Clinical Datasets Research Articles

Uncovering Disease Mechanisms through AI and Visual Computing: A Multi-Scale Approach to Biological Data Interpretation

This research focuses on uncovering complex disease mechanisms by employing artificial intelligence (AI) and visual computing to interpret multi-scale biological data. By analyzing data from molecular to organ-level systems, this study aims to bridge the gap between cellular insights and clinical applications, offering new perspectives on diseases such as cancer, neurodegenerative disorders, and cardiovascular diseases. Through a comprehensive approach to data integration, the project compiles omics and imaging data to enable holistic analysis across biological scales. The use of hierarchical neural networks, graph neural networks (GNNs), and convolutional neural networks (CNNs) allows for multi-dimensional modeling and the extraction of disease-related patterns. The findings are validated with clinical data to support actionable insights, laying the groundwork for personalized disease models and precision therapeutic strategies. Expected outcomes include breakthroughs in disease understanding, advancements in precision medicine, and the establishment of a robust AI framework that will contribute to future biomedical applications. This multidisciplinary approach promises to transform diagnostics, prognosis, and treatment, facilitating a deeper understanding of disease mechanisms and their translation into clinical practice. The integration of multi-scale data enables a nuanced view of disease, capturing interactions that may be obscured in single-level analyses. By employing advanced visual computing techniques, the project highlights structural abnormalities within tissue samples, linking cellular changes to systemic disease pathways. This approach not only aids in identifying novel biomarkers but also supports the development of individualized treatment models that predict patient-specific responses. With validation from clinical datasets, the AI framework is poised to enhance clinical decision-making, offering insights tailored to the unique biological profiles of patients. The project’s multi-scale methodology serves as a foundation for future biomedical research, setting a new standard for AI-driven disease analysis.

(A)voiding misdiagnosis: prediction of detrusor underactivity vs. bladder outlet obstruction using pre-urodynamic nomogram in male patients with LUTS.

Our study aimed to develop a noninvasive model using a combination of the set of clinical data and uroflowmetry (UFL) to differentiate between detrusor underactivity (DU) and bladder outlet obstruction (BOO) in non-neurogenic male patients with lower urinary tract symptoms (LUTS). Data from 229 men with LUTS, diagnosed with DU or BOO on a pressure-flow study (PFS), were retrospectively analyzed, including medical history, Core Lower Urinary Tract Symptoms score (CLSS) questionnaire, UFL and PFS. Uni- and multivariate logistic regression were utilized for the prediction analyses. Of the cohort, 128 (55.9%) patients were diagnosed with DU. A multivariate logistic regression analysis identified less prevalent nocturia (OR 0.27, p < 0.002), more prevalent intermittency (OR 2.33, p = 0.03), less prevalent weak stream (OR 0.14, p = 0.0004), lower straining points in CLSS (OR 0.67, p = 0.02), higher slow stream points in CLSS (OR 1.81, p = 0.002), higher incomplete emptying points in CLSS (OR 1.31, p < 0.02), lower PVR ratio (OR 0.20, p = 0.03), and present features of fluctuating (OR 2.00, p = 0.05), fluctuating-intermittent (OR 3.09, p < 0.006), and intermittent (OR 8.11, p = 0.076) UFL curve shapes as independent predictors of DU. The above prediction model demonstrated satisfactory accuracy (c-index of 0.783). Our 10-factor model provides a noninvasive approach to differentiate DU from BOO in male patients with non-neurogenic LUTS, offering a valuable alternative to invasive PFS.

Over-expression of Transmembrane Protein 158 Predicts Aggressive Tumor Behavior and Poor Prognosis in Lung Cancer.

Transmembrane protein 158 (TMEM158) has emerged as a potential contributor to cancer progression. While TMEM158 has been studied in various cancer types, its role in lung cancer remains unclear. Kaplan-Meier curves were used to evaluate the association between TMEM158 expression and overall survival rates. Gene set enrichment analysis (GSEA) was performed to explore pathways related to TMEM158, and sequence analysis was conducted to identify putative hypoxia-responsive element (HRE) sites in the TMEM158 promoter region. Hypoxic conditions were induced, and TMEM158 expression levels were measured by qPCR. The effect of hypoxia-inducible factor-1α (HIF-1α) depletion on TMEM158 expression was also examined. Additionally, lung cancer cells with either over-expressed or reduced TMEM158 levels were analyzed for epithelial-mesenchymal transition (EMT) and cell migration. Analysis of clinical datasets revealed elevated TMEM158 expression in tumors, particularly in patients with advanced stages. High TMEM158 expression was correlated with lower overall survival. TMEM158 was associated with EMT, hypoxia, and other tumor-promoting pathways. Under hypoxic conditions, TMEM158 expression was at least partially induced in a HIF-1α-dependent manner. Functional studies showed that over-expression of TMEM158 promoted EMT and increased lung cancer cell migration, while depletion of TMEM158 reduced these effects, indicating its role in aggressive tumor behavior. TMEM158 is highly expressed in lung cancer, is associated with hypoxia, and promotes EMT and cell migration. These findings suggest TMEM158 as a potential target for lung cancer therapies.

The cascade integration model based on machine learning to predict gestational diabetes

Abstract Machine learning has significant superiority in disease prediction and auxiliary clinical decisions due to its powerful data analysis and exploration capabilities. Making accurate predictions about gestational diabetes mellitus (GDM) with clinical data is not always an easy task. In previous work, GDM was predicted using machine learning models and integrated learning models such as Stacking. This work improves on previous work by developing a new integrated learning model building approach for better exploiting the benefits of machine learning models. Initially, the clinical data set is normalized. Then, according to the principle of removing the redundant features of each machine learning model, the first nine high-importance features of the five single models are filtered respectively. Finally, the GDM Cascade integration prediction model is constructed and compared with the Blending model and Stacking model, it is obvious that the proposed model construction method has superior performance and the AUC value reaches 0.9536.&amp;#xD;

Clinical phenotype and functional influence of GRIN2A variants in epilepsy-aphasia syndrome.

N-methyl-D-aspartate receptors are glutamate-gated ion channels that play a crucial role in brain function. Numerous inherited or de novo variants in the GRIN2A gene, encoding the GluN2A subunit of the receptor, have been identified in patients with epilepsy. In addition, it is worth noting that GRIN2A variants exhibit a strong correlation with epilepsy-aphasia syndromes, a group of age-dependent epileptic, cognitive, and language disorders with a characteristic electroencephalographic pattern. Whole exome sequencing was conducted in enrolled patients with epilepsy-aphasia syndromes, and GRIN2A variants were screened. The conservation of substituted residues, conformational changes of mutant subunits, and in silico predictions of pathogenicity were thoroughly assessed in our study. Functional alterations of the variants were examined using whole-cell voltage-clamp current recordings while the relative surface expression levels of subunit proteins were assessed via immunofluorescence assays. A summary of previously published GRIN2A missense variants was conducted to investigate the genotypic-phenotypic-functional correlations. Two missense GRIN2A variants (c. 2482A >G/p. M828V, c.2627 T >C/p. I876T) were identified, which are located in the transmembrane helix M4 and C-terminus domain of the GluN2A subunit, respectively. Both variants exhibited reduced current density of NMDARs and surface/total expression levels of GluN2A subunits, while M828V showed a decreased extent of desensitization as well. A further summary of the previously reported GRIN2A variants demonstrated that more variable phenotypes were observed for variants situated in the C-terminus domain or those with loss-of-function effects. Our study expands the phenotypic and functional range of GRIN2A-related disorders. In order to optimally establish the domain-function-phenotype correlations in GRIN2A variants, it is imperative to gather a more extensive set of clinical and functional data. This study has identified two genetic variants of the GRIN2A gene in patients with epilepsy-aphasia syndrome. We assess the variants' harmfulness through a variety of functional experiments, including evaluating the expression level of the mutated protein and the resulting changes in electrophysiological activities. Also, we reviewed previously published papers about GRIN2A variants in epilepsy to learn more about the correlations between their locations, functional changes, and clinical manifestations.

Question Answering for Electronic Health Records: Scoping Review of Datasets and Models.

Question answering (QA) systems for patient-related data can assist both clinicians and patients. They can, for example, assist clinicians in decision-making and enable patients to have a better understanding of their medical history. Substantial amounts of patient data are stored in electronic health records (EHRs), making EHR QA an important research area. Because of the differences in data format and modality, this differs greatly from other medical QA tasks that use medical websites or scientific papers to retrieve answers, making it critical to research EHR QA. This study aims to provide a methodological review of existing works on QA for EHRs. The objectives of this study were to identify the existing EHR QA datasets and analyze them, study the state-of-the-art methodologies used in this task, compare the different evaluation metrics used by these state-of-the-art models, and finally elicit the various challenges and the ongoing issues in EHR QA. We searched for articles from January 1, 2005, to September 30, 2023, in 4 digital sources, including Google Scholar, ACL Anthology, ACM Digital Library, and PubMed, to collect relevant publications on EHR QA. Our systematic screening process followed PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. A total of 4111 papers were identified for our study, and after screening based on our inclusion criteria, we obtained 47 papers for further study. The selected studies were then classified into 2 non-mutually exclusive categories depending on their scope: "EHR QA datasets" and "EHR QA models." A systematic screening process obtained 47 papers on EHR QA for final review. Out of the 47 papers, 53% (n=25) were about EHR QA datasets, and 79% (n=37) papers were about EHR QA models. It was observed that QA on EHRs is relatively new and unexplored. Most of the works are fairly recent. In addition, it was observed that emrQA is by far the most popular EHR QA dataset, both in terms of citations and usage in other papers. We have classified the EHR QA datasets based on their modality, and we have inferred that Medical Information Mart for Intensive Care (MIMIC-III) and the National Natural Language Processing Clinical Challenges datasets (ie, n2c2 datasets) are the most popular EHR databases and corpuses used in EHR QA. Furthermore, we identified the different models used in EHR QA along with the evaluation metrics used for these models. EHR QA research faces multiple challenges, such as the limited availability of clinical annotations, concept normalization in EHR QA, and challenges faced in generating realistic EHR QA datasets. There are still many gaps in research that motivate further work. This study will assist future researchers in focusing on areas of EHR QA that have possible future research directions.

Seizure Onset Zone Detection Based on Convolutional Neural Networks and EEG Signals

Background: The localization of seizure onset zones (SOZs) is a critical step before the surgical treatment of epilepsy. Methods and Results: In this paper, we propose an SOZ detection method based on convolutional neural networks and EEG signals. This method aims to locate SOZs through the seizure status of each channel in multi-channel EEG signals. First, we preprocess the data with filtering, segmentation, resampling, and standardization to ensure their quality and consistency. Then, the single-channel UCI epilepsy seizure recognition dataset is used to train and test the convolutional neural network (CNN) model, achieving an accuracy of 98.70%, a sensitivity of 97.53%, and a specificity of 98.98%. Next, the multi-channel clinical EEG dataset collected by a hospital is divided into 21 single-channel site datasets and input into the model for detection, and then the seizure results of 21 sites per second are obtained. Finally, the seizure sites are visualized through the international 10–20 system electrode distribution map, diagrams of the change process of the seizure sites during seizures are drawn, and patients’ SOZs are located. Conclusions: Our proposed method well classifies seizure and non-seizure data and successfully locates SOZs by detecting the seizure results of 21 sites through a single-channel model. This study can effectively assist doctors in locating the SOZs of patients and provide help for the diagnosis and treatment of epilepsy.

Interaction between subclinical carotid atherosclerosis endotype and biological sex: a study from IMPROVE cohort

Abstract Background Carotid-intima media thickness (c-IMT) is a measure of subclinical carotid atherosclerosis and a predictor of future atherosclerotic cardiovascular events (ASCVD). We have recently generated, using an artificial intelligence approach, four endotypes (E) of subclinical atherosclerosis, i.e. subgroups of individuals predisposed to both c-IMT progression and ASCVD. Each E is defined by a set of demographic, clinical, and molecular data and characterize individuals with a low (E1), intermendiate (E2), high (E3) and very high (E4) cardiovascular risk profile. We observed an unbalanced distribution of males and females across the 4 endotypes, with a higher prevalence of female in the low risk E1. Purpose To estimate the interaction between biological sex and the 4 endotypes on measures of c-IMT progression and ASCVD risk. Methods We performed our study on 3121 individuals from the IMPROVE cohort, which includes participants with at least three ASCVD risk factors. We estimated the interaction between biological sex and the endotypes generated in the IMPROVE on (1) c-IMT and carotid plaque measurements after 30 months of follow-up (c-IMTfastest-progr, Area of plaquesbulb-progr) by linear regression (β, SE) and (2) with the 3-year ASCVD risk by Cox [hazard ratio (HR), 95% confidence interval (CI)] regression model. Crude estimates were adjusted by batch effect for biomarker measurement, geographical region and/or baseline ultrasonographic measures. Adjusted models further included ASCVD common risk factors. Results As shown in Figure 1, we did not observe any significant interaction between endotypes and biological sex on 30-month carotid ultrasound progression measures, and 3-year-ASCVD risk. In both the crude and adjusted models on c-IMTfastest-progr, Area of plaquesbulb-progr, and 3-year ASCVD, the p-value for interaction ranged from 0.18 to 0.98. While E4 was associated with the strongest cardiovascular risk profile in both males and females. Conclusions Our results indicate that the association between endotypes and progression of subclinical atherosclerosis and ASCVD risk mirrors a specific risk profile not entirely explained by differences in biological sex.

MetaboScope: A statistical toolbox for analyzing 1H nuclear magnetic resonance spectra from human clinical studies

Abstract Motivation Metabolic phenotyping, using high-resolution spectroscopic molecular fingerprints of biological samples, has demonstrated diagnostic, prognostic, and mechanistic value in clinical studies. However, clinical translation is hindered by the lack of viable workflows and challenges in converting spectral data into usable information. Results MetaboScope is an analytical and statistical workflow for learning, designing and analyzing clinically relevant 1H nuclear magnetic resonance data. It features modular pre-processing pipelines, multivariate modeling tools including principal components analysis (PCA), orthogonal-projection to latent structure discriminant analysis, and biomarker discovery tools (multiblock PCA and statistical spectroscopy). A simulation tool is also provided, allowing users to create synthetic spectra for hypothesis testing and power calculations. Availability and implementation MetaboScope is built as a pipeline where each module accepts the output generated by the previous one. This provides flexibility and simplicity of use, while being straightforward to maintain. The system and its libraries were developed in JavaScript and run as a web app; therefore, all the operations are performed on the local computer, circumventing the need to upload data. The MetaboScope tool is available at https://www.cheminfo.org/flavor/metabolomics/index.html. The code is open-source and can be deployed locally if necessary. Module notes, video tutorials, and clinical spectral datasets are provided for modeling. Supplementary information Supplementary data are available at Bioinformatics Advances online.

An ensemble machine learning approach for predicting heart failure risk in acute coronary syndrome patients: heartguardai

Abstract Background Heart failure (HF) constitutes a major post-event morbidity in patients experiencing Acute Coronary Syndrome (ACS), underscoring the imperative for precise risk stratification methodologies. Traditional prognostic models, although beneficial, are limited by their reliance on a narrow scope of variables, thereby overlooking the comprehensive analytical potential afforded by extensive clinical datasets. Aims To develop "HeartGuardAI," an ensemble machine learning model that predicts HF risk in ACS patients. By incorporating a wide array of variables including ejection fraction (EF), glomerular filtration rate (GFR), body mas index(BMI), male sex, this model seeks to offer a more practical tool for clinical decision-making. Methods We queried ACTION-ACS a large-scale multicenter database of 3335 patients at 10-year follow-up. We employed a feature selection process informed by Random Forest Classifier to identify the most predictive variables. An ensemble model combining the strengths of Random Forest and Gradient Boosting methodologies was then developed. The model's predictive performance was validated through Stratified K-Fold cross-validation, focusing on Area Under the Curve (AUC) and confidence intervals to assess its efficacy. The model's predictive performance was validated through Stratified K-Fold cross-validation, focusing on Area Under the Curve (AUC) and confidence intervals to assess its efficacy. Results Demonstrating a mean AUC of 0.76 ± 0.036, "HeartGuardAI" showed a high capability to predict HF risk in post-ACS patients and emphasizes the significance of incorporating multifactorial clinical assessments Conclusions HF risk can be predicted by a handy tool developed with advanced machine learning methods. Future studies are needed to prospectively validate this algorithm.Graphic 1Graphic 2

Temporomandibular joint CBCT image segmentation via multi-view ensemble learning network.

Accurate segmentation of the temporomandibular joint (TMJ) from cone beam CT (CBCT) images holds significant clinical value for diagnosing temporomandibular joint osteoarthrosis (TMJOA) and related conditions. Convolutional neural network-based medical image segmentation methods have achieved state-of-the-art performance in various segmentation tasks. However, 3D medical images segmentation requires substantial global context and rich spatial semantic information, demanding much more GPU memory and computational resources. To address these challenges in 3D medical image segmentation, we propose a novel network- the MVEL-Net (Multi-view Ensemble Learning Network) for TMJ CBCT image segmentation. By resampling images along three dimensions, we generate multiple weak learners with different spatial semantic information. A subsequent strong learning network effectively integrates the outputs from these weak learners to achieve more accurate segmentation results. We evaluated our network model using a clinical dataset comprising 88 subjects with TMJ CBCT images. The average Dice similarity coefficient (DSC) was 0.9817 ± 0.0049, the average surface distance was 0.0540 ± 0.0179mm, and the 95% Hausdorff distance was 0.1743 ± 0.0550mm. Our proposed MVEL-Net demonstrates excellent segmentation performance on TMJ from CBCT images, while using fewer GPU memory resources compared to other 3D networks. The effectiveness of this method in capturing spatial context could be leveraged for tasks like organ segmentation from volumetric scans. This may facilitate wider adoption of AI-based solutions for automated analysis of 3D medical images.

Significant of Blood Glucose Time-Series Forecasting with Temporal Fusion Transformer Model for Type 1 Diabetes

Blood glucose (BG) prediction has advanced to a state of the art thanks to deep learning models, which have been demonstrated to enhance type 1 diabetes (T1D) therapy. Nevertheless, most current models are limited to single-horizon prediction and have several practical drawbacks, including poor interpretability. In this study, we develop a novel approach to optimize the forecasting model in the training processes using the Optuna function, cross-validation, and the latest dataset. We offer a novel deep learning framework for multi-horizon BG prediction: the temporal fusion transformer (TFT). TFT employs a self-attention mechanism to extract long-term temporal dependencies and enables a model with an auto-tuning adjustment approach on hyperparameters and a cross-validation function on univariate and multivariate input models. On a clinical dataset with T1D subjects, namely ShanghaiT1DM (16 subjects), D1NAMO (9 subjects), and OhioT1DM (6 subjects) datasets, it achieved an average root mean square error (RMSE) of the three datasets used. The TFT model gave a lower error score than the baseline models of neural hierarchical interpolation for time series (N-HiTS)and long short-term memory (LSTM) with the values of 10.08+0.31 mg/dL and 12.34+0.62 mg/dL on the univariate input model and 9.18+1.21 mg/dL and 14.33+0.52 mg/dL on the multivariate input model for the prediction horizons of 30 and 60 minutes, respectively. This result explained that the TFT model was adequate for carrying out multi-horizon BG value-level forecasting and potentially deploying on-edge devices to improve clinical efforts to manage BG levels for T1D patients in real-time application.

Virtual resection evaluation based on sEEG propagation network for drug-resistant epilepsy

Drug-resistant epilepsy with frequent seizures are considered to undergo surgery to become seizure-free, but seizure-free rates have not dramatically improved, partly due to imprecise intervention locations. To address this clinical need, we construct effective connectivity to reveal epilepsy brain dynamics. Based on the propagation path captured by the high order effective connectivity, calculate the control centrality evaluation scheme of the excised area. We used three datasets: simulation dataset, clinical dataset, and public dataset. The epileptogenic propagation network was quantified by calculating high-order effective connection to obtain accurate propagation path, based on this, combined with the outdegree index for virtual resection. By removing electrodes and recalculating control centrality, we quantify each electrode or region’s control centrality to evaluate the virtual resection scheme. Three datasets obtained consistent results. We track the accurate propagation path and find the obvious inflection points occurring during the excision process. The minimum intervention targets were obtained by comparing different schemes without recurrence. The clinical data with multiple seizures found that after resection, the brain reaches a stable state and is less likely to continue spreading. By quantitative analysis of control centrality to evaluate the possible excision scheme, finally we obtain the best intervention area for epilepsy, which assist in developing surgical plans.

Colon polyp detection based on multi-scale and multi-level feature fusion and lightweight convolutional neural network

Early diagnosis and treatment of colorectal polyps are crucial for preventing colorectal cancer. This paper proposes a lightweight convolutional neural network for the automatic detection and auxiliary diagnosis of colorectal polyps. Initially, a 53-layer convolutional backbone network is used, incorporating a spatial pyramid pooling module to achieve feature extraction with different receptive field sizes. Subsequently, a feature pyramid network is employed to perform cross-scale fusion of feature maps from the backbone network. A spatial attention module is utilized to enhance the perception of polyp image boundaries and details. Further, a positional pattern attention module is used to automatically mine and integrate key features across different levels of feature maps, achieving rapid, efficient, and accurate automatic detection of colorectal polyps. The proposed model is evaluated on a clinical dataset, achieving an accuracy of 0.9982, recall of 0.9988, F1 score of 0.9984, and mean average precision (mAP) of 0.9953 at an intersection over union (IOU) threshold of 0.5, with a frame rate of 74 frames per second and a parameter count of 9.08 M. Compared to existing mainstream methods, the proposed method is lightweight, has low operating configuration requirements, high detection speed, and high accuracy, making it a feasible technical method and important tool for the early detection and diagnosis of colorectal cancer.

Key genes and pathways in the molecular landscape of pancreatic ductal adenocarcinoma: A bioinformatics and machine learning study

Pancreatic ductal adenocarcinoma (PDAC) is recognized for its aggressive nature, dismal prognosis, and a notably low five-year survival rate, underscoring the critical need for early detection methods and more effective therapeutic approaches. This research rigorously investigates the molecular mechanisms underlying PDAC, with a focus on the identification of pivotal genes and pathways that may hold therapeutic relevance and prognostic value. Through the construction of a protein-protein interaction (PPI) network and the examination of differentially expressed genes (DEGs), the study uncovers key hub genes such as CDK1, KIF11, and BUB1, demonstrating their substantial role in the pathogenesis of PDAC. Notably, the dysregulation of these genes is consistent across a spectrum of cancers, positing them as potential targets for wide-ranging cancer therapeutics. This study also brings to the fore significant genes encoding intrinsically disordered proteins, in particular GPRC5A and KRT7, unveiling promising new pathways for therapeutic intervention. Advanced machine learning techniques were harnessed to classify PDAC patients with high accuracy, utilizing the key genetic markers as a dataset. The Support Vector Machine (SVM) model leveraged the hub genes to achieve a sensitivity of 91 % and a specificity of 85 %, while the RandomForest model notched a sensitivity of 91 % and specificity of 92.5 %. Crucially, when the identified genes were cross-referenced with TCGA-PAAD clinical datasets, a tangible correlation with patient survival rates was discovered, reinforcing the potential of these genes as prognostic biomarkers and their viability as targets for therapeutic intervention. This study's findings serve as a potent testament to the value of molecular analysis in enhancing the understanding of PDAC and in advancing the pursuit for more effective diagnostic and treatment strategies.

Investigation of scatter energy window width and count levels for deep learning-based attenuation map estimation in cardiac SPECT/CT imaging.

Deep learning (DL) is becoming increasingly important in generating attenuation maps for accurate attenuation correction in cardiac perfusion SPECT imaging. Typically, DL models take inputs from initial reconstructed SPECT images, which are performed on the photopeak window and often also on scatter windows. While prior studies have demonstrated improvements in DL performance when scatter window images are incorporated into the DL input, the comprehensive analysis of the impact of employing different scatter windows remains unassessed. Additionally, existing research mainly focuses on applying DL to SPECT scans obtained at clinical standard count levels. This study aimed to assess utilities of DL from two aspects: 1) investigating the impact when different scatter windows were used as input to DL, and 2) evaluating the performance of DL when applied on SPECT scans acquired at a reduced count level. We utilized 1517 subjects, with 386 subjects for testing and the remaining 1131 for training and validation. The results showed that as scatter window width increased from 4% to 30%, a slight improvement was observed in DL estimated attenuation maps. The application of DL models to quarter-count (¼-count) SPECT scans, compared to full-count scans, showed a slight reduction in performance. Nonetheless, discrepancies across different scatter window configurations and between count levels were minimal, with all normalized mean square error (NMSE) values remaining within 2.1% when comparing the different DL attenuation maps to the reference CT maps. For attenuation corrected SPECT slices using DL estimated maps, NMSE values were within 0.5% when compared to CT correction. This study, leveraging an extensive clinical dataset, showed that the performance of DL seemed to be consistent across the use of varied scatter window settings. Moreover, our investigation into reduced count studies indicated that DL could provide accurate attenuation correction even at a ¼-count level.

PET image reconstruction using weighted nuclear norm maximization and deep learning prior

The ill-posed Positron emission tomography (PET) reconstruction problem usually results in limited resolution and significant noise. Recently, deep neural networks have been incorporated into PET iterative reconstruction framework to improve the image quality. In this paper, we propose a new neural network-based iterative reconstruction method by using weighted nuclear norm (WNN) maximization, which aims to recover the image details in the reconstruction process. The novelty of our method is the application of WNN maximization rather than WNN minimization in PET image reconstruction. Meanwhile, a neural network is used to control the noise originated from WNN maximization. Our method is evaluated on simulated and clinical datasets. The simulation results show that the proposed approach outperforms state-of-the-art neural network-based iterative methods by achieving the best contrast/noise tradeoff with a remarkable contrast improvement on the lesion contrast recovery. The study on clinical datasets also demonstrates that our method can recover lesions of different sizes while suppressing noise in various low-dose PET image reconstruction tasks. Our code is available athttps://github.com/Kuangxd/PETReconstruction.

X-ray coronary angiography background subtraction by adaptive weighted total variation regularized online RPCA

Objective.X-ray coronary angiograms (XCA) are widely used in diagnosing and treating cardiovascular diseases. Various structures with independent motion patterns in the background of XCA images and limitations in the dose of injected contrast agent have resulted in low-contrast XCA images. Background subtraction methods have been developed to enhance the visibility and contrast of coronary vessels in XCA sequences, consequently reducing the requirement for excessive contrast agent injections.Approach.The current study proposes an adaptive weighted total variation regularized online RPCA (WTV-ORPCA) method, which is a low-rank and sparse subspaces decomposition approach to subtract the background of XCA sequences. In the proposed method, the images undergo initial preprocessing using morphological operators to eliminate large-scale background structures and achieve image homogenization. Subsequently, the decomposition algorithm decomposes the preprocessed images into background and foreground subspaces. This step applies an adaptive weighted TV constraint to the foreground subspace to ensure the spatial coherency of the finally extracted coronary vessel images.Main results.To evaluate the effectiveness of the proposed background subtraction method, some qualitative and quantitative experiments are conducted on two clinical and synthetic low-contrast XCA datasets containing videos from 21 patients. The obtained results are compared with six state-of-the-art methods employing three different assessment criteria. By applying the proposed method to the clinical dataset, the mean values of the global contrast-to-noise ratio, local contrast-to-noise ratio, structural similarity index, and reconstruction error (RE) are obtained as5.976,3.173,0.987, and0.026, respectively. These criteria over the low-contrast synthetic dataset were4.851,2.942,0.958, and0.034, respectively.Significance.The findings demonstrate the superiority of the proposed method in improving the contrast and visibility of coronary vessels, preserving the integrity of the vessel structure, and minimizing REs without imposing excessive computational complexity.

Automated quantification of cerebral microbleeds in susceptibility-weighted MRI: association with vascular risk factors, white matter hyperintensity burden, and cognitive function.

To train and validate a deep learning (DL)-based segmentation model for cerebral microbleeds (CMB) on susceptibility-weighted MRI; and to find associations between CMB, cognitive impairment, and vascular risk factors. Participants in this single-institution retrospective study underwent brain MRI to evaluate cognitive impairment between January-September 2023. For training the DL model, the nnU-Net framework was used without modifications. The DL model's performance was evaluated on independent internal and external validation datasets. Linear regression analysis was used to find associations between log-transformed CMB numbers, cognitive function (mini-mental status examination [MMSE]), white matter hyperintensity (WMH) burden, and clinical vascular risk factors (age, sex, hypertension, diabetes, lipid profiles, and body mass index). Training of the DL model (n = 287) resulted in a robust segmentation performance with an average dice score of 0.73 (95% CI, 0.67-0.79) in an internal validation set, (n = 67) and modest performance in an external validation set (dice score = 0.46, 95% CI, 0.33-0.59, n = 68). In a temporally independent clinical dataset (n = 448), older age, hypertension, and WMH burden were significantly associated with CMB numbers in all distributions (total, lobar, deep, and cerebellar; all P <.01). MMSE was significantly associated with hyperlipidemia (β = 1.88, 95% CI, 0.96-2.81, P <.001), WMH burden (β = -0.17 per 1% WMH burden, 95% CI, -0.27-0.08, P <.001), and total CMB number (β = -0.01 per 1 CMB, 95% CI, -0.02-0.001, P = .04) after adjusting for age and sex. The DL model showed a robust segmentation performance for CMB. In all distributions, CMB had significant positive associations with WMH burden. Increased WMH burden and CMB numbers were associated with decreased cognitive function. CMB = cerebral microbleed; DL = deep learning, DSC = dice similarity coefficient; MMSE = mini-mental status examination; SVD = small vessel disease; SWI = susceptibility-weighted image; WMH = white matter hyperintensity.

Clinical benchmark dataset for AI accuracy analysis: quantifying radiographic annotation of pelvic tilt

Radiographic landmark annotation determines patients’ anatomical parameters and influences diagnoses. However, challenges arise from ambiguous region-based definitions, human error, and image quality variations, potentially compromising patient care. Additionally, AI landmark localization often presents its predictions in a probability-based heatmap format, which lacks a corresponding clinical standard for accuracy validation. This Data Descriptor presents a clinical benchmark dataset for pelvic tilt landmarks, gathered through a probabilistic approach to measure annotation accuracy within clinical environments. A retrospective analysis of 115 pelvic sagittal radiographs was conducted for annotating pelvic tilt parameters by five annotators, revealing landmark cloud sizes of 6.04 mm-17.90 mm at a 95% dataset threshold, corresponding to 9.51°–16.55° maximum angular disagreement in clinical settings. The outcome provides a quantified point cloud dataset for each landmark corresponding to different probabilities, which enables assessment of directional annotation distribution and parameter-wise impact, providing clinical benchmarks. The data is readily reusable for AI studies analyzing the same landmarks, and the method can be easily replicated for establishing clinical accuracy benchmarks of other landmarks.