recent years, but in actual engineering, the fault sample size of air conditioning systems is much smaller than the normal sample size, and the resulting sample imbalance problem makes conventional data-driven diagnostic methods based on low accuracy and poor stability. In order to solve the problem of unbalanced fault diagnosis of air-conditioning chillers, this paper proposes an integrated learning-based diagnostic model, which achieves diagnosis by combining multiple base models and by majority voting. The method uses four classification models, namely, random forest model, decision tree model, k nearest neighbor model, and isomorphic integration model, as base classifiers, and synthesizes the four base classifiers into a heterogeneous integration algorithmic model (IMV) through integrated learning, and performs diagnostic detection of seven types of typical faults of chiller units using the majority voting method of integrated learning. The effectiveness of the proposed algorithm is verified on the RP-1043 dataset, and the experimental results show that the accuracy of the heterogeneous integrated algorithm model (IMV) can reach 96.87%, which is a significant improvement compared with the accuracy of the other four base classifier models (81.04%–96.25%). Therefore, the integrated learning model has some application prospects in fault diagnosis when targeting unbalanced datasets.

A ferroptosis-related signature predicts the clinical diagnosis and prognosis, and associates with the immune microenvironment of lung cancer

Targeting ferroptosis-related pathway is a potential strategy for treatment of lung cancer (LC). Consequently, exploration of ferroptosis-related markers is important for treating LC. We collected LC clinical data and mRNA expression profiles from TCGA and GEO database. Ferroptosis-related genes (FRGs) were obtained through FerrDB database. Expression analysis was performed to obtain differentially expressed FRGs. Diagnostic and prognostic models were constructed based on FRGs by LASSO regression, univariate, and multivariate Cox regression analysis, respectively. External verification cohorts GSE72094 and GSE157011 were used for validation. The interrelationship between prognostic risk scores based on FRGs and the tumor immune microenvironment was analyzed. Immunocytochemistry, Western blotting, and RT-qPCR detected the FRGs level. Eighteen FRGs were used for diagnostic models, 8 FRGs were used for prognostic models. The diagnostic model distinguished well between LC and normal samples in training and validation cohorts of TCGA. The prognostic models for TCGA, GSE72094, and GSE157011 cohorts significantly confirmed lower overall survival (OS) in high-risk group, which demonstrated excellent predictive properties of the survival model. Multivariate Cox regression analysis further confirmed risk score was an independent risk factor related with OS. Immunoassays revealed that in high-risk group, a significantly higher proportion of Macrophages_M0, Neutrophils, resting Natural killer cells and activated Mast cells and the level of B7H3, CD112, CD155, B7H5, and ICOSL were increased. In conclusion, diagnostic and prognostic models provided superior diagnostic and predictive power for LC and revealed a potential link between ferroptosis and TIME.

Diagnosis of cervical lymphoma using a YOLO-v7-based model with transfer learning

To investigate the ability of an auxiliary diagnostic model based on the YOLO-v7-based model in the classification of cervical lymphadenopathy images and compare its performance against qualitative visual evaluation by experienced radiologists. Three types of lymph nodes were sampled randomly but not uniformly. The dataset was randomly divided into for training, validation, and testing. The model was constructed with PyTorch. It was trained and weighting parameters were tuned on the validation set. Diagnostic performance was compared with that of the radiologists on the testing set. The mAP of the model was 96.4% at the 50% intersection-over-union threshold. The accuracy values of it were 0.962 for benign lymph nodes, 0.982 for lymphomas, and 0.960 for metastatic lymph nodes. The precision values of it were 0.928 for benign lymph nodes, 0.975 for lymphomas, and 0.927 for metastatic lymph nodes. The accuracy values of radiologists were 0.659 for benign lymph nodes, 0.836 for lymphomas, and 0.580 for metastatic lymph nodes. The precision values of radiologists were 0.478 for benign lymph nodes, 0.329 for lymphomas, and 0.596 for metastatic lymph nodes. The model effectively classifies lymphadenopathies from ultrasound images and outperforms qualitative visual evaluation by experienced radiologists in differential diagnosis.

Cross-modality transfer learning with knowledge infusion for diabetic retinopathy grading

BackgroundUltra-wide-field (UWF) fundus photography represents an emerging retinal imaging technique offering a broader field of view, thus enhancing its utility in screening and diagnosing various eye diseases, notably diabetic retinopathy (DR). However, the application of computer-aided diagnosis for DR using UWF images confronts two major challenges. The first challenge arises from the limited availability of labeled UWF data, making it daunting to train diagnostic models due to the high cost associated with manual annotation of medical images. Secondly, existing models' performance requires enhancement due to the absence of prior knowledge to guide the learning process.PurposeBy leveraging extensively annotated datasets within the field, which encompass large-scale, high-quality color fundus image datasets annotated at either image-level or pixel-level, our objective is to transfer knowledge from these datasets to our target domain through unsupervised domain adaptation.MethodsOur approach presents a robust model for assessing the severity of diabetic retinopathy (DR) by leveraging unsupervised lesion-aware domain adaptation in ultra-wide-field (UWF) images. Furthermore, to harness the wealth of detailed annotations in publicly available color fundus image datasets, we integrate an adversarial lesion map generator. This generator supplements the grading model by incorporating auxiliary lesion information, drawing inspiration from the clinical methodology of evaluating DR severity by identifying and quantifying associated lesions.ResultsWe conducted both quantitative and qualitative evaluations of our proposed method. In particular, among the six representative DR grading methods, our approach achieved an accuracy (ACC) of 68.18% and a precision (pre) of 67.43%. Additionally, we conducted extensive experiments in ablation studies to validate the effectiveness of each component of our proposed method.ConclusionIn conclusion, our method not only improves the accuracy of DR grading, but also enhances the interpretability of the results, providing clinicians with a reliable DR grading scheme.

Screening and diagnosis of cardiovascular disease using artificial intelligence-enabled cardiac magnetic resonance imaging.

Cardiac magnetic resonance imaging (CMR) is the gold standard for cardiac function assessment and plays a crucial role in diagnosing cardiovascular disease (CVD). However, its widespread application has been limited by the heavy resource burden of CMR interpretation. Here, to address this challenge, we developed and validated computerized CMR interpretation for screening and diagnosis of 11 types of CVD in 9,719 patients. We propose a two-stage paradigm consisting of noninvasive cine-based CVD screening followed by cine and late gadolinium enhancement-based diagnosis. The screening and diagnostic models achieved high performance (area under the curve of 0.988 ± 0.3% and 0.991 ± 0.0%, respectively) in both internal and external datasets. Furthermore, the diagnostic model outperformed cardiologists in diagnosing pulmonary arterial hypertension, demonstrating the ability of artificial intelligence-enabled CMR to detect previously unidentified CMR features. This proof-of-concept study holds the potential to substantially advance the efficiency and scalability of CMR interpretation, thereby improving CVD screening and diagnosis.

Diagnostic implications of ubiquitination-related gene signatures in Alzheimer's disease

The purpose of this study was to explore the diagnostic implications of ubiquitination-related gene signatures in Alzheimer's disease. In this study, we first collected 161 samples from the GEO database (including 87 in the AD group and 74 in the normal group). Subsequently, through differential expression analysis and the iUUCD 2.0 database, we obtained 3450 Differentially Expressed Genes (DEGs) and 806 Ubiquitin-related genes (UbRGs). After taking the intersection, we obtained 128 UbR-DEGs. Secondly, by conducting GO and KEGG enrichment analysis on these 128 UbR-DEGs, we identified the main molecular functions and biological pathways related to AD. Furthermore, through the utilization of GSEA analysis, we have gained insight into the enrichment of functions and pathways within both the AD and normal groups. Further, using lasso regression analysis and cross-validation techniques, we identified 22 characteristic genes associated with AD. Subsequently, we constructed a logistic regression model and optimized it, resulting in the identification of 6 RUbR-DEGs: KLHL21, WDR82, DTX3L, UBTD2, CISH, and ATXN3L. In addition, the ROC result showed that the diagnostic model we built has excellent accuracy and reliability in identifying AD patients. Finally, we constructed a lncRNA-miRNA-mRNA (competing endogenous RNA, ceRNA) regulatory network for AD based on six RUbR-DEGs, further elucidating the interaction between UbRGs and lncRNA, miRNA. In conclusion, our findings will contribute to further understanding of the molecular pathogenesis of AD and provide a new perspective for AD risk prediction, early diagnosis and targeted therapy in the population.

Bioinformatics analysis for constructing a cellular senescence-related age-related macular degeneration diagnostic model and identifying relevant disease subtypes to guide treatment.

Age-related macular degeneration (AMD) is a condition causing progressive central vision loss. Growing evidence suggests a link between cellular senescence and AMD. However, the exact mechanism by which cellular senescence leads to AMD remains unclear. Employing machine learning, we established an AMD diagnostic model. Through unsupervised clustering, two distinct AMD subtypes were identified. GO, KEGG, and GSVA analyses explored the diverse biological functions associated with the two subtypes. By WGCNA, we constructed a coexpression network of differential genes between the subtypes, revealing the regulatory role of hub genes at the level of transcription factors and miRNAs. We identified 5 genes associated with inflammation for the construction of the AMD diagnostic model. Additionally, we observed that the level of cellular senescence and pathways related to programmed cell death (PCD), such as ferroptosis, necroptosis, and pyroptosis, exhibited higher expression levels in subtype B than A. Immune microenvironments also differed between the subtypes, indicating potentially distinct pathogenic mechanisms and therapeutic targets. In summary, by leveraging cellular senescence-associated gene expression, we developed an AMD diagnostic model. Furthermore, we identified two subtypes with varying expression patterns of senescence genes, revealing their differential roles in programmed cell death, disease progression, and immune microenvironments within AMD.

Development and validation of a clinical prediction model for glioma grade using machine learning.

Histopathological evaluation is currently the gold standard for grading gliomas; however, this technique is invasive. This study aimed to develop and validate a diagnostic prediction model for glioma by employing multiple machine learning algorithms to identify risk factors associated with high-grade glioma, facilitating the prediction of glioma grading. Data from 1114 eligible glioma patients were obtained from The Cancer Genome Atlas (TCGA) database, which was divided into a training set (n= 781) and a test set (n= 333). Fifty machine learning algorithms were employed, and the optimal algorithm was selected to construct a prediction model. The performance of the machine learning prediction model was compared to the clinical prediction model in terms of discrimination, calibration, and clinical validity to assess the performance of the prediction model. The area under the curve (AUC) values of the machine learning prediction models (training set: 0.870 vs. 0.740, test set: 0.863 vs. 0.718) were significantly improved from the clinical prediction models. Furthermore, significant improvement in discrimination was observed for the Integrated Discrimination Improvement (IDI) (training set: 0.230, test set: 0.270) and Net Reclassification Index (NRI) (training set: 0.170, test set: 0.170) from the clinical prognostic model. Both models showed a high goodness of fit and an increased net benefit. A strong prediction accuracy model can be developed using machine learning algorithms to screen for high-grade glioma risk predictors, which can serve as a non-invasive prediction tool for preoperative diagnostic grading of glioma.

Identification of DNA methylation characteristics associated with metastasis and prognosis in colorectal cancer

Colorectal cancer (CRC) is prone to metastasis and recurrence after surgery, which is one of the main causes for its poor treatment and prognosis. Therefore, it is essential to identify biomarkers associated with metastasis and recurrence in CRC. DNA methylation has a regulatory role in cancer metastasis, tumor immune microenvironment (TME), and prognosis and may be one of the most valuable biomarkers for predicting CRC metastasis and prognosis. We constructed a diagnostic model and nomogram that can effectively predict CRC metastasis based on the differential methylation CpG sites (DMCs) between metastatic and non-metastatic CRC patients. Then, we identified 17 DMCs associated with progression free survival (PFS) of CRC and constructed a prognostic model. The prognosis model based on 17 DMCs can predict the PFS of CRC with medium to high accuracy. The results of immunohistochemical analysis indicated that the protein expression levels of the genes involved in prognostic DMCs were different between normal and colorectal cancer tissues. According to the results of immune-related analysis, we found that the low-risk patients had better immunotherapy response. In addition, high risk scores were negatively correlated with high tumor mutation burden (TMB) levels, and patients with low TMB levels in the high-risk group had the worst PFS. Our work shows the clinical value of DNA methylation in predicting CRC metastasis and PFS, as well as their correlation with TME, immunotherapy, and TMB, which helps understand the changes of DNA methylation in CRC metastasis and improving the treatment and prognosis of CRC.

An efficient deep neural network for automatic classification of acute intracranial hemorrhages in brain CT scans

BackgroundRecent advancements in deep learning models have demonstrated their potential in the field of medical imaging, achieving remarkable performance surpassing human capabilities in tasks such as classification and segmentation. However, these modern state-of-the-art network architectures often demand substantial computational resources, which limits their practical application in resource-constrained settings. This study aims to propose an efficient diagnostic deep learning model specifically designed for the classification of intracranial hemorrhage in brain CT scans. MethodOur proposed model utilizes a combination of depthwise separable convolutions and a multi-receptive field mechanism to achieve a trade-off between performance and computational efficiency. The model was trained on RSNA datasets and validated on CQ500 dataset and PhysioNet dataset. ResultThrough a comprehensive comparison with state-of-the-art models, our model achieves an average AUROC score of 0.952 on RSNA datasets and exhibits robust generalization capabilities, comparable to SE-ResNeXt, across other open datasets. Furthermore, the parameter count of our model is just 3 % of that of MobileNet V3. ConclusionThis study presents a diagnostic deep-learning model that is optimized for classifying intracranial hemorrhages in brain CT scans. The efficient characteristics make our proposed model highly promising for broader applications in medical settings.

Identification and analysis of chemokine-related and NETosis-related genes in acute pancreatitis to develop a predictive model

Background: Chemokines and NETosis are significant contributors to the inflammatory response, yet there still needs to be a more comprehensive understanding regarding the specific molecular characteristics and interactions of NETosis and chemokines in the context of acute pancreatitis (AP) and severe AP (SAP).Methods: To address this gap, the mRNA expression profile dataset GSE194331 was utilized for analysis, comprising 87 AP samples (77 non-SAP and 10 SAP) and 32 healthy control samples. Enrichment analyses were conducted for differentially expressed chemokine-related genes (DECRGs) and NETosis-related genes (DENRGs). Three machine-learning algorithms were used for the identification of signature genes, which were subsequently utilized in the development and validation of nomogram diagnostic models for the prediction of AP and SAP. Furthermore, single-gene Gene Set Enrichment Analysis (GSEA) and Gene Set Variation Analysis (GSVA) were performed. Lastly, an interaction network for the identified signature genes was constructed.Results: We identified 12 DECRGs and 7 DENRGs, and enrichment analyses indicated they were primarily enriched in cytokine-cytokine receptor interaction, chemokine signaling pathway, TNF signaling pathway, and T cell receptor signaling pathway. Moreover, these machine learning algorithms finally recognized three signature genes (S100A8, AIF1, and IL18). Utilizing the identified signature genes, we developed nomogram models with high predictive accuracy for AP and differentiation of SAP from non-SAP, as demonstrated by area under the curve (AUC) values of 0.968 (95% CI 0.937–0.990) and 0.862 (95% CI 0.742–0.955), respectively, in receiver operating characteristic (ROC) curve analysis. Subsequent single-gene GESA and GSVA indicated a significant positive correlation between these signature genes and the proteasome complex. At the same time, a negative association was observed with the Th1 and Th2 cell differentiation signaling pathways.Conclusion: We have identified three genes (S100A8, AIF1, and IL18) related to chemokines and NETosis, and have developed accurate diagnostic models that might provide a novel method for diagnosing AP and differentiating between severe and non-severe cases.

Early identification of macrophage activation syndrome secondary to systemic lupus erythematosus with machine learning

ObjectiveThe macrophage activation syndrome (MAS) secondary to systemic lupus erythematosus (SLE) is a severe and life-threatening complication. Early diagnosis of MAS is particularly challenging. In this study, machine learning models and diagnostic scoring card were developed to aid in clinical decision-making using clinical characteristics.MethodsWe retrospectively collected clinical data from 188 patients with either SLE or the MAS secondary to SLE. 13 significant clinical predictor variables were filtered out using the Least Absolute Shrinkage and Selection Operator (LASSO). These variables were subsequently utilized as inputs in five machine learning models. The performance of the models was evaluated using the area under the receiver operating characteristic curve (ROC-AUC), F1 score, and F2 score. To enhance clinical usability, we developed a diagnostic scoring card based on logistic regression (LR) analysis and Chi-Square binning, establishing probability thresholds and stratification for the card. Additionally, this study collected data from four other domestic hospitals for external validation.ResultsAmong all the machine learning models, the LR model demonstrates the highest level of performance in internal validation, achieving a ROC-AUC of 0.998, an F1 score of 0.96, and an F2 score of 0.952. The score card we constructed identifies the probability threshold at a score of 49, achieving a ROC-AUC of 0.994 and an F2 score of 0.936. The score results were categorized into five groups based on diagnostic probability: extremely low (below 5%), low (5–25%), normal (25–75%), high (75–95%), and extremely high (above 95%). During external validation, the performance evaluation revealed that the Support Vector Machine (SVM) model outperformed other models with an AUC value of 0.947, and the scorecard model has an AUC of 0.915. Additionally, we have established an online assessment system for early identification of MAS secondary to SLE.ConclusionMachine learning models can significantly improve the diagnostic accuracy of MAS secondary to SLE, and the diagnostic scorecard model can facilitate personalized probabilistic predictions of disease occurrence in clinical environments.