Med Phys. 2025 Feb 17. doi: 10.1002/mp.17704. Online ahead of print.

ABSTRACT

BACKGROUND: Plan review is critical for safely delivering radiation dose to a patient under radiotherapy and mainly performed by medical physicist in routine clinical practice. Recently, the deep-learning models have been used to assist this manual process. As black-box models the reason for their predictions are unknown. Thus, it is important to improve the model interpretability to make them more reliable for clinical deployment.

PURPOSE: To alleviate this issue, a deep generative model, adversarial autoencoder networks (AAE), was employed to automatically detect anomalies in intensity-modulated radiotherapy plans.

METHODS: The typical plan parameters (collimator position, gantry angle, monitor unit, etc.) were collected to form a feature vector for the training sample. The reconstruction error was the difference between the output and input of the model. Based on the distribution of reconstruction errors of the training samples, a detection threshold was determined. For a test plan, its reconstruction error obtained by the learned model was compared with the threshold to determine its category (anomaly or regular). The model was tested with four network settings. It was also compared with the vanilla AE and the other six classic models. The area under receiver operating characteristic curve (AUC) along with other statistical metrics was employed for evaluation.

RESULTS: The AAE model achieved the highest accuracy (AUC = 0.997). The AUCs of the other seven classic methods are 0.935 (AE), 0.981 (K-means), 0.896 (principle component analysis), 0.978 (one-class support vector machine), 0.934 (local outlier factor), and 0.944 (hierarchical density-based spatial clustering of applications with noise), and 0.882 (isolation forest). This indicates that AAE model could detect more anomalous plans with less false positive rate.

CONCLUSIONS: The AAE model can effectively detect anomaly in radiotherapy plans for lung cancer patients. Comparing with the vanialla AE and other classic detection models, the AAE model is more accurate and transparent. The proposed AAE model can improve the interpretability of the results for radiotherapy plan review.

PMID:39960256 | DOI:10.1002/mp.17704

J Diabetes Sci Technol. 2025 Feb 17:19322968251317521. doi: 10.1177/19322968251317521. Online ahead of print.

ABSTRACT

INTRODUCTION: Early detection of diabetic foot complications is essential for effective management and prevention of complications. Artificial intelligence (AI) technology based on digital image analysis offers a promising noninvasive method for diabetic foot screening. This systematic review aims to identify a study on the development of an AI model for diabetic foot screening using digital image analysis.

METHODOLOGY: The review scrutinized articles published between 2018 and 2023, sourced from PubMed, ProQuest, and ScienceDirect. The keyword-based search resulted in 2214 relevant articles and nine articles that met the inclusion criteria. The article quality assessment was done through Quality Assessment of Diagnostic Accuracy Studies (QUADAS). Data were extracted and analyzed using NVivo.

RESULTS: Thermal imagery or foot thermogram was the main data source, with plantar temperature distribution patterns as an important indicator. Deep learning methods, specifically artificial neural networks (ANNs) and convolutional neural networks (CNNs), are the most commonly used methods. The highest performance is demonstrated by the ANN model with MATLAB's Image Processing Toolbox that is able to classify each type of macula with 97.5% accuracy. The findings show the great potential of AI in improving the accuracy and efficiency of diabetic foot screening.

CONCLUSION: This research provides important insights into the development of AI in digital image-based diabetic foot screening. Future studies need to focus on evaluating clinical applicability, including ethical aspects and patient data security, as well as developing more comprehensive data sets.

PMID:39960227 | DOI:10.1177/19322968251317521

Hepatology. 2025 Mar 1;81(3):755-756. doi: 10.1097/HEP.0000000000001220. Epub 2025 Feb 17.

NO ABSTRACT

PMID:39960202 | DOI:10.1097/HEP.0000000000001220

Minerva Med. 2025 Feb 17. doi: 10.23736/S0026-4806.25.09643-0. Online ahead of print.

ABSTRACT

BACKGROUND: Early diagnosis of idiopathic pulmonary fibrosis (IPF) is fundamental to slow disease progression; multidisciplinary teams (MDTs) play a central role in posing the final diagnosis of IPF, thus aiming to improve patient outcomes. However, the practical implementation of MDTs in clinical real-life settings may be hindered by the lack of local expertise or time constraints, with the diagnosis being made without the support of complementary professional health care figures. This study aims to evaluate the impact of MDT meetings on the latency between the symptom onset and the final diagnosis of IPF.

METHODS: Patients referred to a regional center for IPF between January 2019 and August 2019 were included. The length of time to pose a definite diagnosis by means of MDT evaluation was compared with that of patients diagnosed elsewhere (no MDT evaluation) in an observational case-control investigation.

RESULTS: Among 24 IPF patients, those evaluated by MDT (M/F: 14/2, age: 69.8±8.2 yrs) showed a time interval from the first outpatient visit to the definite diagnosis of 3±2.3 months; on the other hand, patients in the control group (M/F: 7/1, age: 76.9±7.7 yrs) showed a time interval of 12.8±9.4 months (P=0.02). The time elapsed between the onset of symptoms and the definite diagnosis was 11.1±5.3 months for patients evaluated within the MDT, compared to 33.8±21.5 months for the control group (P=0.02).

CONCLUSIONS: These exploratory findings confirm the essential role of the MDT in the early diagnosis of IPF, thus discouraging the acquisition of diagnosis solely on individual basis. The current findings highlight the need for the implementation of MDTs in clinical practice to optimize patient care.

PMID:39960753 | DOI:10.23736/S0026-4806.25.09643-0

Funct Plant Biol. 2025 Feb;52:FP24051. doi: 10.1071/FP24051.

ABSTRACT

We applied a systems biology approach to gain a deep insight into the regulatory mechanisms of barley (Hordeum vulgare ) under drought and waterlogging stress conditions. To identify informative models related to stress conditions, we constructed meta-analysis and two distinct weighted gene co-expression networks. We then performed module trait association analyses. Additionally, we conducted functional enrichment analysis of significant modules to shed light on the biological performance of underlying genes in the two contrasting stresses. In the next step, we inferred the gene regulatory networks between top hub genes of significant modules, kinases, and transcription factors (TFs) using a machine learning algorithm. Our results showed that at power=10, the scale-free topology fitting index (R2) was higher than 0.8 and the connectivity mean became stable. We identified 31 co-expressed gene modules in barley, with 13 and 14 modules demonstrating significant associations with drought and waterlogging stress, respectively. Functional enrichment analysis indicated that these stress-responsive modules are involved in critical processes, including ADP-rybosylation factors (ARF) protein signal transduction, ethylene-induced autophagy, and phosphoric ester hydrolase activity. Specific TFs and kinases, such as C2C2-GATA, HB-BELL, and MADS-MIKC, were identified as key regulators under these stress conditions. Furthermore, certain TFs and kinases established unique connections with hub genes in response to waterlogging and drought conditions. These findings enhance our understanding of the molecular networks that modulate barley's response to drought and waterlogging stresses, offering insights into the regulatory mechanisms essential for stress adaptation.

PMID:39960829 | DOI:10.1071/FP24051

Elife. 2025 Feb 17;12:RP87930. doi: 10.7554/eLife.87930.

ABSTRACT

Cell mechanical properties determine many physiological functions, such as cell fate specification, migration, or circulation through vasculature. Identifying factors that govern the mechanical properties is therefore a subject of great interest. Here, we present a mechanomics approach for establishing links between single-cell mechanical phenotype changes and the genes involved in driving them. We combine mechanical characterization of cells across a variety of mouse and human systems with machine learning-based discriminative network analysis of associated transcriptomic profiles to infer a conserved network module of five genes with putative roles in cell mechanics regulation. We validate in silico that the identified gene markers are universal, trustworthy, and specific to the mechanical phenotype across the studied mouse and human systems, and demonstrate experimentally that a selected target, CAV1, changes the mechanical phenotype of cells accordingly when silenced or overexpressed. Our data-driven approach paves the way toward engineering cell mechanical properties on demand to explore their impact on physiological and pathological cell functions.

PMID:39960760 | DOI:10.7554/eLife.87930

J Occup Health. 2025 Feb 17:uiaf013. doi: 10.1093/joccuh/uiaf013. Online ahead of print.

ABSTRACT

The novel properties and functions of nanomaterials have naturally alerted the toxicologists to the fact that such materials may also have novel effects on the human body and living organisms. In particular, materials with high stability or biopersisteny have been shown to have a tendency to accumulate in the body, leading to chronic toxicity including carcinogenicity. However, at the early stages of toxicity research, the information is often limited to the effects of short-term exposure studies, and findings on chronic effects are very much delayed. In this context, it was rather exceptional that studies on multiwall carbon nanotubes (MWCNTs) have started with the verification of their potential to induce mesothelioma. This toxicological endpoint was expected on the basis of existing knowledge of asbestos and asbestos-like fiber particles. This movement has led to the achievement of the original mission of the "Modern Toxicology", which is "to achieve a win-win situation where both industrial promotion and safety assurance are ensured by communicating and sharing toxicity information to developers and consumers at a stage before mass production and consumption begins, that is, before massive exposure of the general public begins". Inaccurate toxicity assessments of asbestos in the 1980s and 1990s allowed its spread to our living environment, which is difficult to decontaminate, and the damage still continues to this day. However, the case described here could be an example of realizing the proposition that 'nanomaterials, the flagship of high technology, must not repeat the same mistakes.'

PMID:39960454 | DOI:10.1093/joccuh/uiaf013

Plant Cell Environ. 2025 Feb 17. doi: 10.1111/pce.15430. Online ahead of print.

ABSTRACT

Legumes have the beneficial capacity to establish symbiotic interactions with rhizobia, which provide their host plants with fixed nitrogen. However, in the presence of nitrogen, this process is rapidly repressed to avoid unnecessary investments of carbon in the symbiosis. Several players involved in regulating nodulation in response to nitrate availability have been identified, including peptide hormones, microRNAs and transcription factors. Nevertheless, how these molecular players are linked to each other and what underlying molecular mechanisms are at play to inhibit nodulation remain unresolved. Nitrate-mediated control of nodulation seems to differ between model legumes, such as Medicago and Lotus, compared to legume crops such as soybean. A deeper understanding of these regulatory processes, particularly in soybean, is expected to contribute to establishing increased nodulation efficiency in modern agricultural systems, hence improving sustainability by reducing the need for environmentally hazardous nitrogen fertilizers. This review describes the state of the art of nitrate-regulated nodulation in soybean, while drawing parallels with molecular mechanisms described in other legumes and addressing knowledge gaps that require future study.

PMID:39960031 | DOI:10.1111/pce.15430

J Biochem Mol Toxicol. 2025 Feb;39(2):e70172. doi: 10.1002/jbt.70172.

ABSTRACT

Cyclophosphamide (CYC) is one of the most potent antineoplastic drugs; however, hepatonephrotoxicity, observed following its use, remains one of its most severe side effects. Previous studies have reported that syringaldehyde (SYA), a flavonoid compound, exhibits anti-inflammatory and antioxidant properties. However, it is unclear whether SYA has any effects on hepatonephrotoxicity caused by the side effects of antineoplastic drugs. In the present research, we thoroughly evaluated the effects of SYA on cyclophosphamide-induced hepatonephrotoxicity in a mouse model, focusing on Nrf2/HO-1 pathway activation. In the present study, SYA (25 and 50 mg/kg, p.o.) and CYC (30 mg/kg, i.p.) were delivered to male mice for 10 days to induce hepatonephrotoxicity. SYA treatment alleviated the elevated levels of AST, ALT, BUN, and creatinine caused by CYC. It further suppressed lipid peroxidation by lowering MDA levels and enhanced antioxidant defense by elevating GSH, SOD, and CAT levels. Additionally, SYA increased the mRNA expression levels of HO-1, Nrf2, and Bcl-2, which had been reduced due to oxidative stress, inflammatory, and apoptotic pathways, while suppressing the elevated gene expression levels of NFκB, TNF-α, Bax, and Cas-3. Furthermore, SYA regulated the altered protein expression levels of Nrf2, Cas-3, Bax, and Bcl-2 induced by CYC. Microscopically, SYA also mitigated liver and kidney tissue damage caused by CYC. In conclusion, SYA significantly reduced CYC-induced hepatonephrotoxicity by inhibiting inflammation, oxidative stress, and apoptosis by employing the Nrf2/NFκB/HO-1 pathway. These findings indicate that SYA has the possibility as a treatment option agent in the case of prevention of liver and kidney damage.

PMID:39959927 | DOI:10.1002/jbt.70172

In Silico Pharmacol. 2025 Feb 13;13(1):27. doi: 10.1007/s40203-025-00307-7. eCollection 2025.

ABSTRACT

The global spread of monkeypox, caused by the double-stranded DNA monkeypox virus (MPXV), has underscored the urgent need for effective antiviral treatments. In this study, we aim to identify a potent inhibitor for MPXV DNA polymerase (DNAP), a critical enzyme in the virus replication process. Using a computational drug repurposing approach, we performed a virtual screening of 1615 FDA-approved drugs based on drug-likeness and molecular docking against DNAP. Among these, 1430 compounds met Lipinski's rule of five for drug-likeness, with Doxycycline emerging as the most promising competitive inhibitor, binding strongly to the DNAP active site with a binding affinity of - 9.3 kcal/mol. This interaction involved significant hydrogen bonds, electrostatic interactions, and hydrophobic contacts, with Doxycycline demonstrating a stronger affinity than established antivirals for smallpox, including Cidofovir, Brincidofovir, and Tecovirimat. Stability and flexibility analyses through a 200 ns molecular dynamics simulation and normal mode analysis confirmed the robustness of Doxycycline binding to DNAP. Overall, our results suggest Doxycycline as a promising candidate for monkeypox treatment, though additional experimental and clinical studies are needed to confirm its therapeutic potential and clinical utility.

SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s40203-025-00307-7.

PMID:39958784 | PMC:PMC11825436 | DOI:10.1007/s40203-025-00307-7

Mini Rev Med Chem. 2025 Feb 14. doi: 10.2174/0113895575360959250117073046. Online ahead of print.

ABSTRACT

Sleep disorders and the resultant sleep deprivation (SD) are very common nowadays, resulting in depressed mood, poor memory and concentration, and various important changes in health, performance and safety. They may provoke further impairment of the cell lining of the blood vessels, as acting as a risk factor for cardiovascular disease (CVD) onset and progression. SD may lead to low neuronal regaining and plasticity, drastically affecting brain function. Thus, SD is a known risk factor for mental, behavioral and developmental disorders. Due to the inflammatory and oxidative stressful nature of SD, immune response modulation and antioxidants could be another therapeutic approach, apart from the already known symptomatic treatment with sedatives. Additionally, many drugs approved for other indications and under investigation, have been revisited due to their wide array of pharmacological activities. This review summarizes the main aspects of SD pathology and SD interrelated comorbidities and presents direct and indirect antioxidant molecules and drugs with multi-targeting potential that could assist in the prevention or management of these factors. A number of research groups have investigated well-known antioxidant compounds with multi-targeting cores, combining structural characteristics with properties including antiinflammatory, metal chelatory, gene transcription and immune modulatory that may add towards the effective SD and its associated comorbidities treatment.

PMID:39957704 | DOI:10.2174/0113895575360959250117073046

Pharmacogenomics. 2025 Feb 16:1-10. doi: 10.1080/14622416.2025.2463866. Online ahead of print.

ABSTRACT

AIM: To share the experience of a US national reference laboratory, offering genotyping for TPMT and NUDT15, TPMT enzyme phenotyping and detection of thiopurine metabolites.

METHODS: Retrospective review of archived datasets related to thiopurines drug therapy including patients' data that underwent TPMT and NUDT15 genotyping, and smaller data sets where genotyping was performed with TPMT enzyme levels (phenotyping) +/- therapeutic drug monitoring (TDM).

RESULTS: Thirteen percent of patients had variants in one or both genes tested. Testing for NUDT15 revealed 3.9% additional patients requiring thiopurines dosing recommendations. A correlation between TPMT enzyme activity and TPMT polymorphisms (odds ratio OD = 71.41, p-value <0.001) and between older age and higher enzyme levels (OD = 0.98, p-value = 0.002) was identified. No correlation between sex and TPMT enzyme levels, nor between TPMT genotyping and the level of thiopurine metabolites was found.

CONCLUSION: Adding NUDT15 to TPMT genotyping, identified additional 3.9% patients to benefit from thiopurine dose modifications. A significant correlation between genetic variants in TPMT and TPMT enzyme levels and between age and enzyme levels was established, while no correlation was identified between sex and enzyme levels nor between TPMT variation and thiopurine metabolites. Providers rely more significantly on genotyping only approach, rather than genotyping and phenotyping.

PMID:39957149 | DOI:10.1080/14622416.2025.2463866

Br J Clin Pharmacol. 2025 Feb 16. doi: 10.1002/bcp.70004. Online ahead of print.

ABSTRACT

AIMS: Phenoconversion, a genotype-phenotype mismatch, challenges a successful implementation of personalized medicine. The aim of this study was to detect and determine phenoconversion using the solanidine metabolites 3,4-seco-solanidine-3,4-dioic acid (SSDA) and 4-OH-solanidine as diet-derived cytochrome P450 2D6 (CYP2D6) biomarkers in a geriatric, multimorbid cohort with high levels of polypharmacy.

METHODS: Blood samples and data of geriatric, multimedicated patients were collected during physician counsel (CT: NCT05247814). Solanidine and its metabolites were determined via liquid chromatography/tandem mass spectrometry and used for CYP2D6 phenotyping. CYP2D6 genotyping was performed and activity scores (AS) were assigned. Complete medication intake was assessed. A shift of the AS predicted via genotyping as measured by phenotyping was calculated.

RESULTS: Solanidine and its metabolites were measured in 88 patients with complete documentation of drug use. Patients had a median age of 83 years (interquartile range [IQR] 77-87) and the majority (70.5%, n = 62) were female. Patients took a median of 15 (IQR 12-17) medications. The SSDA/solanidine metabolic ratio correlated significantly with the genotyping-derived AS (P < .001) and clearly detected poor metabolizers. In the model adjusted for age, sex, Charlson Comorbidity Index and estimated glomerular filtration rate each additional CYP2D6 substrate/inhibitor significantly lowered the expected AS by 0.53 (95% confidence interval 0.85-0.21) points in patients encoding functional CYP2D6 variants (R2 = 0.242).

CONCLUSIONS: Phenotyping of CYP2D6 activity by measurement of diet-derived biomarkers elucidates phenoconversion in geriatric patients. These results might serve as a prerequisite for the validation and establishment of a bedside method to measure CYP2D6 activity in multimorbid patients for successful application of personalized drug prescribing.

PMID:39957076 | DOI:10.1002/bcp.70004

ACS Omega. 2025 Jan 28;10(5):4336-4352. doi: 10.1021/acsomega.4c06520. eCollection 2025 Feb 11.

ABSTRACT

1,2,4-Oxadiazoles are well recognized for their exceptional physical, chemical, and pharmacokinetic properties, making them promising candidates for various therapeutic applications. These include treatments for cystic fibrosis, Duchenne muscular dystrophy, Alzheimer's disease, and a broad spectrum of other therapeutic interventions such as antituberculosis, anticancer, antibiotic, anti-inflammatory, and anticonvulsant activities. In this study, single crystals of a novel 1,2,4-oxadiazole derivative, methyl-4-(5-(4-(octyloxy)phenyl)-1,2,4-oxadiazol-3-yl)benzoate, were grown by a slow evaporation technique. The structural elucidation was performed using X-ray diffraction analysis, confirming the compound's crystalline structure in the triclinic system. The analysis revealed a linear conformation with bond lengths closely aligned with Cambridge Structural Database (CSD) averages, signifying high precision in the molecular structure. A detailed CSD study identified nine principal configurations of the phenyl octyloxy moiety, underscoring the structural diversity of the compound. Hirshfeld surface analysis highlighted the predominance of C-H···O and C-H···π interactions, with dispersion energy playing a critical role in stabilizing the crystal lattice. Docking studies against key microbial targets, particularly E. coli FabH, demonstrated superior binding energies, suggesting significant antimicrobial potential. The comprehensive suite of structural and computational analyses underscores the potential of the synthesized 1,2,4-oxadiazole derivative, which may be one of the promising candidates for antimicrobial drug development. Future in vitro, in vivo studies will be supportive in optimizing the derivative for enhanced efficacy and further elucidating its pharmacological mechanisms, paving the way for potential clinical applications. This study not only provides insights into the structural and functional properties of a novel 1,2,4-oxadiazole derivative but also highlights its promising role in antimicrobial drug discovery.

PMID:39959081 | PMC:PMC11822514 | DOI:10.1021/acsomega.4c06520

Liver Res. 2024 Sep 16;8(3):131-140. doi: 10.1016/j.livres.2024.09.003. eCollection 2024 Sep.

ABSTRACT

Compared with the widely used rodents, pigs are anatomically, physiologically, and genetically more similar to humans, making them high-quality models for the study of liver diseases. Here, we review the latest research progress on pigs as a model of human liver disease, including methods for establishing them and their advantages in studying cystic fibrosis liver disease, acute liver failure, liver regeneration, non-alcoholic fatty liver disease, liver tumors, and xenotransplantation. We also emphasize the importance of genetic engineering techniques, mainly the CRISPR/Cas9 system, which has greatly enhanced the utility of porcine models as a tool for substantially advancing liver disease research. Genetic engineering is expected to propel the pig as one of the irreplaceable animal models for future biomedical research.

PMID:39957748 | PMC:PMC11771255 | DOI:10.1016/j.livres.2024.09.003

Data Brief. 2025 Jan 27;59:111328. doi: 10.1016/j.dib.2025.111328. eCollection 2025 Apr.

ABSTRACT

Leaf dimensioning is relevant for analyzing plant responses to several conditions such as soil fertility, availability of light, agricultural pesticide effect, and access to water in the soil or periods of drought. In this paper, we present a dataset composed of 6981 images of 612 common bean leaves (Phaseolus vulgaris). We captured the images of each leaf accompanied by a fiducial marker and annotated the known leaf dimensions (area, perimeter, length, and width). We provide annotations concerning image segmentation, known area uniformly distributed over the leaf region, real area of the marker region, marker pose, capture conditions, and camera calibration. This dataset can be useful for developing deep learning algorithms for leaf dimensioning and related problems. Therefore, there is a potential to contribute to computer vision and plant physiology researchers and specialists.

PMID:39959655 | PMC:PMC11830349 | DOI:10.1016/j.dib.2025.111328

Middle East Afr J Ophthalmol. 2024 Dec 2;30(4):189-202. doi: 10.4103/meajo.meajo_154_24. eCollection 2023 Oct-Dec.

ABSTRACT

Numerous artificial intelligence (AI) models, including deep learning techniques, are being developed to segment choroids in optical coherence tomography (OCT) images. However, there is a need for consensus on which specific models to use, requiring further synthesis of their efficacy and role in choroid visualization in myopic patients. A systematic literature search was conducted on three main databases (PubMed, Web of Science, and Scopus) using the search terms: "Machine learning" OR "Artificial Intelligence" OR "Deep learning" AND "Myopia" AND "Choroid" OR "Choroidal" from inception to February 2024 removing duplicates. A total of 12 studies were included. The populations included myopic patients with varying degrees of myopia. The AI models applied were primarily deep learning models, including U-Net with a bidirectional Convolutional Long Short-Term Memory module, LASSO regression, Attention-based Dense U-Net network, ResNeSt101 architecture training five models, and Mask Region-Based Convolutional Neural Network. The reviewed AI models demonstrated high diagnostic accuracy, including sensitivity, specificity, and area under the curve values, in identifying and assessing myopia-related changes. Various biomarkers were assessed, such as choroidal thickness, choroidal vascularity index, choroidal vessel volume, luminal volume, and stromal volume, providing valuable insights into the structural and vascular changes associated with the condition. The integration of AI models in ophthalmological imaging represents a significant advancement in the diagnosis and management of myopia. The high diagnostic accuracy and efficiency of these models underscore their potential to revolutionize myopia care, improving patient outcomes through early detection and precise monitoring of disease progression. Future studies should focus on standardizing AI methodologies and expanding their application to broader clinical settings to fully realize their potential in ophthalmology.

PMID:39959595 | PMC:PMC11823532 | DOI:10.4103/meajo.meajo_154_24

IEEE Sens J. 2024 Jun 15;24(12):19590-19600. doi: 10.1109/jsen.2024.3396052. Epub 2024 May 7.

ABSTRACT

Nocturnal blood pressure (BP) monitoring offers valuable insights into various aspects of human wellbeing, particularly cardiovascular health. Despite recent advancements in medical technology, there remains a pressing need for a non-invasive, cuffless, and less burdensome method for overnight BP measurements. A range of machine learning models have been developed to estimate daytime BP using photoplethysmography (PPG), a readily available sensor embedded in modern wearable devices. However, investigations into nocturnal BP estimation, especially concerning long-term data patterns during sleep, are still lacking. This paper investigates the estimation of nocturnal BP from overnight PPG signals collected in a clinical-grade sleep laboratory setting. To address this, we propose SleepBP-Net, a lightweight time-distributed convolutional recurrent network. This novel model leverages long-term patterns within PPG waveforms to estimate systolic and diastolic BP (SBP and DBP), considering Portapres BP measurements as a reference. Our experiments, based on leave-one-subject-out validation on 1-minute sequences of PPG, resulted in a mean absolute error (MAE) of 15.7 mmHg (SBP) and 12.1 mmHg (DBP). Model personalization improved the results to 7.8 mmHg (SBP) and 5.9 mmHg (DBP). Further enhancements were observed when extending the sequence length to 30 minutes, resulting in MAE values of 7.2 mmHg (SBP) and 5.7 mmHg (DBP). These findings underscore the significance of learning long-term temporal patterns from sleep PPG data. Additionally, we demonstrate the superiority of hybrid convolutional recurrent networks over their convolutional network counterparts. Based on our results, SleepBP-Net holds promise for unobtrusive real-world nocturnal BP estimation, particularly in scenarios where computational efficiency is crucial.

PMID:39959563 | PMC:PMC11824277 | DOI:10.1109/jsen.2024.3396052

Transl Anim Sci. 2025 Feb 5;9:txae168. doi: 10.1093/tas/txae168. eCollection 2025.

ABSTRACT

The aim of the study was to deploy computer vision models for real-time detection of digital dermatitis (DD) lesions in cows using Android or iOS mobile applications. Early detection of DD lesions in dairy cows is crucial for prompt treatment and animal welfare. Android and iOS apps could facilitate routine and early DD detection in cows' feet on dairy and beef farms. Upon detecting signs of DD, dairy farmers could implement preventive and treatment methods, including foot baths, topical treatment, hoof trimming, or quarantining cows affected by DD to prevent its spread. We applied transfer-learning to DD image data for 5 lesion classes, M0, M4H, M2, M2P, and M4P, on pretrained YOLOv5 model architecture using COCO-128 pretrained weights. The combination of localization loss, classification loss, and objectness loss was used for the optimization of prediction performance. The custom DD detection model was trained on 363 images of size 416 × 416 pixels and tested on 46 images. During model training, data were augmented to increase model robustness in different environments. The model was converted into TFLite format for Android devices and CoreML format for iOS devices. Techniques such as quantization were implemented to improve inference speed in real-world settings. The DD models achieved a mean average precision (mAP) of 0.95 on the test dataset. When tested in real-time, iOS devices resulted in Cohen's kappa value of 0.57 (95% CI: 0.49 to 0.65) averaged across the 5 lesion classes denoting the moderate agreement of the model detection with human investigators. The Android device resulted in a Cohen's kappa value of 0.38 (95% CI: 0.29 to 0.47) denoting fair agreement between model and investigator. Combining M2 and M2P classes and M4H and M4P classes resulted in a Cohen's kappa value of 0.65 (95% CI: 0.54 to 0.76) and 0.46 (95% CI: 0.35 to 0.57), for Android and iOS devices, respectively. For the 2-class model (lesion vs. non-lesion), a Cohen's kappa value of 0.74 (95% CI: 0.63 to 0.85) and 0.65 (95% CI: 0.52 to 0.78) was achieved for iOS and Android devices, respectively. iOS achieved a good inference time of 20 ms, compared to 57 ms on Android. Additionally, we deployed models on Ultralytics iOS and Android apps giving kappa scores of 0.56 (95% CI: 0.48 to 0.64) and 0.46 (95% CI: 0.37 to 0.55), respectively. Our custom iOS app surpassed the Ultralytics apps in terms of kappa score and confidence score.

PMID:39959562 | PMC:PMC11829201 | DOI:10.1093/tas/txae168

Heliyon. 2025 Jan 25;11(3):e42257. doi: 10.1016/j.heliyon.2025.e42257. eCollection 2025 Feb 15.

ABSTRACT

Congenital heart disease (CHD), impacting around 1 % of infants worldwide, constitutes a significant healthcare challenge. Early detection is crucial, however constrained by the intricacies of conventional diagnostic techniques such as auscultation and echocardiography. This research presents a tailored one-dimensional convolutional neural network (1D-CNN) for the classification of phonocardiogram (PCG) signals into normal or abnormal categories, providing an automated and efficient solution for congenital heart disease (CHD) diagnosis. The model was trained on a composite dataset consisting of local pediatric PCG signals and publicly accessible dataset. Preprocessing methods, such as low- and high-pass filtering (60-650 Hz), resampling, and noise reduction, were utilized to enhance signal quality. Data augmentation techniques, including chunking, padding, and pitch-shifting, were employed to rectify dataset imbalances and improve model efficacy. Experimental results indicate substantial enhancements, attaining an accuracy of 98.56 %, precision of 98.56 %, F1 score of 98.55 %, sensitivity of 0.98, and specificity of 0.99. The comparative analysis demonstrates the proposed approach's superiority over current methods regarding accuracy and reliability. The research highlights the promise of combining modern signal processing with deep learning for efficient CHD screening. The suggested model exhibits outstanding performance yet, issues like dataset variability and noise persist. Future endeavors involve extending to multiclass categorization and assessing performance across a wider range of medical problems. This study represents a significant advancement in accessible, automated CHD diagnoses, enhancing clinical competence to elevate pediatric treatment.

PMID:39959496 | PMC:PMC11830292 | DOI:10.1016/j.heliyon.2025.e42257