PLoS One. 2024 Nov 25;19(11):e0312739. doi: 10.1371/journal.pone.0312739. eCollection 2024.

ABSTRACT

Fine art recognition, traditionally dependent on human expertise, is undergoing a significant transformation with the integration of Artificial Intelligence (AI) and deep learning. This article introduces a novel AI-based approach for fine art recognition, utilizing Convolutional Neural Networks (CNNs) and advanced feature extraction techniques. Addressing the inherent challenges within this domain, we present a systematic methodology to enhance automated fine art recognition. By leveraging critical dataset characteristics such as objective type, genre, material, technique, and department, our method exhibits exceptional performance in classifying fine art pieces across diverse attributes. Our approach significantly improves accuracy and efficiency by integrating advanced feature extraction techniques with a customized CNN architecture. Experimental validation on a benchmark dataset highlights the efficacy of our method, indicating substantial contributions to the interdisciplinary field of fine art analysis.

PMID:39585839 | DOI:10.1371/journal.pone.0312739

PLOS Digit Health. 2024 Nov 25;3(11):e0000436. doi: 10.1371/journal.pdig.0000436. eCollection 2024 Nov.

ABSTRACT

The detection of heart disease using a stethoscope requires significant skill and time, making it expensive and impractical for widespread screening in low-resource environments. Machine learning analysis of heart sound recordings can improve upon the accessibility and accuracy of diagnoses, but existing approaches require further validation on larger and more representative clinical datasets. For many previous algorithms, segmenting the signal into its individual sound components is a key first step. However, segmentation algorithms often struggle to find S1 or S2 sounds in the presence of strong murmurs or noise that significantly alter or mask the expected sound. Segmentation errors then propagate to the subsequent disease classifier steps. We propose a novel recurrent neural network and hidden semi-Markov model (HSMM) algorithm that can both segment the signal and detect a heart murmur, removing the need for a two-stage algorithm. This algorithm formed the 'CUED_Acoustics' entry to the 2022 George B. Moody PhysioNet challenge, where it won the first prize in both the challenge tasks. The algorithm's performance exceeded that of many end-to-end deep learning approaches that struggled to generalise to new test data. As our approach both segments the heart sound and detects a murmur, it can provide interpretable predictions for a clinician. The model also estimates the signal quality of the recording, which may be useful for a screening environment where non-experts are using a stethoscope. These properties make the algorithm a promising tool for screening of abnormal heart murmurs.

PMID:39585836 | DOI:10.1371/journal.pdig.0000436

Bioinformatics. 2024 Nov 25:btae708. doi: 10.1093/bioinformatics/btae708. Online ahead of print.

ABSTRACT

MOTIVATION: Peptides and their derivatives hold potential as therapeutic agents. The rising interest in developing peptide drugs is evidenced by increasing approval rates by the FDA of USA. To identify the most potential peptides, study on peptide-protein interactions presents a very important approach but poses considerable technical challenges. In experimental aspects, the transient nature of peptide-protein interactions (PepPIs) and the high flexibility of peptides contribute to elevated costs and inefficiency. Traditional docking and molecular dynamics simulation methods require substantial computational resources, and the predictive accuracy of their results remain unsatisfactory.

RESULTS: To address this gap, we proposed TPepPro, a Transformer-based model for PepPI prediction. We trained TPepPro on a dataset of 19,187 pairs of peptide-protein complexes with both sequential and structural features. TPepPro utilizes a strategy that combines local protein sequence feature extraction with global protein structure feature extraction. Moreover, TPepPro optimizes the architecture of structural featuring neural network in BN-ReLU arrangement, which notably reduced the amount of computing resources required for peptide-protein interactions prediction. According to comparison analysis, the accuracy reached 0.855 in TPepPro, achieving an 8.1% improvement compared to the second-best model TAGPPI. TPepPro achieved an AUC of 0.922, surpassing the second-best model TAGPPI with 0.844. Moreover, the newly developed TPepPro identify certain PepPIs that can be validated according to previous experimental evidence, thus indicating the efficiency of TPepPro to detect high potential PepPIs that would be helpful for amino acid drug applications.

AVAILABILITY: The source code of TPepPro is available at https://github.com/wanglabhku/TPepPro.

SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.\.

PMID:39585721 | DOI:10.1093/bioinformatics/btae708

Invest Ophthalmol Vis Sci. 2024 Nov 4;65(13):54. doi: 10.1167/iovs.65.13.54.

ABSTRACT

PURPOSE: To compare changes in macular curvature following acute IOP elevation across a range of myopic conditions.

METHODS: We studied 328 eyes from 184 subjects, comprising 32 emmetropic controls (between +2.75 and -2.75 diopters), 50 eyes with high myopia (<-5 diopters; HM), 108 highly myopic with glaucoma (HMG) and 105 pathologic myopia (PM) eyes, and 33 PM with staphyloma (PM+S) eyes. For each eye, we imaged the macula using optical coherence tomography (OCT) under the baseline condition and under acute IOP elevation (to ∼40 mm Hg) achieved through ophthalmodynamometry. We manually aligned the scans (baseline and IOP elevation) using three vascular landmarks in the macula tissue. We then automatically segmented the sclera and the choroid tissues using a deep learning algorithm and extracted the sclera-choroid interface. We calculated the macula curvatures, determined by the radius of curvature of the sclera-choroid interface in the nasal-temporal and superior-inferior direction. Differences in macula curvatures between baseline and elevated IOP scans were calculated at corresponding locations, and the mean curvature difference was reported for each eye.

RESULTS: IOP elevation resulted in a significantly higher macula curvature change along the nasal-temporal direction in the PM+S (13.5 ± 8.2 × 10-5 µm-1), PM (9.0 ± 7.9 × 10-5 µm-1), and HMG (5.2 ± 5.1 × 10-5 µm-1) eyes as compared to HM (3.1 ± 2.7 × 10-5 µm-1) eyes (all P < 0.05). Interestingly, HM and HMG eyes had the same curvature change in the nasal-temporal direction as emmetropic control eyes (4.2 ± 4.3 × 10-5 µm-1).

CONCLUSIONS: Our findings indicate that the macula in HMG, PM, and PM+S eyes showed greater curvature changes under IOP elevation compared to HM and emmetropic eyes. These preliminary results suggest that HM eyes with conditions such as glaucoma or staphyloma are more sensitive to acute IOP elevation.

PMID:39585674 | DOI:10.1167/iovs.65.13.54

Jpn J Radiol. 2024 Nov 25. doi: 10.1007/s11604-024-01699-w. Online ahead of print.

ABSTRACT

PURPOSE: Despite a global decrease in the number of COVID-19 patients, early prediction of the clinical course for optimal patient care remains challenging. Recently, the usefulness of image generation for medical images has been investigated. This study aimed to generate short-term follow-up chest CT images using a latent diffusion model in patients with COVID-19.

MATERIALS AND METHODS: We retrospectively enrolled 505 patients with COVID-19 for whom the clinical parameters (patient background, clinical symptoms, and blood test results) upon admission were available and chest CT imaging was performed. Subject datasets (n = 505) were allocated for training (n = 403), and the remaining (n = 102) were reserved for evaluation. The image underwent variational autoencoder (VAE) encoding, resulting in latent vectors. The information consisting of initial clinical parameters and radiomic features were formatted as a table data encoder. Initial and follow-up latent vectors and the initial table data encoders were utilized for training the diffusion model. The evaluation data were used to generate prognostic images. Then, similarity of the prognostic images (generated images) and the follow-up images (real images) was evaluated by zero-mean normalized cross-correlation (ZNCC), peak signal-to-noise ratio (PSNR), and structural similarity (SSIM). Visual assessment was also performed using a numerical rating scale.

RESULTS: Prognostic chest CT images were generated using the diffusion model. Image similarity showed reasonable values of 0.973 ± 0.028 for the ZNCC, 24.48 ± 3.46 for the PSNR, and 0.844 ± 0.075 for the SSIM. Visual evaluation of the images by two pulmonologists and one radiologist yielded a reasonable mean score.

CONCLUSIONS: The similarity and validity of generated predictive images for the course of COVID-19-associated pneumonia using a diffusion model were reasonable. The generation of prognostic images may suggest potential utility for early prediction of the clinical course in COVID-19-associated pneumonia and other respiratory diseases.

PMID:39585556 | DOI:10.1007/s11604-024-01699-w

Int J Comput Assist Radiol Surg. 2024 Nov 25. doi: 10.1007/s11548-024-03295-9. Online ahead of print.

ABSTRACT

PURPOSE: In the last decade, the development of Deep Learning and its variants, based on the application of artificial neural networks, has reinvigorated Artificial Intelligence (AI). As a result, many new applications of AI in medicine, especially Radiology, have been introduced. This resurgence in AI, and its diverse clinical and nonclinical applications throughout healthcare, requires a thorough understanding to reap the potential benefits and avoid the potential pitfalls.

METHODS: To realize the full potential of AI in medicine, a highly coordinated approach should be undertaken to select, support and finance more highly focused AI projects. By studying and understanding the successes and failures, and strengths and limitations, of AI in Radiology, it is possible to seek and develop the most clinically relevant AI algorithms. The authors have reviewed their clinical practice regarding the use of AI to determine applications in which AI can add both clinical and remunerative benefits.

RESULTS: Review of our policies and applications regarding AI in the Department of Radiology emphasized that, at the time of this writing, AI has been useful in the detection of specific clinical entities for which the AI algorithms have been designed. In addition to helping to reduce diagnostic errors, AI offers an important opportunity to prioritize positive cases, such as pulmonary embolism or intracranial hemorrhage. It has become apparent that the detection of certain conditions, such as incidental and unsuspected cerebral aneurysms can be used to initiate a variety of patient-oriented activities. Finding an unsuspected brain aneurysm is not only of clinical importance to the patient, but the required clinical workup and management of the patient can help generate reimbursement that helps defray the cost of AI implementations. A program for screening, clinical management, and follow-up, facilitated by the AI detection of incidental brain aneurysms, has been implemented at our multi-hospital healthcare system.

CONCLUSION: We feel that it is possible to avoid missed opportunities for AI in Radiology and create AI tools to enhance medical wisdom and improve patient care, within a fiscally responsive environment.

PMID:39585545 | DOI:10.1007/s11548-024-03295-9

Clin Neuroradiol. 2024 Nov 25. doi: 10.1007/s00062-024-01471-7. Online ahead of print.

ABSTRACT

PURPOSE: Deep learning (DL) methods for detecting large vessel occlusion (LVO) in acute ischemic stroke (AIS) show promise, but the effect of computed tomography angiography (CTA) image quality on DL performance is unclear. Our study investigates the impact of improved image quality from Photon Counting Computed Tomography (PCCT) on LVO detection in AIS using a DL-based software prototype developed by a commercial vendor, which incorporates a novel deep learning architecture.

MATERIALS AND METHODS: 443 cases that underwent stroke diagnostics with CTA were included. Positive cases featured vascular occlusions in the Internal Carotid Artery (ICA), M1, and M2 segments of the Middle Cerebral Artery (MCA). Negative cases showed no vessel occlusion on CTA. The performance of the DL-based LVO detection software prototype was assessed using Syngo.via version VB80.

RESULTS: Our study included 267 non-occlusion cases and 176 cases. Among them, 150 cases were scanned via PCCT (no occlusion = 100, ICA and M1 = 41, M2 = 9), while 293 cases were scanned using conventional CT (no occlusion = 167, ICA and M1 = 89, M2 = 37). Independent of scanner type, the algorithm showed sensitivity and specificity of 70.5 and 98.9% for the detection of all occlusions. DL algorithm showed improved performance after excluding M2 occlusions (sensitivity 86.2%). After stratification by scanner type, the algorithm showed significantly a trend towards better performance (p = 0.013) on PCCT CTA images for the detection of all occlusions (sensitivity 84.0%, specificity 99%) compared to CTA images from conventional CT scanner (sensitivity 65.1%, specificity 98.8%). The detection of M2 occlusions was also better on PCCT CTA images (sensitivity 55.6%) compared to conventional scanner CTA images (sensitivity 18.9%), but the sample size for M2 occlusions was limited, and further research is needed to confirm these findings.

CONCLUSION: Our study suggests that PCCT CTA images may offer improved detection of large vessel occlusion, particularly for M2 occlusions. However further research is needed to confirm these findings. One of the limitations of our study is the inability to exclude the presence of a perfusion deficit, despite ruling out vascular occlusion, due to the lack of CT perfusion (CTP) imaging data. Future research may investigate CNNs by leveraging both CTA and CTP images from PCCT for improved performance.

PMID:39585389 | DOI:10.1007/s00062-024-01471-7

Eur J Nucl Med Mol Imaging. 2024 Nov 25. doi: 10.1007/s00259-024-06994-2. Online ahead of print.

ABSTRACT

PURPOSE: Recent development in positron emission tomography (PET) dramatically increased the effective sensitivity by increasing the geometric coverage leading to total-body PET imaging. This encouraging breakthrough brings the hope of ultra-low dose PET imaging equivalent to transatlantic flight with the assistance of deep learning (DL)-based methods. However, conventional DL approaches face limitations in addressing the heterogeneous domain of PET imaging. This study aims to develop a wavelet-based DL method capable of restoring high-quality imaging from ultra-low-dose PET scans.

MATERIALS AND METHODS: In contrast to conventional DL techniques that denoise images in the spatial domain, we introduce WaveNet, a novel approach that inputs wavelet-decomposed frequency components of PET imaging to perform denoising in the frequency domain. A dataset comprising total-body 18F -FDG PET images of 1447, acquired using total-body PET scanners including Biograph Vision Quadra (Siemens Healthineers) and uEXPLORER (United Imaging) in Bern and Shanghai, was utilized for developing and testing the proposed method. The quality of enhanced images was assessed using a customized scoring system, which incorporated weighted global physical metrics and local indices.

RESULTS: Our proposed WaveNet consistently outperforms the baseline UNet model across all levels of dose reduction factors (DRF), with greater improvements observed as image quality decreases. Statistical analysis (p < 0.05) and visual inspection validated the superiority of WaveNet. Moreover, WaveNet demonstrated superior generalizability when applied to two cross-scanner datasets (p < 0.05).

CONCLUSION: WaveNet developed with total-body PET scanners may offer a computational-friendly and robust approach to recover image quality from ultra-low-dose PET imaging. Its adoption may enhance the reliability and clinical acceptance of DL-based dose reduction techniques.

PMID:39585354 | DOI:10.1007/s00259-024-06994-2

Genomics Proteomics Bioinformatics. 2024 Nov 25:qzae084. doi: 10.1093/gpbjnl/qzae084. Online ahead of print.

ABSTRACT

Functional peptides are short amino acid fragments that have a wide range of beneficial functions for living organisms. The majority of previous research focused on mono-functional peptides, but a growing number of multi-functional peptides have been discovered. Although there have been enormous experimental efforts to assay multi-functional peptides, only a small fraction of millions of known peptides have been explored. Effective and precise techniques for identifying multi-functional peptides can facilitate their discovery and mechanistic understanding. In this article, we presented a method iMFP-LG for identifying multi-functional peptides based on protein language models (pLMs) and graph attention networks (GATs). Comparison results showed that iMFP-LG outperforms state-of-the-art methods on both multi-functional bioactive peptides and multi-functional therapeutic peptides datasets. The interpretability of iMFP-LG was also illustrated by visualizing attention patterns in pLMs and GATs. Regarding the outstanding performance of iMFP-LG on the identification of multi-functional peptides, we employed iMFP-LG to screen novel candidate peptides with both ACP and AMP functions from millions of known peptides in the UniRef90. As a result, 8 candidate peptides were identified, and 1 candidate that exhibits both antibacterial and anticancer effects was confirmed through molecular structure alignment and biological experiments. We anticipate that iMFP-LG can assist in the discovery of multi-functional peptides and contribute to the advancement of peptide drug design.

PMID:39585308 | DOI:10.1093/gpbjnl/qzae084

Noncoding RNA. 2024 Nov 8;10(6):55. doi: 10.3390/ncrna10060055.

ABSTRACT

Understanding the structures of noncoding RNAs (ncRNAs) is important for the development of RNA-based therapeutics. There are inherent challenges in employing current experimental techniques to determine the tertiary (3D) structures of RNAs with high complexity and flexibility in folding, which makes computational methods indispensable. In this study, we compared the utilities of three advanced computational tools, namely RNAComposer, Rosetta FARFAR2, and the latest AlphaFold 3, to predict the 3D structures of various forms of RNAs, including the small interfering RNA drug, nedosiran, and the novel bioengineered RNA (BioRNA) molecule showing therapeutic potential. Our results showed that, while RNAComposer offered a malachite green aptamer 3D structure closer to its crystal structure, the performances of RNAComposer and Rosetta FARFAR2 largely depend upon the secondary structures inputted, and Rosetta FARFAR2 predictions might not even recapitulate the typical, inverted "L" shape tRNA 3D structure. Overall, AlphaFold 3, integrating molecular dynamics principles into its deep learning framework, directly predicted RNA 3D structures from RNA primary sequence inputs, even accepting several common post-transcriptional modifications, which closely aligned with the experimentally determined structures. However, there were significant discrepancies among three computational tools in predicting the distal loop of human pre-microRNA and larger BioRNA (tRNA fused pre-miRNA) molecules whose 3D structures have not been characterized experimentally. While computational predictions show considerable promise, their notable strengths and limitations emphasize the needs for experimental validation of predictions besides characterization of more RNA 3D structures.

PMID:39585047 | DOI:10.3390/ncrna10060055

Epigenomics. 2024 Nov 25:1-9. doi: 10.1080/17501911.2024.2432854. Online ahead of print.

ABSTRACT

Over the past century, human lifespan has increased remarkably, yet the inevitability of aging persists. The disparity between biological age, which reflects pathological deterioration and disease, and chronological age, indicative of normal aging, has driven prior research focused on identifying mechanisms that could inform interventions to reverse excessive age-related deterioration and reduce morbidity and mortality. DNA methylation has emerged as an important predictor of age, leading to the development of epigenetic clocks that quantify the extent of pathological deterioration beyond what is typically expected for a given age. Machine learning technologies offer promising avenues to enhance our understanding of the biological mechanisms governing aging by further elucidating the gap between biological and chronological ages. This perspective article examines current algorithmic approaches to epigenetic clocks, explores the use of machine learning for age estimation from DNA methylation, and discusses how refining the interpretation of ML methods and tailoring their inferences for specific patient populations and cell types can amplify the utility of these technologies in age prediction. By harnessing insights from machine learning, we are well-positioned to effectively adapt, customize and personalize interventions aimed at aging.

PMID:39584810 | DOI:10.1080/17501911.2024.2432854

Stress Health. 2024 Nov 25:e3510. doi: 10.1002/smi.3510. Online ahead of print.

ABSTRACT

Earthquakes, as significant natural disasters, still cannot be accurately predicted today. Although current earthquake early warning systems can provide alerts several seconds in advance, acute stress responses (ASR) in emergency situations can waste these precious escape seconds. To investigate the correlation between personality and ASR, this study collected the temperament and character of all participants using the Chen Huichang-60 Temperament Scale and the DISC Personality Inventory. In addition, this study simulated growing earthquakes in an earthquake experience hall, collecting heart rate variability and respiration signal variations throughout the process from subjects. Multivariate analysis of variance (MANOVA) and Toeplitz Inverse Covariance-Based Clustering methods were used to analyse the differences and connections between them. Furthermore, this study employed a deep learning model that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) to predict ASR across personalities. This model used datasets from the majority dataset of a certain personality and a single participant, respectively, and showed different performance. The results are as follows. After categorising participants based on personality test results, MANOVA revealed significant differences between the personality groups Influence-Choleric and Influence-Sanguine (p = 0.001), Influence-Phlegmatic and Steadiness-Sanguine (p = 0.023), Influence-Sanguine and Steadiness-Sanguine (p < 0.001) and Influence-Sanguine and Steadiness-Phlegmatic (p < 0.001), as well as across different earthquake stages (p < 0.01). The clustering method quantified stress responses over time for different personalities and labelled ASR levels for use in supervised learning. Ultimately, the CNN-LSTM model performed predictions of ASR using both personality and individual datasets, achieving the AUC of 0.795 and 0.72, demonstrating better prediction and classification effectiveness with the former. This study provides a new personality-based method for earthquake stress management, creating possibilities for longitudinal stress research and prediction. It aids the general public in comprehending their own acute stress and allows authorities and communities to make practical, efficient disaster evacuation plans based on the overall situation of public ASR.

PMID:39584748 | DOI:10.1002/smi.3510

Pest Manag Sci. 2024 Nov 25. doi: 10.1002/ps.8554. Online ahead of print.

ABSTRACT

BACKGROUND: Reliable, fast, and accurate weed detection in farmland is crucial for precision weed management but remains challenging due to the diverse weed species present across different fields. While deep learning models for direct weed detection have been developed in previous studies, creating a training dataset that encompasses all possible weed species, ecotypes, and growth stages is practically unfeasible. This study proposes a novel approach to detect weeds by integrating semantic segmentation with image processing. The primary aim is to simplify the weed detection process by segmenting crop pixels and identifying all vegetation outside the crop mask as weeds.

RESULTS: The proposed method employs a semantic segmentation model to generate a mask of corn (Zea mays L.) crops, identifying all green plant pixels outside the mask as weeds. This indirect segmentation approach reduces model complexity by avoiding the need for direct detection of diverse weed species. To enhance real-time performance, the semantic segmentation model was optimized through knowledge distillation, resulting in a faster, lighter-weight inference. Experimental results demonstrated that the DeepLabV3+ model, after applying knowledge distillation, achieved an average accuracy (aAcc) exceeding 99.5% and a mean intersection over union (mIoU) across all categories above 95.5%. Furthermore, the model's operating speed surpassed 34 frames per second (FPS).

CONCLUSION: This study introduces a novel method that accurately segments crop pixels to form a mask, identifying vegetation outside this mask as weeds. By focusing on crop segmentation, the method avoids the complexity associated with diverse weed species, varying densities, and different growth stages. This approach offers a practical and efficient solution to facilitate the training of effective computer vision models for precision weed detection and control. © 2024 Society of Chemical Industry.

PMID:39584373 | DOI:10.1002/ps.8554

J Alzheimers Dis. 2024 Nov 25:13872877241300278. doi: 10.1177/13872877241300278. Online ahead of print.

ABSTRACT

BACKGROUND: Quantitative susceptibility mapping (QSM) is pivotal for analyzing neurodegenerative diseases. However, accurate hippocampal segmentation remains a challenge.

OBJECTIVE: This study introduces a method for extracting hippocampal magnetic susceptibility values using a convolutional neural network (CNN) model referred to as 3D residual UNET.

METHODS: The model was pre-trained on whole QSM images and hippocampal segmentations from 3D T1-weighted images of 297 patients with Alzheimer's disease and mild cognitive impairment. Fine-tuning was conducted through manually annotated hippocampal segmentations from the QSM images of 60 patients. The performance was assessed using the Dice similarity coefficient (DSC) and Pearson correlation coefficient.

RESULTS: The developed model was applied to another 98 patients, 49 with AD and 49 with mild cognitive impairment (MCI), and the correlation between the hippocampal magnetic susceptibility and volume was evaluated. The mean DSC for the hippocampal segmentation model was 0.716 ± 0.045. The correlation coefficient between the magnetic susceptibility values derived from manual segmentation and the CNN model was 0.983. The Pearson correlation coefficient between magnetic susceptibility and hippocampal volume from the CNN model was -0.252 (p = 0.012) on the left side and -0.311 (p = 0.002) on the right.

CONCLUSIONS: The 3D residual UNET model enhances hippocampal analysis precision using QSM, which is capable of accurately extracting magnetic susceptibility.

PMID:39584366 | DOI:10.1177/13872877241300278

Sheng Wu Gong Cheng Xue Bao. 2024 Nov 25;40(11):3912-3929. doi: 10.13345/j.cjb.240087.

ABSTRACT

Proteins with specific functions and characteristics play a crucial role in biomedicine and nanotechnology. De novo protein design enables the customization of sequences to produce proteins with desired structures that do not exist in the nature. In recent years, with the rapid development of artificial intelligence (AI), deep learning-based generative models have increasingly become powerful tools, enabling the design of functional proteins with atomic-level precision. This article provides an overview of the evolution of de novo protein design, with focus on the latest algorithmic models, and then analyzes existing challenges such as low design success rates, insufficient accuracy, and dependence on experimental validation. Furthermore, this article discusses the future trends in protein design, aiming to provide insights for researchers and practitioners in this field.

PMID:39584325 | DOI:10.13345/j.cjb.240087

Echocardiography. 2024 Dec;41(12):e70042. doi: 10.1111/echo.70042.

ABSTRACT

In the last decade, artificial intelligence (AI) has influenced the field of cardiac computed tomography (CT), with its scope further enhanced by advanced methodologies such as machine learning (ML) and deep learning (DL). The AI-driven techniques leverage large datasets to develop and train algorithms capable of making precise evaluations and predictions. The realm of cardiac CT is expanding day by day and multiple tools are offered to answer different questions. Coronary artery calcium score (CACS) and CT angiography (CTA) provide high-resolution images that facilitate the detailed anatomical evaluation of coronary plaque burden. New tools such as myocardial CT perfusion (CTP) and fractional flow reserve (FFRCT) have been developed to add a functional evaluation of the stenosis. Moreover, epicardial adipose tissue (EAT) is gaining interest as its role in coronary artery plaque development has been deepened. Seen the great added value of these tools, the demand for new exams has increased such as the burden on imagers. Due to its ability to fast compute multiple data, AI can be helpful in both the acquisition and post-processing phases. AI can possibly reduce radiation dose, increase image quality, and shorten image analysis time. Moreover, different types of data can be used for risk assessment and patient risk stratification. Recently, the focus of the scientific community on AI has led to numerous studies, especially on CACS and CTA. This narrative review concentrates on AI's role in the post-processing of CACS, CTA, FFRCT, CTP, and EAT, discussing both current capabilities and future directions in the field of cardiac imaging.

PMID:39584228 | DOI:10.1111/echo.70042

J Bone Oncol. 2024 Nov 4;49:100644. doi: 10.1016/j.jbo.2024.100644. eCollection 2024 Dec.

ABSTRACT

OBJECTIVE: The incidence of osteosarcoma (OS) is low, but primary malignant bone tumors rank third among the causes of death in cancer patients under the age of 20. Currently, analysis of cellular structure and tumor morphology through microscopic images remains one of the main diagnostic methods for osteosarcoma. However, this completely manual approach is tedious, time-consuming, and difficult to diagnose accurately due to the similarities in certain characteristics of malignant and benign tumors.

METHODS: Leveraging the potential of artificial intelligence (AI) in assessing and classifying images, this study explored a modified CA-MobileNet V3 model that was embedded into innovative microscope products to enhance the microscope's feature extraction capabilities and help reduce misclassification during diagnosis.

RESULTS: The intelligent recognition model method introduced in this paper has significant advantages in retrieval and classification of osteosarcoma cells and other cell types. Compared with models such as ShuffleNet V2, EfficientNet V2, Mobilenet V3 (without transfer learning), TL-MobileNet V3 (with transfer learning), etc., the model size is only 5.33 MB, is a lightweight model, and the accuracy of the improved model reached 98.69 %. In addition, the artificial intelligence microscope (AIM) with integrated design based on this model can also help improve diagnostic efficiency.

CONCLUSION: The innovative method of the CA-MobileNet V3 automatic classification model based on deep learning provides an efficient and reliable solution for the pathological diagnosis of osteosarcoma. This study contributes to medical image analysis and provides doctors with an accurate and valuable tool for microscopic diagnosis. It also promotes the advancement of artificial intelligence in medical imaging technology.

PMID:39584044 | PMC:PMC11585738 | DOI:10.1016/j.jbo.2024.100644

Surg Open Sci. 2024 Oct 30;22:53-60. doi: 10.1016/j.sopen.2024.10.010. eCollection 2024 Dec.

ABSTRACT

BACKGROUND: The COVID-19 pandemic prompted the transition of all teaching and learning of final-year General Surgery students to an online platform. Despite the utility of online methods, challenges exist such as a sense of impersonal learning, and poor student engagement. Student engagement with course content is important for deep learning. An Online Student Engagement Scale (OSE) and a revised Biggs Two-Factor Study Process Questionnaire (R-SPQ-2F) were used to evaluate student engagement and learning approaches respectively.

METHODS: A cross-sectional study was conducted in 2021 at a South African university. The OSE and R-SPQ-2F online survey tools were administered to all final-year students (n = 325) enrolled in the surgical online module. Quantitative data was collected, and the data was analysed statistically using R-Statistical computing software. Results are presented in the form of descriptive and inferential statistics. The reliability of the tools was evaluated by Cronbach's alpha.

RESULTS: The survey response rate was 35.4 % (115/325). Students were engaged at a high level, and the median (IQR) scores of the OSE tool were 71.0 (63.0-78.0). Overall, students adopted a deep approach (DA) to learning, with median (IQR) scores of 34.0 (30.0-39.0) on the R-SPQ-2F tool. There was a moderate positive correlation between the total OSE score and DA (0.53, p < 0.001). Both the OSE and R-SPQ-2F tools showed an acceptable level of internal consistency of 0.893 and 0.806 respectively.

CONCLUSIONS: Student engagement was associated with deep learning approaches. The OSE and R-SPQ-2F tools were reliable tools to measure student engagement and learning approaches among medical students.

PMID:39584028 | PMC:PMC11582439 | DOI:10.1016/j.sopen.2024.10.010

IEEE Int Conf Commun. 2024 Jun;2024:214-219. doi: 10.1109/icc51166.2024.10622995. Epub 2024 Aug 20.

ABSTRACT

In this study, we introduce BedDot, the first contact-free and bed-mounted continuous blood pressure monitoring sensor. Equipped with a seismic sensor, BedDot eliminates the need for external wearable devices and physical contact, while avoiding privacy or radiation concerns associated with other technologies such as cameras or radars. Using advanced preprocessing techniques and innovative AI algorithms, we extract time-series features from the collected bedseismogram signals and accurately estimate blood pressure with remarkable stability and robustness. Our user-friendly prototype has been tested with over 75 participants, demonstrating exceptional performance that meets all three major industry standards, which are Association for the Advancement of Medical Instrumentation (AAMI), Food and Drug Administration (FDA) and the British and Irish Hypertension Society (BHS), and outperforms current state-of-the-art deep learning models for time series analysis. As a non-invasive solution for monitoring blood pressure during sleep and assessing cardiovascular health, BedDot holds immense potential for revolutionizing the field.

PMID:39583890 | PMC:PMC11583795 | DOI:10.1109/icc51166.2024.10622995

Biomed Rep. 2024 Nov 12;22(1):11. doi: 10.3892/br.2024.1889. eCollection 2025 Jan.

ABSTRACT

The potential of artificial intelligence (AI) to significantly transform numerous aspects of contemporary civilization is substantial. Advancements in research show an increasing interest in creating AI solutions in the healthcare sector. This interest is driven by the broad spectrum and extensive nature of easily accessible patient data-including medical imaging, digitized data collection, and electronic health records - and by the ability to analyze and interpret complex data, facilitating more accurate and timely diagnoses. This review's goal is to provide a comprehensive overview of the advancements achieved by AI in healthcare, to elucidate the present state of AI in enhancing the healthcare system and improving the quality and efficiency of healthcare decision making, and to discuss selected medical applications of AI. Furthermore, the barriers and constraints that may impede the use of AI in healthcare are outlined, and the potential future directions of AI-augmented healthcare systems are discussed.

PMID:39583770 | PMC:PMC11582508 | DOI:10.3892/br.2024.1889