Brief Bioinform. 2024 Sep 23;25(6):bbae591. doi: 10.1093/bib/bbae591.

ABSTRACT

Cell tracking is an essential function needed in automated cellular activity monitoring. In practice, processing methods striking a balance between computational efficiency and accuracy as well as demonstrating robust generalizability across diverse cell datasets are highly desired. This paper develops a central-metric fully test-time adaptive framework for cell tracking (CMTT-JTracker). Firstly, a CMTT mechanism is designed for the pre-segmentation of cell images, which enables extracting target information at different resolutions without additional training. Next, a multi-task learning network with the spatial attention scheme is developed to simultaneously realize detection and re-identification tasks based on features extracted by CMTT. Experimental results demonstrate that the CMTT-JTracker exhibits remarkable biological and tracking performance compared with benchmarking tracking methods. It achieves a multiple object tracking accuracy (MOTA) of $0.894$ on Fluo-N2DH-SIM+ and a MOTA of $0.850$ on PhC-C2DL-PSC. Experimental results further confirm that the CMTT applied solely as a segmentation unit outperforms the SOTA segmentation benchmarks on various datasets, particularly excelling in scenarios with dense cells. The Dice coefficients of the CMTT range from a high of $0.928$ to a low of $0.758$ across different datasets.

PMID:39552066 | DOI:10.1093/bib/bbae591

Sci Rep. 2024 Nov 18;14(1):28388. doi: 10.1038/s41598-024-79610-w.

ABSTRACT

Magnetic resonance imaging remains the gold standard for diagnosing osteoporotic vertebral compression fractures (OVCF), but the use of X-ray imaging, particularly anteroposterior (AP) and lateral views, is prevalent due to its accessibility and cost-effectiveness. We aim to assess whether the performance of AP images-based deep learning is comparable compared to those using lateral images. This retrospective study analyzed X-ray images from two tertiary teaching hospitals, involving 1,507 patients for the training and internal test, and 104 patients for the external test. The EfficientNet-B5-based algorithms were employed to classify OVCF and non-OVCF group. The model was trained with a 1:1 balanced dataset and validated through 5-fold cross validation. Performance outcomes were compared with the area under receiver operating characteristic (AUROC) curve. Out of a total of 1,507 patients, 799 were included in the training dataset and 708 were included in the internal test dataset. The training and internal test datasets were matched 1:1 as OVCF and non-OVCF patients. The DL model showed comparable classifying performance with internal test data (N = 708, AUROC for AP, 0.915; AUROC for lateral, 0.953) and external test data (N = 104, AUROC for AP, 0.982; AUROC for lateral, 0979), respectively. The other performances including F1 score and accuracy were also comparable. Especially, The AUROC of AP and lateral x-ray image-based DL was not significantly different (p for DeLong test = 0.604). The EfficientNet-B5 algorithms using AP X-ray images shows comparable efficacy for classifying OVCF and non-OVCF compared to lateral images.

PMID:39551876 | DOI:10.1038/s41598-024-79610-w

Sci Rep. 2024 Nov 17;14(1):28377. doi: 10.1038/s41598-024-79581-y.

ABSTRACT

Detecting glucose levels is crucial for diabetes patients as it enables timely and effective management, preventing complications and promoting overall health. In this endeavor, we have designed a novel, affordable point-of-care diagnostic device utilizing microfluidic principles, a smartphone camera, and established laboratory colorimetric methods for accurate glucose estimation. Our proposed microfluidic device comprises layers of adhesive poly-vinyl films stacked on a poly methyl methacrylate (PMMA) base sheet, with micro-channel contours precision-cut using a cutting printer. Employing the gold standard glucose-oxidase/peroxidase reaction on this microfluidic platform, we achieve enzymatic glucose determination. The resulting colored complex, formed by phenol and 4-aminoantipyrine in the presence of hydrogen peroxide generated during glucose oxidation, is captured at various glucose concentrations using a smartphone camera. Raw images are processed and utilized as input data for a 2-D convolutional neural network (CNN) deep learning classifier, demonstrating an impressive 95% overall accuracy against new images. The glucose predictions done by CNN are compared with ISO 15197:2013/2015 gold standard norms. Furthermore, the classifier exhibits outstanding precision, recall, and F1 score of 94%, 93%, and 93%, respectively, as validated through our study, showcasing its exceptional predictive capability. Next, a user-friendly smartphone application named "GLUCOLENS AI" was developed to capture images, perform image processing, and communicate with cloud server containing the CNN classifier. The developed CNN model can be successfully used as a pre-trained model for future glucose concentration predictions.

PMID:39551869 | DOI:10.1038/s41598-024-79581-y

Curr Med Sci. 2024 Nov 18. doi: 10.1007/s11596-024-2928-5. Online ahead of print.

ABSTRACT

Artificial intelligence (AI) is an interdisciplinary field that combines computer technology, mathematics, and several other fields. Recently, with the rapid development of machine learning (ML) and deep learning (DL), significant progress has been made in the field of AI. As one of the fastest-growing branches, DL can effectively extract features from big data and optimize the performance of various tasks. Moreover, with advancements in digital imaging technology, DL has become a key tool for processing high-dimensional medical image data and conducting medical image analysis in clinical applications. With the development of this technology, the diagnosis of orthopedic diseases has undergone significant changes. In this review, we describe recent research progress on DL in fracture diagnosis and discuss the value of DL in this field, providing a reference for better integration and development of DL technology in orthopedics.

PMID:39551854 | DOI:10.1007/s11596-024-2928-5

NPJ Precis Oncol. 2024 Nov 17;8(1):263. doi: 10.1038/s41698-024-00754-z.

ABSTRACT

Accurate treatment response assessment using serial CT scans is essential in oncological clinical trials. However, oncologists' assessment following the Response Evaluation Criteria in Solid Tumors (RECIST) guideline is subjective, time-consuming, and sometimes fallible. Advanced liver cancer often presents multifocal hepatic lesions on CT imaging, making accurate characterization more challenging than with other malignancies. In this work, we developed a tumor volume guided comprehensive objective response evaluation based on deep learning (RECORD) for liver cancer. RECORD performs liver tumor segmentation, followed by sum of the volume (SOV)-based treatment response classification and new lesion assessment. Then, it can provide treatment evaluations of response, stability, and progression, and calculates progression-free survival (PFS) and response time. The RECORD pipeline was developed with both CNN and ViT backbones. Its performance was evaluated in three longitudinal cohorts involving 60 multi-national centers, 206 patients, 891 CT scans, using internal five-fold cross-validation and external validations. RECORD with the most effective backbone achieved an average AUC-response of 0.981, AUC-stable of 0.929, and AUC-progression of 0.969 for SOV-based disease status classification, F1-score of 0.887 for new lesion identification, and accuracy of 0.889 for final treatment outcome assessments across all cohorts. RECORD's PFS and response time predictions strongly correlated with clinician's assessments (P < 0.001). Moreover, RECORD can better stratify high-risk versus low-risk patients for overall survival compared to the human-assessed RECIST results. In conclusion, RECORD demonstrates efficiency and objectivity in analyzing liver lesions for treatment response evaluation. Further research should extend the pipeline to other metastatic organ sites.

PMID:39551847 | DOI:10.1038/s41698-024-00754-z

Sci Rep. 2024 Nov 18;14(1):28402. doi: 10.1038/s41598-024-79028-4.

ABSTRACT

Neural networks are reported to be vulnerable under minor and imperceptible attacks. The underlying mechanism and quantitative measure of the vulnerability still remains to be revealed. In this study, we explore the intrinsic trade-off between accuracy and robustness in neural networks, framed through the lens of the "uncertainty principle". By examining the fundamental limitations imposed by this principle, we reveal how neural networks inherently balance precision in feature extraction with susceptibility to adversarial perturbations. Our analysis highlights that as neural networks achieve higher accuracy, their vulnerability to adversarial attacks increases, a phenomenon rooted in the uncertainty relation. By using the mathematics from quantum mechanics, we offer a theoretical foundation and analytical method for understanding the vulnerabilities of deep learning models.

PMID:39551816 | DOI:10.1038/s41598-024-79028-4

Sci Rep. 2024 Nov 17;14(1):28379. doi: 10.1038/s41598-024-78067-1.

ABSTRACT

Glioma refers to a highly prevalent type of brain tumor that is strongly associated with a high mortality rate. During the treatment process of the disease, it is particularly important to accurately perform segmentation of the glioma from Magnetic Resonance Imaging (MRI). However, existing methods used for glioma segmentation usually rely solely on either local or global features and perform poorly in terms of capturing and exploiting critical information from tumor volume features. Herein, we propose a local and global dual transformer with an attentional supervision U-shape network called DTASUnet, which is purposed for glioma segmentation. First, we built a pyramid hierarchical encoder based on 3D shift local and global transformers to effectively extract the features and relationships of different tumor regions. We also designed a 3D channel and spatial attention supervision module to guide the network, allowing it to capture key information in volumetric features more accurately during the training process. In the BraTS 2018 validation set, the average Dice scores of DTASUnet for the tumor core (TC), whole tumor (WT), and enhancing tumor (ET) regions were 0.845, 0.905, and 0.808, respectively. These results demonstrate that DTASUnet has utility in assisting clinicians with determining the location of gliomas to facilitate more efficient and accurate brain surgery and diagnosis.

PMID:39551805 | DOI:10.1038/s41598-024-78067-1

Sci Rep. 2024 Nov 18;14(1):28399. doi: 10.1038/s41598-024-80090-1.

ABSTRACT

This study aims to quantify fundus tessellated (FT) density and optic disc (OD) morphology using deep learning (DL) techniques and to investigate the correlations between these fundus characteristics and refractive function in young patients with myopia. We constructed two DL-based segmentation models to delineate the FT, OD, peripapillary atrophy (PPA), and macula at a pixel-level resolution. The study sought to identify differences in fundus characteristics between eyes categorized as having high myopia versus mild or moderate myopia. Furthermore, the correlation between fundus measurements and various ocular parameters was statistically analyzed. Correlation analysis indicated that the spherical equivalent and axial length were significantly associated with all fundus measurements (p < 0.001). Additionally, corneal curvature (K1, K2), lens thickness, and foveal thickness exhibited significant correlations with some of the fundus measurements at a 0.01 significance level. Using DL algorithms, it is feasible to automatically quantify FT and OD characteristics in young myopic patients. The study findings suggest that both FT and OD characteristics are highly correlated with the severity of myopia, particularly as it progresses from mild or moderate to high levels. Moreover, a significant relationship exists between most of these fundus characteristics and a spectrum of refractive function parameters.

PMID:39551799 | DOI:10.1038/s41598-024-80090-1

Nat Commun. 2024 Nov 17;15(1):9955. doi: 10.1038/s41467-024-54375-y.

ABSTRACT

Behavioral larval zebrafish screens leverage a high-throughput small molecule discovery format to find neuroactive molecules relevant to mammalian physiology. We screen a library of 650 central nervous system active compounds in high replicate to train deep metric learning models on zebrafish behavioral profiles. The machine learning initially exploited subtle artifacts in the phenotypic screen, necessitating a complete experimental re-run with rigorous physical well-wise randomization. These large matched phenotypic screening datasets (initial and well-randomized) provide a unique opportunity to quantify and understand shortcut learning in a full-scale, real-world drug discovery dataset. The final deep metric learning model substantially outperforms correlation distance-the canonical way of computing distances between profiles-and generalizes to an orthogonal dataset of diverse drug-like compounds. We validate predictions by prospective in vitro radio-ligand binding assays against human protein targets, achieving a hit rate of 58% despite crossing species and chemical scaffold boundaries. These neuroactive compounds exhibit diverse chemical scaffolds, demonstrating that zebrafish phenotypic screens combined with metric learning achieve robust scaffold hopping capabilities.

PMID:39551797 | DOI:10.1038/s41467-024-54375-y

Mol Cancer. 2024 Nov 18;23(1):258. doi: 10.1186/s12943-024-02169-7.

ABSTRACT

Millions of new cases of cancer are diagnosed worldwide each year, making it a serious public health concern. Developments in customized therapy and early detection have significantly enhanced treatment for and results from cancer. Therefore, it is important to investigate new molecular biomarkers. In this study, we created an efficient transient receptor potential melastatin (TRPM) family members-related TRPM-Score for 17 solid tumors. CCNE1, produced from TRPM-Score, was found to be an exceptional biomarker through several sophisticated machine learning and deep learning computational techniques. TRPM-Score and CCNE1 immunotherapeutic prediction, immunological characteristics, and predictive value were thoroughly assessed. In most cancer types, CCNE1 was a substantially dangerous marker. Additional in vitro tests validated CCNE1's immunomodulatory properties, demonstrating that silencing impeded macrophage movement and decreased PD-L1 expression. Additionally, CCNE1 may accurately predict responses to cancer immunotherapy. These findings indicate that the TRPM family-particularly CCNE1, which is associated with TRPM-is a significant player in the pan-cancer domain and can be utilized as a therapeutic target and prognostic biomarkers, especially in immuno-oncology. The thorough characterization of the TRPM family and the discovery of CCNE1 as a crucial downstream effector mark important developments in our comprehension of pan-cancer biology.

PMID:39551726 | DOI:10.1186/s12943-024-02169-7

Ultrasound Med Biol. 2024 Nov 16:S0301-5629(24)00384-3. doi: 10.1016/j.ultrasmedbio.2024.10.005. Online ahead of print.

ABSTRACT

Prostate cancer (PCa) poses a significant threat to men's health, with early diagnosis being crucial for improving prognosis and reducing mortality rates. Transrectal ultrasound (TRUS) plays a vital role in the diagnosis and image-guided intervention of PCa. To facilitate physicians with more accurate and efficient computer-assisted diagnosis and interventions, many image processing algorithms in TRUS have been proposed and achieved state-of-the-art performance in several tasks, including prostate gland segmentation, prostate image registration, PCa classification and detection and interventional needle detection. The rapid development of these algorithms over the past 2 decades necessitates a comprehensive summary. As a consequence, this survey provides a narrative review of this field, outlining the evolution of image processing methods in the context of TRUS image analysis and meanwhile highlighting their relevant contributions. Furthermore, this survey discusses current challenges and suggests future research directions to possibly advance this field further.

PMID:39551652 | DOI:10.1016/j.ultrasmedbio.2024.10.005

Network. 2024 Nov 16:1-27. doi: 10.1080/0954898X.2024.2426580. Online ahead of print.

ABSTRACT

Cancer involves abnormal cell growth, with types like intestinal and oesophageal cancer often diagnosed in advanced stages, making them hard to cure. Symptoms are like burning sensations in the stomach and swallowing difficulties are specified as colorectal cancer. Deep learning significantly impacts the medical image processing and diagnosis, offering potential improvements in accuracy and efficiency. The Kruskal Szekeres Generative Adversarial Network Augmented Deep Autoencoder (KSGANA-DA) is introduced for early colorectal cancer detection and it comprises two stages; Initial stage, data augmentation uses Affine Transform via Random Horizontal Rotation and Geometric Transform via Kruskal-Szekeres that coordinates to improve the training dataset diversity, boosting detection performance. The second stage, a Deep Autoencoder Anatomical Landmark-based Image Segmentation preserves edge pixel spatial locations, improving precision and recall for early boundary detection. Experiments validate KSGANA-DA performance and different existing methods are implemented into Python. The results of KSGANA-DA are to provide higher precision by 41%, recall by 7%, and lesser training time by 46% than compared to conventional methods.

PMID:39550608 | DOI:10.1080/0954898X.2024.2426580

Sci Rep. 2024 Nov 16;14(1):28342. doi: 10.1038/s41598-024-80033-w.

ABSTRACT

The article proposes a novel approach to assess rotor angle stability in microgrids by enhancing the Modified Galerkin Method (MGM), which is based on the Polynomial Approximation, using real-time RFID data acquisition. Due to their reliance on assumptions, traditional rotor angle stability methodologies frequently fail in online transient stability testing. MGM successfully captures the dynamic behavior of microgrids by approximating state variables using a sequence of polynomials and coefficients. Redundant data, like as vibrations or noise signals, can cause delays in defect diagnosis and decrease diagnostic accuracy. This problem is addressed by integrating RFID technology. RFID technology could potentially be used with a hybrid CNN-LSTM model to develop a sophisticated fault diagnostic system. This entails identifying fault characteristics through the use of signal processing techniques and feature extraction methods, such as the Fourier transform and time-domain statistical features. In addition, we use Total Harmonic Distortion (THD) to reduce superfluous data. The suggested techniques significantly increase fault detection efficiency and precision, outperforming existing techniques with a 0.94 classification accuracy. An extensive case study on an IEEE 3-machine 9-bus system is used to illustrate its efficacy, showing observable improvements in fault detection speed and accuracy that make microgrid operations safer and more dependable.

PMID:39550455 | DOI:10.1038/s41598-024-80033-w

Sci Rep. 2024 Nov 16;14(1):28321. doi: 10.1038/s41598-024-79799-w.

ABSTRACT

In the race to combat ever-evolving diseases, the drug discovery process often faces the hurdles of high-cost and time-consuming procedures. To tackle these challenges and enhance the efficiency of identifying new therapeutic agents, we introduce VirtuDockDL, which is a streamlined Python-based web platform utilizing deep learning for drug discovery. This pipeline employs a Graph Neural Network to analyze and predict the effectiveness of various compounds as potential drug candidates. During the validation phase, VirtuDockDL was instrumental in identifying non-covalent inhibitors against the VP35 protein of the Marburg virus, a critical target given the virus's high fatality rate and limited treatment options. Further, in benchmarking, VirtuDockDL achieved 99% accuracy, an F1 score of 0.992, and an AUC of 0.99 on the HER2 dataset, surpassing DeepChem (89% accuracy) and AutoDock Vina (82% accuracy). Compared to RosettaVS, MzDOCK, and PyRMD, VirtuDockDL outperformed them by combining both ligand- and structure-based screening with deep learning. While RosettaVS excels in accurate docking but lacks high-throughput screening, and PyRMD focuses on ligand-based methods without AI integration, VirtuDockDL offers superior predictive accuracy and full automation for large-scale datasets, making it ideal for comprehensive drug discovery workflows. These results underscore the tool's capability to identify high-affinity inhibitors accurately across various targets, including the HER2 protein for cancer therapy, TEM-1 beta-lactamase for bacterial infections, and the CYP51 enzyme for fungal infections like Candidiasis. To sum up, VirtuDockDL combines user-friendly interface design with powerful computational capabilities to facilitate rapid, cost-effective drug discovery and development. The integration of AI in drug discovery could potentially transform the landscape of pharmaceutical research, providing faster responses to global health challenges. The VirtuDockDL is available at https://github.com/FatimaNoor74/VirtuDockDL .

PMID:39550439 | DOI:10.1038/s41598-024-79799-w

Nat Commun. 2024 Nov 16;15(1):9951. doi: 10.1038/s41467-024-54441-5.

ABSTRACT

Understanding how genetic variants affect the epigenome is key to interpreting GWAS, yet profiling these effects across the non-coding genome remains challenging due to experimental scalability. This necessitates accurate computational models. Existing machine learning approaches, while progressively improving, are confined to the cell types they were trained on, limiting their applicability. Here, we introduce Enformer Celltyping, a deep learning model which incorporates distal effects of DNA interactions, up to 100,000 base-pairs away, to predict epigenetic signals in previously unseen cell types. Using DNA and chromatin accessibility data for epigenetic imputation, Enformer Celltyping outperforms current best-in-class approaches and generalises across cell types and biological regions. Moreover, we propose a framework for evaluating models on genetic variant effect prediction using regulatory quantitative trait loci mapping studies, highlighting current limitations in genomic deep learning models. Despite this, Enformer Celltyping can also be used to study cell type-specific genetic enrichment of complex traits.

PMID:39550354 | DOI:10.1038/s41467-024-54441-5

J Voice. 2024 Nov 15:S0892-1997(24)00353-9. doi: 10.1016/j.jvoice.2024.10.010. Online ahead of print.

ABSTRACT

Integrating Convolutional Neural Networks and Vision Transformers in voice analysis has unveiled a new horizon in mental health identification. Human voice, a powerful indicator of mental health, was the focus of this study. Human voice data representing stable and unstable conditions were gathered from various mental health institutions in Bangladesh. The results of the experiment suggest that the proposed model achieved 91% accuracy, precision of 92% for the "Unstable" category and 90% for the "Stable" category, and recall of 91% for the "Stable" category and 92% for the "Unstable" category. In addition, a high F1 score of 91% was achieved. This study significantly contributes to computer-aided diagnosis in mental health by using deep learning (DL) to diagnose mental well-being. Our research underscores the substantial impact of DL on the advancement of mental health care, instilling hope for a brighter future in mental health care.

PMID:39550322 | DOI:10.1016/j.jvoice.2024.10.010

Brief Bioinform. 2024 Sep 23;25(6):bbae601. doi: 10.1093/bib/bbae601.

ABSTRACT

RNA 2'-O-methylation (Nm) is a crucial post-transcriptional modification with significant biological implications. However, experimental identification of Nm sites is challenging and resource-intensive. While multiple computational tools have been developed to identify Nm sites, their predictive performance, particularly in terms of precision and generalization capability, remains deficient. We introduced Nmix, an advanced computational tool for precise prediction of Nm sites in human RNA. We constructed the largest, low-redundancy dataset of experimentally verified Nm sites and employed an innovative multi-feature fusion approach, combining one-hot, Z-curve and RNA secondary structure encoding. Nmix utilizes a meticulously designed hybrid deep learning architecture, integrating 1D/2D convolutional neural networks, self-attention mechanism and residual connection. We implemented asymmetric loss function and Bayesian optimization-based ensemble learning, substantially improving predictive performance on imbalanced datasets. Rigorous testing on two benchmark datasets revealed that Nmix significantly outperforms existing state-of-the-art methods across various metrics, particularly in precision, with average improvements of 33.1% and 60.0%, and Matthews correlation coefficient, with average improvements of 24.7% and 51.1%. Notably, Nmix demonstrated exceptional cross-species generalization capability, accurately predicting 93.8% of experimentally verified Nm sites in rat RNA. We also developed a user-friendly web server (https://tubic.org/Nm) and provided standalone prediction scripts to facilitate widespread adoption. We hope that by providing a more accurate and robust tool for Nm site prediction, we can contribute to advancing our understanding of Nm mechanisms and potentially benefit the prediction of other RNA modification sites.

PMID:39550226 | DOI:10.1093/bib/bbae601

J Stomatol Oral Maxillofac Surg. 2024 Nov 14:102151. doi: 10.1016/j.jormas.2024.102151. Online ahead of print.

ABSTRACT

PURPOSES: One notable anomaly, presence of distomolars, arises beyond the typical sequence of the human dental system. In this study, convolutional neural networks (CNNs) based machine learning methods were employed to classify distomolar tooth existence using panoramic radiography (PR).

METHODS: PRs dataset, composed of 117 subjects with distomolar teeth and 146 subjects without distomolar teeth, was constructed. These images were assessed using AlexNet, DarkNet, DenseNet, EfficientNet, GoogLeNet, ResNet, MobileNet, NasNet-Mobile, VGG, and XceptionNet frameworks for distomolar teeth existence. Considering the moderate number dataset samples, transfer learning was also utilized to improve the performance of these CNN based networks along with 5-fold cross-validation. The final classification was obtained through the fusion of the classifiers results.

RESULTS: Performance of the experimental studies was assessed utilizing accuracy (Acc), sensitivity (sen), specificity (spe) and precision (pre) metrics. Best accuracy value of 96.2% was obtained for the fusion of DarkNet, DenseNet, and ResNet, three best individual performing architectures, in distomolar teeth classification problem.

CONCLUSION AND PRACTICAL IMPLICATIONS: In summary, this study has demonstrated the significant potential of CNNs in accurately detecting distomolar teeth in dental radiographs, a critical task for clinical diagnosis and treatment planning. The fusion of CNN architectures, particularly ResNet, Darknet, and DenseNet, has shown exceptional performance, pointing towards the future of artificial intelligence (AI) driven dental diagnostics. Our findings showed that these systems can help clinicians during radiologic examinations.

PMID:39550006 | DOI:10.1016/j.jormas.2024.102151

Environ Res. 2024 Nov 14:120356. doi: 10.1016/j.envres.2024.120356. Online ahead of print.

ABSTRACT

Predicting the bioconcentration of chemical compounds plays a crucial role in assessing environmental risks and toxicological impacts. This study presents a robust multitask deep learning model for predicting the bioconcentration potential. The model can predict the bioconcentration of compounds in multiple categories, including non-bioconcentrative (non-BC), weakly bioconcentrative (weak-BC), and strongly bioconcentrative (strong-BC). We also employed the SHapley Additive exPlanations (SHAP) technology for the model interpretation. The binary classification models (non-BC vs BC and weak-BC vs strong-BC) showed good predictive performance, which achieved accuracy values over 90% and area under the curve (AUC) values with 0.95. The final ternary classification model provided an overall accuracy with 91.11%. Comparative analysis of molecular physicochemical properties showed that the importance of molecular weight, polar surface area, solubility, and hydrogen bonding are important for chemical bioconcentration. Besides, we identified eight structural alerts responsible for chemical bioconcentration. We made the model available as an online tool named BCdpi-predictor, which is accessible at http://bcdpi.sapredictor.cn/. Users can predict the bioconcentration potential of chemical compounds freely. The model has significant implications for environmental policy and regulatory frameworks, such as REACH, by providing a more accurate and interpretable method for assessing chemical risks. We hope that the results of this study can provide helpful tools and meaningful information for chemical bioconcentration prediction in environmental risk assessment.

PMID:39549907 | DOI:10.1016/j.envres.2024.120356

Bone. 2024 Nov 14:117330. doi: 10.1016/j.bone.2024.117330. Online ahead of print.

ABSTRACT

Vertebral compression fractures (VCFs) are the most common type of osteoporotic fractures, yet they are often clinically silent and undiagnosed. Chest frontal radiographs (CFRs) are frequently used in clinical practice and a portion of VCFs can be detected through this technology. This study aimed to develop an automatic artificial intelligence (AI) tool using deep learning (DL) model for the opportunistic screening of VCFs from CFRs. The datasets were collected from four medical centers, comprising 19,145 vertebrae (T6-T12) from 2735 patients. Patients from Center 1, 2 and 3 were divided into the training and internal testing datasets in an 8:2 ratio (n = 2361, with 16,527 vertebrae). Patients from Center 4 were used as the external test dataset (n = 374, with 2618 vertebrae). Model performance was assessed using sensitivity, specificity, accuracy and the area under the curve (AUC). A reader study with five clinicians of different experience levels was conducted with and without AI assistance. In the internal testing dataset, the model achieved a sensitivity of 83.0 % and an AUC of 0.930 at the fracture level. In the external testing dataset, the model demonstrated a sensitivity of 78.4 % and an AUC of 0.942 at the fracture level. The model's sensitivity outperformed that of five clinicians with different levels of experience. Notably, AI assistance significantly improved sensitivity at the patient level for both junior clinicians (from 56.1 % without AI to 81.6 % with AI) and senior clinicians (from 65.0 % to 85.6 %). In conclusion, the automatic AI tool significantly increases clinicians' sensitivity in diagnosing fractures on CFRs, showing great potential for the opportunistic screening of VCFs.

PMID:39549901 | DOI:10.1016/j.bone.2024.117330