BMC Biol. 2025 Aug 4;23(1):236. doi: 10.1186/s12915-025-02359-9.

ABSTRACT

BACKGROUND: Prediction of protein-protein interactions (PPIs) is fundamental for identifying drug targets and understanding cellular processes. The rapid growth of PPI studies necessitates the development of efficient and accurate tools for automated prediction of PPIs. In recent years, several robust deep learning models have been developed for PPI prediction and have found widespread application in proteomics research. Despite these advancements, current computational tools still face limitations in modeling both the pairwise interactions and the hierarchical relationships between proteins.

RESULTS: We present HI-PPI, a novel deep learning method that integrates hierarchical representation of PPI network and interaction-specific learning for protein-protein interaction prediction. HI-PPI extracts the hierarchical information by embedding structural and relational information into hyperbolic space. A gated interaction network is then employed to extract pairwise features for interaction prediction. Experiments on multiple benchmark datasets demonstrate that HI-PPI outperforms the state-of-the-art methods; HI-PPI improves Micro-F1 scores by 2.62%-7.09% over the second-best method. Moreover, HI-PPI offers explicit interpretability of the hierarchical organization within the PPI network. The distance between the origin and the hyperbolic embedding computed by HI-PPI naturally reflects the hierarchical level of proteins.

CONCLUSIONS: Overall, the proposed HI-PPI effectively addresses the limitations of existing PPI prediction methods. By leveraging the hierarchical structure of PPI network, HI-PPI significantly enhances the accuracy and robustness of PPI predictions.

PMID:40754535 | DOI:10.1186/s12915-025-02359-9

Enferm Infecc Microbiol Clin (Engl Ed). 2025 Aug-Sep;43(7):402-410. doi: 10.1016/j.eimce.2025.06.004.

ABSTRACT

OBJECTIVE: Glucocorticoids are vital in treating COVID-19, but standard dosage for noncritical patients remain controversial. To determine the optimal glucocorticoid dosage for noncritical COVID-19 patients, we analyzed factors influencing dosage and developed a predictive model.

METHODS: We retrospectively analyzed 273 noncritical COVID-19 pneumonia patients underwent pulmonary CT and treated with glucocorticoids in a tertiary hospital (12/2022-01/2023). Patients were divided into low and high glucocorticoid dosage groups based on a daily 40mg methylprednisolone or equivalent. Artificial intelligence (AI)-based deep learning was utilized to assess pulmonary CT images for accurate lesion area, which then analyzed through multivariable logistic regression to explore their correlation with glucocorticoid dosage. A predictive model was developed and validated for dosage prediction.

RESULTS: The primary analysis included 243 patients, with 168 in the training set and 75 in the validation set. High-dose treatment was administered to 139 patients (82.7%) and low-dose to 29 patients (17.3%) in the training cohort. A predictive model incorporating normally inflated ratio, ground-glass opacity (GGO) ratio, and consolidation ratio accurately predicted selection of high- or low-dose, in both training (AUC=0.803) and validation cohorts (AUC=0.836), respectively. In 30 patients with post-CT adjusted dosages, the predicted dosages highly matched with the actual adjusted dosages.

CONCLUSION: Glucocorticoid dosages for noncritical COVID-19 pneumonia treatment are influenced by pulmonary CT features. Our predictive model can predict glucocorticoid dosage, however, should be validated by larger, prospective studies.

PMID:40754353 | DOI:10.1016/j.eimce.2025.06.004

Ann Oncol. 2025 Aug 1:S0923-7534(25)00910-X. doi: 10.1016/j.annonc.2025.07.013. Online ahead of print.

NO ABSTRACT

PMID:40754034 | DOI:10.1016/j.annonc.2025.07.013

Neuropsychologia. 2025 Aug 2:109242. doi: 10.1016/j.neuropsychologia.2025.109242. Online ahead of print.

ABSTRACT

Action suppresses the neural responses to its sensory feedback. The phenomenon, termed action-induced suppression, highlights the predictive processes in sensorimotor integration but remains controversial regarding the underlying mechanisms. The predictive coding framework posits that action-induced suppression is a general, non-action-specific process driven by predictions. In contrast, the Dual-Stream Prediction Model (DSPM) argues that motor-based and memory-based predictions are mediated by distinct processes - motor predictions rely on precise action-perception mappings and temporal synchrony, whereas memory predictions are based on learned associations. To test these competing theories, we compared auditory ERP responses elicited by self-initiated keypresses (motor-based) and visually cued auditory events (memory-based) in a matching judgment task. Results revealed significant suppression at the P2 component, when the prediction matched the auditory feedback only in the motor-auditory task but not in the visual-auditory task. The findings qualitatively replicated common observations of action-induced suppression; the suppression effects are at a later component rather than N1, indicating the interaction between prediction and perception at a higher level, such as syllable categorization in the current experimental design. Surprisingly, we observed N1 enhancement to the auditory probe in both conditions, with greater enhancement in the motor-auditory task compared to the visual-auditory task. The enhancement effects likely reflect a prediction-induced attentional-like modulation at an early auditory processing stage, potentially driven by the demands of the matching judgment task. Together, these findings support the DSPM by demonstrating functional dissociable mechanisms of motor-based and memory-based predictions.

PMID:40754023 | DOI:10.1016/j.neuropsychologia.2025.109242

Comput Biol Med. 2025 Aug 2;196(Pt B):110860. doi: 10.1016/j.compbiomed.2025.110860. Online ahead of print.

ABSTRACT

Leukemia is a hematologic tumor that proliferates in bone marrow and seriously affects the survival of patients. Early and accurate diagnosis is crucial for effective leukemia treatment. Traditional diagnostic methods rely on experts' subjective analysis of bone marrow smears microscopic images. This approach is time-consuming and complex. Despite recent advances in deep learning, automated leukemia detection remains limited due to the scarcity of high-quality datasets, the prevailing focus on single-cell image classification rather than precise cell-level detection in whole slide images, along with challenges such as morphological heterogeneity, uneven staining, scale variation, and occluded cell boundary in bone marrow smears. To address these challenges, we construct a novel dataset comprising 1794 high-quality microscopic images, establishing a new benchmark for lymphocytic leukemia detection. Additionally, we develop a fully automated diagnostic method based on spatially-guided learning (SGLNet), enabling rapid whole slide analysis of leukemia. Specifically, we introduce several innovative enhancements to the baseline algorithm, including the spatially-guided learning framework, scale-aware fusion module, small object-enhancing mechanisms, and efficient intersection over union loss function. These improvements effectively address the impact of morphological similarity and complex backgrounds in leukemia detection, significantly enhancing detection accuracy. Finally, the results show that SGLNet achieves mean average precision scores of 95.9 % and 98.6 % in detecting acute lymphoblastic leukemia and chronic lymphocytic leukemia, respectively. These results demonstrate the efficiency and accuracy of our method in identifying lymphoblastic leukemia cells, significantly enhancing large-scale clinical diagnosis, and supporting clinicians in developing personalized treatment plans.

PMID:40753948 | DOI:10.1016/j.compbiomed.2025.110860

Adv Gerontol. 2025;38(2):171-180.

ABSTRACT

Osteoporosis of the jawbones is a significant concern in dental practice, particularly for implant treatment planning. This review summarizes current diagnostic approaches with a focus on the use of artificial intelligence (AI) algorithms, including convolutional neural networks, for analyzing panoramic radiographs and cone-beam computed tomography. The findings demonstrate that AI models achieve high diagnostic accuracy in the automated classification of radiographic images, comparable to dual-energy X-ray absorptiometry. AI reduces subjectivity in image interpretation, although further standardization, dataset expansion, and development of explainable models are necessary. The review highlights comparative metrics of various neural network architectures and their potential for integration into clinical workflows.

PMID:40753551

Brief Bioinform. 2025 Jul 2;26(4):bbaf392. doi: 10.1093/bib/bbaf392.

ABSTRACT

P-glycoprotein (P-gp), a key member of the ATP-binding cassette (ABC) transporter family, plays a significant role in drug absorption and distribution by binding to diverse xenobiotics and actively transporting them out of cells. Given P-gp's widespread expression, including its critical presence at the blood-brain barrier, identifying whether a compound functions as a P-gp substrate or inhibitor is essential in drug development to evaluate its ability to penetrate the central nervous system. However, most studies on P-gp focus on inhibitor models rather than substrate models. This study presents a robust graph neural network approach to predict P-gp substrates, leveraging graph convolutional networks, AttentiveFP, and an ensemble model. Using a dataset of 1995 drug molecules (1202 substrates, 793 nonsubstrates), AttentiveFP outperformed traditional methods, achieving an ROC-AUC of 0.848 and an accuracy of 0.815. Integrated gradient analysis identified 20 key substructures associated with P-gp substrates. Most noteworthy is that the top four conferring a >70% probability of substrate classification which can be used a quick assessment in the future. This interpretable framework enhances P-gp prediction and broader drug development efforts.

PMID:40753539 | DOI:10.1093/bib/bbaf392

BMC Oral Health. 2025 Aug 2;25(1):1292. doi: 10.1186/s12903-025-06576-0.

ABSTRACT

BACKGROUND: Accurate restoration and reconstruction of tooth morphology are crucial in restorative dentistry, implantology, and forensic odontology. Traditional methods, like manual wax modeling and template-based computer-aided design (CAD), struggle with accuracy, personalization, and efficiency. To address the challenge, we propose an innovative and efficient deep learning-based framework designed for the automatic restoration and reconstruction of tooth morphology.

METHODS: The proposed method contains three stages. Firstly, an RGB image of a defective tooth is inputted into the restoration network, which fills in the missing regions to produce a complete RGB image of the tooth. The resulting image is then converted to a grayscale image in the preprocessing stage to ensure compatibility with the subsequent reconstruction process. Finally, the 3D reconstruction network utilizes the grayscale image to generate a detailed 3D mesh model of the tooth.

RESULTS: The experimental results demonstrate that the proposed method achieves superior performance in restoration quality, reconstruction accuracy, generalization, and inference speed, with an average time of 12 s per image. Notably, compared to the original Pixel2Mesh, the improved ResNet50-based Pixel2Mesh enhances the average F-Score, CD, and EMD for reconstructed tooth models by 26.5%, 34.7%, and 22.3%, respectively.

CONCLUSIONS: The approach proposed in this paper offers a promising solution for personalized intelligent, and efficient tooth restoration and reconstruction, providing a valuable tool for dental diagnostics and treatment planning.

PMID:40753409 | DOI:10.1186/s12903-025-06576-0

Sci Rep. 2025 Aug 2;15(1):28283. doi: 10.1038/s41598-025-13755-0.

ABSTRACT

Early detection of lung cancer, which remains one of the leading causes of death worldwide, is important for improved prognosis, and CT scanning is an important diagnostic modality. Lung cancer classification according to CT scan is challenging since the disease is characterized by very variable features. A hybrid deep architecture, ILN-TL-DM, is presented in this paper for precise classification of lung cancer from CT scan images. Initially, an Adaptive Gaussian filtering method is applied during pre-processing to eliminate noise and enhance the quality of the CT image. This is followed by an Improved Attention-based ResU-Net (P-ResU-Net) model being utilized during the segmentation process to accurately isolate the lung and tumor areas from the remaining image. During the process of feature extraction, various features are derived from the segmented images, such as Local Gabor Transitional Pattern (LGTrP), Pyramid of Histograms of Oriented Gradients (PHOG), deep features and improved entropy-based features, all intended to improve the representation of the tumor areas. Finally, classification exploits a hybrid deep learning architecture integrating an improved LeNet structure with Transfer Learning (ILN-TL) and a DeepMaxout (DM) structure. Both model outputs are finally merged with the help of a soft voting strategy, which results in the final classification result that separates cancerous and non-cancerous tissues. The strategy greatly enhances lung cancer detection's accuracy and strength, showcasing how combining sophisticated neural network structures with feature engineering and ensemble methods could be used to achieve better medical image classification. The ILN-TL-DM model consistently outperforms the conventional methods with greater accuracy (0.962), specificity (0.955) and NPV (0.964).

PMID:40753351 | DOI:10.1038/s41598-025-13755-0

NPJ Precis Oncol. 2025 Aug 2;9(1):271. doi: 10.1038/s41698-025-01056-8.

ABSTRACT

Although immunotherapy combined with chemotherapy (ICT) is the standard treatment for advanced non-small cell lung cancer (NSCLC), identification of reliable prognostic biomarkers remains challenging. In this multicenter study, we performed next-generation sequencing of tumor samples from 162 patients receiving first-line ICT at the Chinese PLA General Hospital and collected their pathological image information. First, we established a model to predict the risk of tumor progression based on genomic characteristics. Furthermore, a deep learning method was employed to recognize different cell types from pathological images, which significantly improved the accuracy of progression-free survival (PFS) and overall survival (OS) prediction. In summary, we constructed a Prognostic Multimodal Classifier for Progression (PMCP) that possesses the capability to precisely forecast PFS and OS. Patients with the PMCP1 subtype exhibit a low risk of progression and demonstrate a higher proportion of epithelial cells. PMCP highlighted the potential value of multimodal biomarkers in guiding clinical decisions regarding ICT. The area under curve (AUC) for predicting PFS was 0.807. This study revealed the importance of integrating genomic and pathological data to improve prognostic accuracy and enable personalized treatment for patients with advanced NSCLC.

PMID:40753345 | DOI:10.1038/s41698-025-01056-8

Eur Radiol. 2025 Aug 3. doi: 10.1007/s00330-025-11865-x. Online ahead of print.

ABSTRACT

OBJECTIVES: This study aims to evaluate the diagnostic accuracy of an open-source deep learning (DL) model for detecting clinically significant prostate cancer (csPCa) in biparametric MRI (bpMRI). It also aims to outline the necessary components of the model that facilitate effective sharing and external evaluation of PCa detection models.

MATERIALS AND METHODS: This retrospective diagnostic accuracy study evaluated a publicly available DL model trained to detect PCa on bpMRI. External validation was performed on bpMRI exams from 151 biologically male patients (mean age, 65 ± 8 years). The model's performance was evaluated using patient-level classification of PCa with both radiologist interpretation and histopathology serving as the ground truth. The model processed bpMRI inputs to generate lesion probability maps. Performance was assessed using the area under the receiver operating characteristic curve (AUC) for PI-RADS ≥ 3, PI-RADS ≥ 4, and csPCa (defined as Gleason ≥ 7) at an exam level.

RESULTS: The model achieved AUCs of 0.86 (95% CI: 0.80-0.92) and 0.91 (95% CI: 0.85-0.96) for predicting PI-RADS ≥ 3 and ≥ 4 exams, respectively, and 0.78 (95% CI: 0.71-0.86) for csPCa. Sensitivity and specificity for csPCa were 0.87 and 0.53, respectively. Fleiss' kappa for inter-reader agreement was 0.51.

CONCLUSION: The open-source DL model offers high sensitivity to clinically significant prostate cancer. The study underscores the importance of sharing model code and weights to enable effective external validation and further research.

KEY POINTS: Question Inter-reader variability hinders the consistent and accurate detection of clinically significant prostate cancer in MRI. Findings An open-source deep learning model demonstrated reproducible diagnostic accuracy, achieving AUCs of 0.86 for PI-RADS ≥ 3 and 0.78 for CsPCa lesions. Clinical relevance The model's high sensitivity for MRI-positive lesions (PI-RADS ≥ 3) may provide support for radiologists. Its open-source deployment facilitates further development and evaluation across diverse clinical settings, maximizing its potential utility.

PMID:40753327 | DOI:10.1007/s00330-025-11865-x

Sci Rep. 2025 Aug 2;15(1):28241. doi: 10.1038/s41598-025-99124-3.

ABSTRACT

Numerous factors contribute to student success in educational settings, with academic support and learning strategies identified as key influences. Existing research highlights that various academic assistance and individual learning approaches shape student success. Although different methods are available for predicting academic achievement, the application of fuzzy logic in this context remains relatively underexplored. This study seeks to address this gap by employing a fuzzy logic model to forecast the exam performance of prospective physical education and sports teachers. Specifically, the study examines how the learning approaches and perceived levels of academic support among these candidates influence their exam outcomes. The results indicate that fuzzy logic provides a viable tool for predicting student success, demonstrating its potential as an alternative to traditional predictive methods. Furthermore, the findings suggest that students' learning approaches and perceptions of academic support play a significant role in shaping their academic achievements. By integrating fuzzy logic into educational research, this study contributes to a broader understanding of how non-linear and complex interactions between academic support and learning strategies impact student performance. These results highlight the practical potential of fuzzy logic models in identifying at-risk students and designing targeted interventions to improve academic outcomes. Educators and institutions can create more effective and personalised learning environments by fostering deep learning approaches and enhancing academic support systems.

PMID:40753281 | DOI:10.1038/s41598-025-99124-3

BMC Urol. 2025 Aug 2;25(1):190. doi: 10.1186/s12894-025-01875-8.

ABSTRACT

PURPOSE: In study, we aimed to dosimetrically evaluate the usability of a new generation autocontouring algorithm (DirectORGANS) that automatically identifies organs and contours them directly in the computed tomography (CT) simulator before creating prostate radiotherapy plans.

METHODS: The CT images of 10 patients were used in this study. The prostates, bladder, rectum, and femoral heads of 10 patients were automatically contoured based on DirectORGANS algorithm at the CT simulator. On the same CT image sets, the same target volumes and contours of organs at risk were manually contoured by an experienced physician using MRI images and used as a reference structure. The doses of manually delineated contours of the target volume and organs at risk and the doses of auto contours of the target volume and organs at risk were obtained from the dose volume histogram of the same plan. Conformity index (CI) and homogeneity index (HI) were calculated to evaluate the target volumes. In critical organ structures, V60, V65, V70 for the rectum, V65, V70, V75, and V80 for the bladder, and maximum doses for femoral heads were evaluated. The Mann-Whitney U test was used for statistical comparison with statistical package SPSS (P < 0.05).

RESULTS: Compared to the doses of the manual contours (MC) with auto contours (AC), there was no significant difference between the doses of the organs at risk. However, there were statistically significant differences between HI and CI values due to differences in prostate contouring (P < 0.05).

CONCLUSION: The study showed that the need for clinicians to edit target volumes using MRI before treatment planning. However, it demonstrated that delineating organs at risk was used safely without the need for correction. DirectORGANS algorithm is suitable for use in RT planning to minimize differences between physicians and shorten the duration of this contouring step.

PMID:40753235 | DOI:10.1186/s12894-025-01875-8

Eur Urol Focus. 2025 Aug 1:S2405-4569(25)00216-0. doi: 10.1016/j.euf.2025.07.011. Online ahead of print.

ABSTRACT

Bladder cancer (BC) ranks among the tenth most common cancers globally, and its management remains a significant challenge for both patients and clinicians in terms of care delivery and decision-making process. The integration of artificial intelligence (AI) tools-primarily machine learning and deep learning methods-into the current BC workflow offers an opportunity for a more personalized approach to treatment. This article provides a brief overview of AI applications across different steps of BC management (ie, detection, grading, staging, risk stratification, treatment, and outcome prediction), highlighting its potential to contribute to individualized management strategies. Despite significant advances, major barriers still impede broad applications of AI in BC clinical workflows. Overcoming these obstacles is critical to realize the full potential of AI-driven personalization of BC care in the coming decade. PATIENT SUMMARY: Our mini review summarizes how artificial intelligence (ie, a machine's ability to mimic human intelligence to perform tasks involving decision-making and problem-solving) has been applied to the management of bladder cancer, and whether it could lead to more precise treatment for patients diagnosed with this disease. Although several promising applications have been developed, more studies are necessary before these can be used in routine clinical practice.

PMID:40753031 | DOI:10.1016/j.euf.2025.07.011

Anesthesiol Clin. 2025 Sep;43(3):507-525. doi: 10.1016/j.anclin.2025.05.005. Epub 2025 Jul 5.

ABSTRACT

The article explores the transformative impact of artificial intelligence (AI) in critical care medicine. AI models offer superior accuracy in mortality risk assessment and personalized treatment recommendations, enhancing patient outcomes in acute and complex settings. However, various challenges exist, such as data quality, model interpretability, integrating models to clinical workflow, and ethical and legal concerns need to be addressed for successful integration into clinical practice. The article emphasizes the demand for ongoing research on data quality and standardization, building model transparency, developing robust validation methods, and ethical oversight to fully leverage AI's potential in improving critical care.

PMID:40752950 | DOI:10.1016/j.anclin.2025.05.005

Anesthesiol Clin. 2025 Sep;43(3):471-489. doi: 10.1016/j.anclin.2025.05.003. Epub 2025 Jul 3.

ABSTRACT

Recent breakthroughs in artificial intelligence (AI) have particularly shone in cardiothoracic anesthesia, where its ability to efficiently analyze complex datasets and process vast amounts of information in mere moments has captured considerable attention. For cardiothoracic anesthesiologists, the challenge of swiftly evaluating myriad variables is paramount to minimizing complications and optimizing patient outcomes. This article explores the current state of AI in cardiac anesthesia, illuminating the compelling evidence supporting its use and charting potential future paths for researchers and clinicians alike. We will also delve into the challenges in this dynamic field and propose inventive concepts for seamlessly integrating AI into future research initiatives.

PMID:40752948 | DOI:10.1016/j.anclin.2025.05.003

Anesthesiol Clin. 2025 Sep;43(3):405-418. doi: 10.1016/j.anclin.2025.04.003. Epub 2025 May 26.

ABSTRACT

The term "artificial intelligence" (AI) refers to the intelligence demonstrated by machines. It is generally defined as creating machines that mimic cognitive functions typically associated with the human mind. This article will define and explore the various levels of AI and examine its historical development. We will also analyze the attention AI has received from different countries. Next, we will investigate the fundamental technologies underpinning AI. After that, we will discuss how AI transforms health care. Finally, we will consider whether AI has the potential to replace physicians and explore its future in the health care sector.

PMID:40752944 | DOI:10.1016/j.anclin.2025.04.003

Surv Ophthalmol. 2025 Jul 31:S0039-6257(25)00128-6. doi: 10.1016/j.survophthal.2025.07.011. Online ahead of print.

ABSTRACT

Less than a decade has passed since deep learning (DL) was first applied in ophthalmology. With tremendous growth in this field since then, DL is expected to transform and enhance the efficiency of traditional ophthalmology practice. Central serous chorioretinopathy (CSC) is a common chorioretinal disorder whose etiopathogenesis remains largely unknown. The diagnosis and management of CSC rely heavily on multimodal imaging data, detailed analysis of which may exceed the capacity of many practices. In this comprehensive review, we examine how DL can address such issues through automated analysis of CSC-related imaging biomarkers, including subretinal fluid, pigment epithelial detachment, subretinal hyperreflective material, hyperreflective foci, retinal pigment epithelium atrophy, ellipsoid zone loss, and choroidal layer, sublayers, vessels, and neovascularization. Their prognostic yield and therapeutic implications are covered as well. We describe how DL enables rapid, noninvasive visualization of choroidal vasculature, a primary source of pathology in CSC, in unprecedented detail. We also review the state-of-the-art DL models designed for automated CSC diagnosis, classification, prognostication, and treatment outcome prediction based on imaging data. We highlight the challenges and gaps in this field, discuss some recommended counter measures, and suggest future research directions.

PMID:40752852 | DOI:10.1016/j.survophthal.2025.07.011

Int J Dev Neurosci. 2025 Aug;85(5):e70034. doi: 10.1002/jdn.70034.

ABSTRACT

BACKGROUND AND OBJECTIVE: Autism spectrum disorder (ASD) is a complex neurological condition that impacts cognitive, social and behavioural abilities. Early and accurate diagnosis is crucial for effective intervention and treatment. Traditional diagnostic methods lack accuracy, efficient feature selection and computational efficiency. This study proposes an integrated approach that combines the electromagnetic interaction algorithm (EIA) for feature selection with the adaptive kernel attention network (AKAttNet) for classification, aiming to improve ASD detection performance across multiple datasets.

METHODS: The proposed methodology consists of two core components: (1) EIA, which optimises feature selection by identifying the most relevant attributes for ASD classification, and (2) AKAttNet, a deep learning model leveraging adaptive kernel attention mechanisms to enhance classification accuracy. The framework is evaluated using four publicly available ASD datasets. The classification performance of AKAttNet is compared against traditional machine learning methods, including logistic regression (LR), support vector machine (SVM) and random forest (RF), as well as competing deep learning models. Statistical evaluation includes precision, recall (sensitivity), specificity and overall accuracy metrics.

RESULTS: The proposed model outperforms conventional machine learning and deep learning approaches, demonstrating higher classification accuracy and robustness across multiple datasets. AKAttNet, combined with EIA-based feature selection, achieves an accuracy improvement ranging from 0.901 to 0.9827, Cohen's kappa values between 0.7789 and 0.9685 and Jaccard similarity scores from 0.8041 to 0.9709 across four different datasets. Comparative analysis highlights the efficiency of the EIA algorithm in reducing feature dimensionality while maintaining high model performance. Additionally, the proposed method exhibits lower computational time and enhanced generalizability, making it a promising approach for ASD detection.

CONCLUSIONS: This study presents a practical ASD detection framework integrating EIA for feature selection with AKAttNet for classification. The results indicate that this hybrid approach enhances diagnostic accuracy while reducing computational overhead, making it a promising tool for early ASD diagnosis. The findings support the potential of deep learning and optimisation techniques in developing more efficient and reliable ASD screening systems. Future work can explore real-world clinical applications and further refinement of the feature selection process.

PMID:40751377 | DOI:10.1002/jdn.70034

J Oral Rehabil. 2025 Aug 1. doi: 10.1111/joor.70030. Online ahead of print.

ABSTRACT

BACKGROUND: The Test of Masticating and Swallowing Solids (TOMASS) is a validated screening tool for assessing masticatory and swallowing functions. However, the conventional TOMASS relies on operator-dependent methods, which limit its objectivity and efficiency. The neck-worn electronic stethoscope (NWES), a contact sensor positioned on the back of the neck, has recently been developed to automatically detect and monitor swallowing actions through deep learning-based analysis of collected sound data.

OBJECTIVE: This study piloted a semi-automated assessment approach using a NWES to objectively measure TOMASS parameters and examine the influence of age and gender.

METHODS: A total of 123 healthy adults (mean age: 58.7 ± 18.5 years) consumed two crackers while audio data recorded using a NWES and visual data were collected by smartphone. Measurements included discrete bite count, swallow count, oral processing and swallowing time (OPST), and first OPST (1st-OPST). Statistical analyses were conducted to assess gender- and age-related changes and differences.

RESULTS: The NWES enabled objective and precise TOMASS measurements. Age-related prolongation of OPST and 1st-OPST was observed, particularly in men (p < 0.001). Women exhibited fewer age-related changes in OPST, although swallow count tended to decrease with age (p < 0.001). Regarding gender differences, younger women demonstrated higher bite (2.3 [interquartile range (IQR): 1.0-3.0] vs. 1 [IQR: 1.0-2.0], p = 0.042) and swallow counts (2.5 [IQR: 2.0-2.5] vs. 2 [IQR: 1.0-2.0], p = 0.026) compared with men.

CONCLUSION: The NWES appeared suitable as an objective, efficient tool for automated TOMASS evaluation. Age-related changes in masticatory and swallowing performance differed according to gender, highlighting the need for tailored assessments. Future research on NWES-based TOMASS measurements should include diverse populations and extension to dysphagia and masticatory dysfunction.

PMID:40751301 | DOI:10.1111/joor.70030