Am J Respir Crit Care Med. 2025 Apr 10. doi: 10.1164/rccm.202411-2231OC. Online ahead of print.

ABSTRACT

RATIONALE: Clinical and real-world studies show elexacaftor/tezacaftor/ivacaftor (ELX/TEZ/IVA) is efficacious and safe in people with cystic fibrosis (CF) ≥12 years of age with at least one F508del allele.

OBJECTIVES: Given the potential for life-long ELX/TEZ/IVA use, long-term safety and efficacy of ELX/TEZ/IVA was assessed.

METHODS: In this phase 3, open-label, single-arm extension study, participants with F508del-minimal function genotypes (from 24-week parent study 445-102 [n = 399]) or with the F508del-F508del genotype (from 4-week parent study 445-103 [n = 107]) received ELX/TEZ/IVA (ELX 200 mg once daily, TEZ 100 mg once daily, and IVA 150 mg every 12 hours) over 192 weeks.

MEASUREMENTS AND MAIN RESULTS: Primary endpoint was safety and tolerability. Mean exposure to ELX/TEZ/IVA was 172.6 weeks. Most participants had adverse events classified as mild (12.8%) or moderate (60.7%) in severity. Eighteen participants (3.6%) had adverse events that led to treatment discontinuation. After starting ELX/TEZ/IVA, participants had consistent increases in percent predicted FEV1 (ppFEV1), Cystic Fibrosis Questionnaire-Revised respiratory domain score, and body mass index, with decreases in sweat chloride concentration and pulmonary exacerbations rates; these improvements were maintained through 192 weeks. The mean annualized rate of change in ppFEV1 was 0.02 percentage points (95% CI, -0.14 to 0.19) after initiation of ELX/TEZ/IVA.

CONCLUSIONS: During this 192-week open label extension study, the longest clinical study of a CFTR modulator to date, ELX/TEZ/IVA remained generally safe and well-tolerated. Participants had sustained improvements in lung function, respiratory symptoms, CFTR function, pulmonary exacerbation rates, and nutritional status. The estimated annualized rate of change in ppFEV1 suggests no evidence of pulmonary function loss across the study population over the 4-year treatment period. These results confirm the favorable long-term safety profile and durable disease-modifying clinical benefits of ELX/TEZ/IVA in adolescents and adults with CF. Clinical trial registration available at www.

CLINICALTRIALS: gov, ID: NCT03525574.

PMID:40209082 | DOI:10.1164/rccm.202411-2231OC

Comput Med Imaging Graph. 2025 Apr 7;123:102538. doi: 10.1016/j.compmedimag.2025.102538. Online ahead of print.

ABSTRACT

Patent foramen ovale (PFO) is one of the main causes of ischemic stroke. Due to the complex characteristics of contrast transthoracic echocardiography (cTTE), PFO classification is time-consuming and laborious in clinical practice. For this reason, a variety of PFO diagnostic models have been presented based on machine learning in recent years. However, existing models have lower diagnostic accuracy due to similar gray values of microbubbles and surrounding myocardial tissue in cTTE. Meanwhile, the greater volume of right-to-left shunt (RLS) volume leads to a higher incidence of migraine and stroke. Existing models do not quantify the severity of RLS, which affects the use of treatment methods in later clinical treatment. To solve these problems, we propose TVUNet++ for left ventricular segmentation and ULSAM-ResNet for PFO classification. More specifically, TVUNet++ can distinguish various local features in cTTE through learnable affinity maps and implicitly capture the semantic relationship between the left heart cavity and the background region. In addition, we provide a benchmark cTTE dataset to evaluate the performance of the proposed model through various experiments. Experimental results show that the average Dice Coefficient of the proposed model can reach 92.11%. Moreover, ULSAM-ResNet can realize multi-scale and multi-frequency feature learning through multiple subspaces and learn cross-channel information for accurate grade classification efficiently. The average recall of static cTTE can reach 84.27%. Furthermore, the proposed model outperforms state-of-the-art models in the grade classification of PFO.

PMID:40209281 | DOI:10.1016/j.compmedimag.2025.102538

J Med Internet Res. 2025 Apr 10;27:e67883. doi: 10.2196/67883.

ABSTRACT

BACKGROUND: Ocular myasthenia gravis (OMG) is a neuromuscular disorder primarily affecting the extraocular muscles, leading to ptosis and diplopia. Effective patient education is crucial for disease management; however, in China, limited health care resources often restrict patients' access to personalized medical guidance. Large language models (LLMs) have emerged as potential tools to bridge this gap by providing instant, AI-driven health information. However, their accuracy and readability in educating patients with OMG remain uncertain.

OBJECTIVE: The purpose of this study was to systematically evaluate the effectiveness of multiple LLMs in the education of Chinese patients with OMG. Specifically, the validity of these models in answering patients with OMG-related questions was assessed through accuracy, completeness, readability, usefulness, and safety, and patients' ratings of their usability and readability were analyzed.

METHODS: The study was conducted in two phases: 130 choice ophthalmology examination questions were input into 5 different LLMs. Their performance was compared with that of undergraduates, master's students, and ophthalmology residents. In addition, 23 common patients with OMG-related patient questions were posed to 4 LLMs, and their responses were evaluated by ophthalmologists across 5 domains. In the second phase, 20 patients with OMG interacted with the 2 LLMs from the first phase, each asking 3 questions. Patients assessed the responses for satisfaction and readability, while ophthalmologists evaluated the responses again using the 5 domains.

RESULTS: ChatGPT o1-preview achieved the highest accuracy rate of 73% on 130 ophthalmology examination questions, outperforming other LLMs and professional groups like undergraduates and master's students. For 23 common patients with OMG-related questions, ChatGPT o1-preview scored highest in correctness (4.44), completeness (4.44), helpfulness (4.47), and safety (4.6). GEMINI (Google DeepMind) provided the easiest-to-understand responses in readability assessments, while GPT-4o had the most complex responses, suitable for readers with higher education levels. In the second phase with 20 patients with OMG, ChatGPT o1-preview received higher satisfaction scores than Ernie 3.5 (Baidu; 4.40 vs 3.89, P=.002), although Ernie 3.5's responses were slightly more readable (4.31 vs 4.03, P=.01).

CONCLUSIONS: LLMs such as ChatGPT o1-preview may have the potential to enhance patient education. Addressing challenges such as misinformation risk, readability issues, and ethical considerations is crucial for their effective and safe integration into clinical practice.

PMID:40209226 | DOI:10.2196/67883

PLoS One. 2025 Apr 10;20(4):e0320006. doi: 10.1371/journal.pone.0320006. eCollection 2025.

ABSTRACT

Obtaining the traces and the characteristics of elongated structures is an important task in computer vision pipelines. In biomedical applications, the analysis of traces of vasculature, nerves or fibres of the extracellular matrix can help characterise processes like angiogenesis or the effect of a certain treatment. This paper presents an objective comparison of six existing methodologies (Edge detection, CT Fire, Scale Space, Twombli, U-Net and Graph Based) and one novel approach called Trace Ridges to trace biomedical images with fibre-like structures. Trace Ridges is a fully automatic and fast algorithm that combines a series of image-processing algorithms including filtering, watershed transform and edge detection to obtain an accurate delineation of the fibre-like structures in a rapid time. To compare the algorithms, four biomedical data sets with vastly distinctive characteristics were selected. Ground truth was obtained by manual delineation of the fibre-like structures. Three pre-processing filtering options were used as a first step: no filtering, Gaussian low-pass and DnCnn, a deep-learning filtering. Three distance error metrics (total, average and maximum distance from the obtained traces to the ground truth) and processing time were calculated. It was observed that no single algorithm outperformed the others in all metrics. For the total distance error, which was considered the most significative, Trace Ridges ranked first, followed by Graph Based, U-Net, Twombli, Scale Space, CT Fire and Edge Detection. In terms of speed, Trace Ridges ranked second, only slightly slower than Edge Detection. Code is freely available at github.com/youssefarafat/Trace_Ridges.

PMID:40209168 | DOI:10.1371/journal.pone.0320006

Int J Med Robot. 2025 Apr;21(2):e70065. doi: 10.1002/rcs.70065.

ABSTRACT

BACKGROUND: Pancreatic diseases such as cancer and pancreatitis pose significant health risks. Early detection requires precise segmentation results. Fully automatic segmentation algorithms cannot integrate clinical expertise and correct output errors, while interactive methods can offer a better chance for higher accuracy and reliability.

METHODS: We proposed a new network-RRM-TransUNet for the interactive pancreas segmentation task in CT images aiming to provide more reliable and precise results. The network incorporates Rotary Position Embedding, Root Mean Square Normalisation, and a Mixture of Experts mechanism. An intuitive interface is constructed for user-aided pancreas segmentation.

RESULTS: RRM-TransUNet achieves outstanding performance on multiple datasets, with a Dice Similarity Coefficient (DSC) of 93.82% and an Average Symmetric Surface Distance error (ASSD) of 1.12 mm on MSD, 93.79%/1.15 mm on AMOS, and 93.68%/1.18 mm on AbdomenCT-1K.

CONCLUSION: Our method outperforms previous methods and provides doctors with an efficient and user-friendly interactive pancreas segmentation experience through the intuitive interface.

PMID:40209153 | DOI:10.1002/rcs.70065

PLoS Genet. 2025 Apr 10;21(4):e1011659. doi: 10.1371/journal.pgen.1011659. Online ahead of print.

ABSTRACT

Genome-wide association studies (GWAS) performed on large cohort and biobank datasets have identified many genetic loci associated with Alzheimer's disease (AD). However, the younger demographic of biobank participants relative to the typical age of late-onset AD has resulted in an insufficient number of AD cases, limiting the statistical power of GWAS and any downstream analyses. To mitigate this limitation, several trait imputation methods have been proposed to impute the expected future AD status of individuals who may not have yet developed the disease. This paper explores the use of imputed AD status in nonlinear transcriptome/proteome-wide association studies (TWAS/PWAS) to identify genes and proteins whose genetically regulated expression is associated with AD risk. In particular, we considered the TWAS/PWAS method DeLIVR, which utilizes deep learning to model the nonlinear effects of expression on disease. We trained transcriptome and proteome imputation models for DeLIVR on data from the Genotype-Tissue Expression (GTEx) Project and the UK Biobank (UKB), respectively, with imputed AD status in UKB participants as the outcome. Next, we performed hypothesis testing for the DeLIVR models using clinically diagnosed AD cases from the Alzheimer's Disease Sequencing Project (ADSP). Our results demonstrate that nonlinear TWAS/PWAS trained with imputed AD outcomes successfully identifies known and putative AD risk genes and proteins. Notably, we found that training with imputed outcomes can increase statistical power without inflating false positives, enabling the discovery of molecular exposures with potentially nonlinear effects on neurodegeneration.

PMID:40209152 | DOI:10.1371/journal.pgen.1011659

Br J Dermatol. 2025 Apr 10:ljaf133. doi: 10.1093/bjd/ljaf133. Online ahead of print.

NO ABSTRACT

PMID:40209097 | DOI:10.1093/bjd/ljaf133

JMIR Hum Factors. 2025 Apr 10;12:e58377. doi: 10.2196/58377.

ABSTRACT

BACKGROUND: The global increase in the Internet of Things (IoT) adoption has sparked interest in its application within the educational sector, particularly in colleges and universities. Previous studies have often focused on individual attitudes toward IoT without considering a multiperspective approach and have overlooked the impact of IoT on the technology acceptance model outside the educational domain.

OBJECTIVE: This study aims to bridge the research gap by investigating the factors influencing IoT adoption in educational settings, thereby enhancing the understanding of collaborative learning through technology. It seeks to elucidate how IoT can facilitate learning processes and technology acceptance among college and university students in the United Arab Emirates.

METHODS: A questionnaire was distributed to students across various colleges and universities in the United Arab Emirates, garnering 463 participants. The data collected were analyzed using a hybrid approach that integrates structural equation modeling (SEM) and artificial neural network (ANN), along with importance-performance map analysis to evaluate the significance and performance of each factor affecting IoT adoption.

RESULTS: The study, involving 463 participants, identifies 2 primary levels at which factors influence the intention to adopt IoT technologies. Initial influences include technology optimism (TOP), innovation, and learning motivation, crucial for application engagement. Advanced influences stem from technology acceptance model constructs, particularly perceived ease of use (PE) and perceived usefulness (PU), which directly enhance adoption intentions. Detailed statistical analysis using partial least squares-SEM reveals significant relationships: TOP and innovativeness impact PE (β=.412, P=.04; β=.608, P=.002, respectively), and PU significantly influences TOP (β=.381, P=.04), innovativeness (β=.557, P=.003), and learning motivation (β=.752, P<.001). These results support our hypotheses (H1, H2, H3, H4, and H5). Further, the intention to use IoT is significantly affected by PE and usefulness (β=.619, P<.001; β=.598, P<.001, respectively). ANN modeling enhances these findings, showing superior predictive power (R2=89.7%) compared to partial least squares-SEM (R2=86.3%), indicating a more effective identification of nonlinear associations. Importance-performance map analysis corroborates these results, demonstrating the importance and performance of PU as most critical, followed by technology innovativeness and optimism, in shaping behavioral intentions to use IoT.

CONCLUSIONS: This research contributes methodologically by leveraging deep ANN architecture to explore nonlinear relationships among factors influencing IoT adoption in education. The study underscores the importance of both intrinsic motivational factors and perceived technological attributes in fostering IoT adoption, offering insights for educational institutions considering IoT integration into their learning environments.

PMID:40209037 | DOI:10.2196/58377

Science. 2025 Apr 11;388(6743):eadq1521. doi: 10.1126/science.adq1521. Epub 2025 Apr 11.

ABSTRACT

Human pelvic evolution following the human-chimpanzee divergence is thought to result in an obstetrical dilemma, a mismatch between large infant brains and narrowed female birth canals, but empirical evidence has been equivocal. By using deep learning on 31,115 dual-energy x-ray absorptiometry scans from UK Biobank, we identified 180 loci associated with seven highly heritable pelvic phenotypes. Birth canal phenotypes showed sex-specific genetic architecture, aligning with reproductive function. Larger birth canals were linked to slower walking pace and reduced back pain but increased hip osteoarthritis risk, whereas narrower birth canals were associated with reduced pelvic floor disorder risk but increased obstructed labor risk. Lastly, genetic correlation between birth canal and head widths provides evidence of coevolution between the human pelvis and brain, partially mitigating the dilemma.

PMID:40208988 | DOI:10.1126/science.adq1521

PLoS Biol. 2025 Apr 10;23(4):e3003099. doi: 10.1371/journal.pbio.3003099. Online ahead of print.

ABSTRACT

Understanding how specific secretory cargoes are targeted to distinct domains of the plasma membrane in epithelial cells requires analyzing the trafficking of post-Golgi vesicles to their sites of secretion. We used the RUSH (retention using selective hooks) system to synchronously release an apical cargo, Cadherin 99C (Cad99C), and a basolateral cargo, the ECM protein Nidogen, from the endoplasmic reticulum and follow their movements to the plasma membrane. We also developed an interactive vesicle tracking framework, MSP-tracker and viewer, that exploits developments in computer vision and deep learning to determine vesicle trajectories in a noisy environment without the need for extensive training data. MSP-tracker outperformed other tracking software in detecting and tracking post-Golgi vesicles, revealing that Cad99c vesicles predominantly move apically with a mean speed of 1.1µm/sec. This is reduced to 0.85 µm/sec by a dominant slow dynein mutant, demonstrating that dynein transports Cad99C vesicles to the apical cortex. Furthermore, both the dynein mutant and microtubule depolymerization cause lateral Cad99C secretion. Thus, microtubule organization plays a central role in targeting apical secretion, suggesting that Drosophila does not have distinct apical versus basolateral vesicle fusion machinery. Nidogen vesicles undergo planar-polarized transport to the leading edge of follicle cells as they migrate over the ECM, whereas most Collagen is secreted at trailing edges. The follicle cells therefore bias secretion of different ECM components to opposite sides of the cell, revealing that the secretory pathway is more spatially organized than previously thought.

PMID:40208901 | DOI:10.1371/journal.pbio.3003099

PLoS One. 2025 Apr 10;20(4):e0320892. doi: 10.1371/journal.pone.0320892. eCollection 2025.

ABSTRACT

BACKGROUND: The allelic variations of the apolipoprotein E (APOE) gene play a critical role in regulating lipid metabolism and significantly impact cardiovascular disease risk (CVD). This study aimed to evaluate the impact of exercise on cardiac structure and function in mouse models expressing different APOE genotypes using photon-counting computed tomography (PCCT) and deep learning-based segmentation.

METHODS: A total of 140 mice were grouped based on APOE genotype (APOE2, APOE3, APOE4), sex, and exercise regimen. All mice were maintained on a controlled diet to isolate the effects of exercise. Low dose cardiac photon counting micro-CT imaging with intrinsic gating was performed using a custom-built micro-PCCT system and data was reconstructed with an iterative algorithm incorporating both temporal and spectral dimensions. A liposomal-iodine nanoparticle contrast agent was intravenously administered to uniformly opacify cardiovascular structures. Cardiac structures were segmented using a 3D U-Net deep learning model that was trained and validated on manually labeled data. Statistical analyses, including ANOVA, post-hoc analysis, and stratified group comparisons, were used to assess the effects of genotype, sex, and exercise on key cardiac metrics, including ejection fraction and cardiac index.

RESULTS: The PCCT imaging pipeline provided high-resolution images with enhanced contrast between blood compartment and myocardium allowing for precise segmentation of cardiac features. Deep learning-based segmentation achieved high accuracy with an average Dice coefficient of 0.85. Exercise significantly improved cardiac performance, with ejection fraction increasing by up to 18% and cardiac index by 46% in exercised males, who generally benefited more from exercise. Females, particularly those with the APOE4 genotype, also showed improvements, with a 31% higher ejection fraction in exercised versus non-exercised mice. Stratified analyses confirmed that both sexes benefited from exercise, with males showing larger effect sizes. APOE3 and APOE4 genotypes derived the greatest benefit, while APOE2 mice showed no significant improvement.

CONCLUSIONS: This study demonstrates the utility of PCCT combined with deep learning segmentation in assessing the cardioprotective effects of exercise in APOE mouse models. These findings highlight the importance of genotype-specific approaches in understanding and potentially mitigating the impact of CVD through lifestyle interventions such as exercise.

PMID:40208877 | DOI:10.1371/journal.pone.0320892

Science. 2025 Apr 11;388(6743):eadq1709. doi: 10.1126/science.adq1709. Epub 2025 Apr 11.

ABSTRACT

We used Hi-C, imaging, proteomics, and polymer modeling to define rules of engagement for SMC (structural maintenance of chromosomes) complexes as cells refold interphase chromatin into rod-shaped mitotic chromosomes. First, condensin disassembles interphase chromatin loop organization by evicting or displacing extrusive cohesin. Second, condensin bypasses cohesive cohesins, thereby maintaining sister chromatid cohesion as sisters separate. Studies of mitotic chromosomes formed by cohesin, condensin II, and condensin I alone or in combination lead to refined models of mitotic chromosome conformation. In these models, loops are consecutive and not overlapping, implying that condensins stall upon encountering each other. The dynamics of Hi-C interactions and chromosome morphology reveal that during prophase, loops are extruded in vivo at ∼1 to 3 kilobases per second by condensins as they form a disordered discontinuous helical scaffold within individual chromatids.

PMID:40208986 | DOI:10.1126/science.adq1709

Science. 2025 Apr 11;388(6743):176-180. doi: 10.1126/science.ads3888. Epub 2025 Apr 10.

ABSTRACT

Denisovans are an extinct hominin group defined by ancient genomes of Middle to Late Pleistocene fossils from southern Siberia. Although genomic evidence suggests their widespread distribution throughout eastern Asia and possibly Oceania, so far only a few fossils from the Altai and Tibet are confidently identified molecularly as Denisovan. We identified a hominin mandible (Penghu 1) from Taiwan (10,000 to 70,000 years ago or 130,000 to 190,000 years ago) as belonging to a male Denisovan by applying ancient protein analysis. We retrieved 4241 amino acid residues and identified two Denisovan-specific variants. The increased fossil sample of Denisovans demonstrates their wider distribution, including warm and humid regions, as well as their shared distinct robust dentognathic traits that markedly contrast with their sister group, Neanderthals.

PMID:40208980 | DOI:10.1126/science.ads3888

PLoS One. 2025 Apr 10;20(4):e0321015. doi: 10.1371/journal.pone.0321015. eCollection 2025.

ABSTRACT

BACKGROUND: HIV/AIDS remains a global health challenge, with significant prevalence in sub-Saharan Africa. Highly active antiretroviral therapy (HAART) is the mainstay treatment for HIV, and the number of people living with HIV (PLWHIV) on HAART has considerably increased worldwide. The use of HAART has led to improved patient outcomes; however, it is associated with adverse drug reactions (ADRs) and drug-drug interactions (DDIs), which pose serious concerns in the management of patients with HIV. The aim of the study was to determine the prevalence and factors associated with ADRs among patients on HAART.

METHODOLOGY: This was a hospital-based cross-sectional study carried out among 312 HIV patients on HAART attending HIV clinics at Mbarara Regional Hospital. Data was collected using an interviewer-administered, semi-structured questionnaire and a review of patient charts. ADRs were assessed for causality and categorized using Naranjo ADR assessment scale into probable, possible and definite, for severity using the modified Hartwig and Siegel criteria into mild, moderate and Severe, and for preventability using Schumock and Thornton criteria into definite, probable and non-preventable. Lexicomp® Drug Interaction Checker software was used to identify and rate clinically significant drug-drug interactions. The prevalence of ADRs and potential DDI was analyzed using descriptive statistics while logistic regression analysis was used to establish the association of variables.

RESULTS: 312 patients were interviewed and their records reviewed. The prevalence of ADRs during this study was 76.0%. On assessment, 78.3% of the ADRs were mild and 76.6% of ADRs were definitely preventable. CD4 count below 200 cells/mm3 (AOR = 1.00, 95% CI: 1.00-1.02; p value = 0.04), primary education level (AOR = 3.27, 95% CI: 1.34-7.95; p value = 0.009), and secondary education level (AOR = 3.64, 95% CI: 1.39-9.52; p value = 0.009) were identified as independent risk factors. Patients who experienced a significant DDI were 5.66 times more likely to experience an ADR (p value = 0.02, 95% CI: 1.32-24.18).

CONCLUSION: There is a high prevalence of adverse drug reactions among patients with HIV on HAART. Low CD4 count and lower education levels are risk factors for ADRs in this population; therefore, tailored interventions to these subgroups should be implemented for early ADR identification and management. Significant drug-drug interactions are highly associated with the occurrence of ADRs among HIV patients on HAART, which calls for intensified pharmacovigilance and pharmaceutical care in this population.

PMID:40208898 | DOI:10.1371/journal.pone.0321015

J Med Syst. 2025 Apr 10;49(1):47. doi: 10.1007/s10916-025-02182-3.

ABSTRACT

Drug response prediction (DRP) is a central task in the era of precision medicine. Over the past decade, the emergence of deep learning (DL) has greatly contributed to addressing DRP challenges. Notably, the prediction of DRP for cancer cell lines benefits significantly from data availability for model development. However, an effective predictive model is still challenging due to issues with data quality, high-dimensional data, and multi-omics data integration. In this study, we introduce MLG2Net, a deep-learning model inspired by graph neural networks designed to predict DRP in lung cancer cell lines based on pharmacogenomics data. Our model comprises two key components: drug SMILES described by local and global graph networks and cell line genomics are illustrated as a map. Our results show that MLG2Net outperforms three reference graph networks. MLG2Net performance reached a Pearson coefficient correlation ( C C p ) of 0.8616 and a root mean square error (RMSE) of 2.94e-6 in predicting drug responses for Lung Adenocarcinoma (LUAD) cell lines. Subsequent testing on the Lung Squamous Cell Carcinoma (LUSC) dataset reveals lower performance ( C C p : 0.7999, RMSE: 4.08e-6), attributed to the dataset's smaller size influencing model capacity. Moreover, we assessed the model's architecture by isolating its components, with results indicating that the global network is particularly effective in this task. In conclusion, MLG2Net exhibited promising applications in DRP for cancer cell lines, with potential advancements by incorporating larger datasets.

PMID:40208442 | DOI:10.1007/s10916-025-02182-3

Ann Am Thorac Soc. 2025 Apr 10. doi: 10.1513/AnnalsATS.202410-1048OC. Online ahead of print.

ABSTRACT

RATIONALE: Incidental features of interstitial lung disease (ILD) are commonly observed on chest computed tomography (CT) scans and are independently associated with poor outcomes. While most studies to date have relied on qualitative assessments of ILD, quantitative imaging algorithms have the potential to effectively detect ILD and assist in risk stratification for population-based cohorts.

OBJECTIVES: To determine whether quantitative measures of ILD are associated with clinically relevant outcomes in the National Lung Screening Trial (NLST).

METHODS: Quantitative measures of ILD were generated using low dose CT (LDCT) data collected as part of the NLST and processed with Computer-Aided Lung Informatics for Pathology Evaluation and Ratings (CALIPER) and deep learning-based usual interstitial pneumonia (DL-UIP) algorithms (Imbio Inc., Minneapolis, MN). A multivariable Cox proportional hazard regression model was used to test the association between ILD measures (percent ground glass opacity, reticular opacity and honeycombing of total lung volume and binary DL-UIP classification) and all-cause mortality. Secondary outcomes of incident lung cancer and lung cancer mortality were also explored.

RESULTS: Quantitative CT data were generated in 11,518 individuals. Mean age was 61.5 years and 58.7% were male. An increased risk of all-cause mortality was observed for each percent increase in CALIPER-derived ground glass opacity (hazard ratio (HR) 1.02, 95% confidence interval (CI) 1.01 - 1.02), reticular opacity (HR 1.18, 95% CI 1.12 - 1.24), and honeycombing (HR 6.23, 95% CI 4.23 - 9.16). Individuals with a positive DL-UIP classification pattern had a 4.8-fold increased risk of all-cause mortality (HR 4.75, 95% CI 2.50 - 9.04). CALIPER derived reticular opacity was also associated with increased lung cancer specific mortality. No quantitative measures of ILD were associated with incident lung cancer.

CONCLUSIONS: Quantitative measures of ILD on LDCT are associated with clinically relevant endpoints in a large at-risk population of individuals with tobacco use history. Primary Source of Funding: This work was supported by the National Institutes of Health Grants K24HL138188 (MKH), F32HL175973 (JMW), T32HL007749 (JMW), R01HL169166 (JMO), R01HL166290 (JMO). Word Count: 324/350.

PMID:40208581 | DOI:10.1513/AnnalsATS.202410-1048OC

J Med Syst. 2025 Apr 10;49(1):47. doi: 10.1007/s10916-025-02182-3.

ABSTRACT

Drug response prediction (DRP) is a central task in the era of precision medicine. Over the past decade, the emergence of deep learning (DL) has greatly contributed to addressing DRP challenges. Notably, the prediction of DRP for cancer cell lines benefits significantly from data availability for model development. However, an effective predictive model is still challenging due to issues with data quality, high-dimensional data, and multi-omics data integration. In this study, we introduce MLG2Net, a deep-learning model inspired by graph neural networks designed to predict DRP in lung cancer cell lines based on pharmacogenomics data. Our model comprises two key components: drug SMILES described by local and global graph networks and cell line genomics are illustrated as a map. Our results show that MLG2Net outperforms three reference graph networks. MLG2Net performance reached a Pearson coefficient correlation ( C C p ) of 0.8616 and a root mean square error (RMSE) of 2.94e-6 in predicting drug responses for Lung Adenocarcinoma (LUAD) cell lines. Subsequent testing on the Lung Squamous Cell Carcinoma (LUSC) dataset reveals lower performance ( C C p : 0.7999, RMSE: 4.08e-6), attributed to the dataset's smaller size influencing model capacity. Moreover, we assessed the model's architecture by isolating its components, with results indicating that the global network is particularly effective in this task. In conclusion, MLG2Net exhibited promising applications in DRP for cancer cell lines, with potential advancements by incorporating larger datasets.

PMID:40208442 | DOI:10.1007/s10916-025-02182-3

Oral Maxillofac Surg. 2025 Apr 10;29(1):79. doi: 10.1007/s10006-025-01371-1.

ABSTRACT

PURPOSE: Not much is known about the applications of artificial intelligence (AI) in cleft lip and/or palate. We aim to perform a scoping review to synthesize the literature in the last 10 years on integrating AI in the approach to this condition and highlight aspects of research into its prediction, diagnosis and treatment.

METHODS: A search was performed via PubMed, Science Direct, Scopus, and LILACS from 2014 to 2024, in which 649 articles were identified, and 3 studies were identified via the snowball method; the title and abstract were identified, and 35 articles were obtained for full reading. Finally, 25 studies were selected after applying the inclusion and exclusion criteria to execute this review.

RESULTS: The articles reviewed included different types of studies, with observational and experimental studies being frequent and systematic reviews and narratives being less frequent. Similarly, there was evidence of a generalized distribution, with a greater concentration in the United States. These studies were analyzed according to the use of AI applied to cleft lip/palate, obtaining 6 subcategories, including diagnosis, prediction, treatment, and education, in which different types of AI models were included, most frequently using deep learning and machine learning.

CONCLUSION: These technologies promise to optimize the care of patients with this condition. Although current advances are promising, further research is essential to expand and refine their beneficial use. AI has driven significant advances in various stages of the cleft lip and/or palate approach, integrating tools such as assisted algorithms, genetics-based predictive models, and advanced surgical planning.

PMID:40208434 | DOI:10.1007/s10006-025-01371-1

J Autism Dev Disord. 2025 Apr 10. doi: 10.1007/s10803-025-06811-1. Online ahead of print.

ABSTRACT

PURPOSE: The objective of this study is to identify the acoustic characteristics of cries of Typically Developing (TD) and Autism Spectrum Disorder (ASD) children via Deep Learning (DL) techniques to support clinicians in the early detection of ASD.

METHODS: We used an existing cry dataset that included 31 children with ASD and 31 TD children aged between 18 and 54 months. Statistical analysis was applied to find differences between groups for different voice acoustic features such as jitter, shimmer and harmonics-to-noise ratio (HNR). A DL model based on Recursive Convolutional Neural Networks (R-CNN) was developed to classify cries of ASD and TD children.

RESULTS: We found a statistical significant increase in jitter and shimmer for ASD cries compared to TD, as well as a decrease in HNR for ASD cries. Additionally, the DL algorithm achieved an accuracy of 90.28% in differentiating ASD cries from TD.

CONCLUSION: Empowering clinicians with automatic non-invasive Artificial Intelligence (AI) tools based on cry vocal biomarkers holds considerable promise in advancing early detection and intervention initiatives for children at risk of ASD, thereby improving their developmental trajectories.

PMID:40208423 | DOI:10.1007/s10803-025-06811-1

Technol Health Care. 2025 Apr 10:9287329251325740. doi: 10.1177/09287329251325740. Online ahead of print.

ABSTRACT

BackgroundBrain tumors pose a significant healthcare challenge, necessitating early detection and precise monitoring to ensure effective treatment.ObjectivesThe study proposes an innovative technique with the integration of hybrid transfer learning with improved microwave imaging. The integration of special feature extraction abilities of pre-trained deep learning methods along with the high-resolution imaging capabilities of the patch antenna.MethodsIt was primarily composed of two phases. The initial stage involves the development of a patch antenna and head phantom model, which are then subjected to SAR analysis to extract pertinent features from transmitted signals. In the second stage, an AI-based detection model that utilizes MobileNet V2 is implemented. The images acquired by the patch antenna system are fed into MobileNet V2, which extracts high-level features by employing depth-wise separable convolutions and inverted residual blocks. The fully connected layer is used to classify brain tumors in an effective manner by passing these extracted features.ResultsThe results of the simulation indicate that the model performs exceptionally well, with an accuracy of 98.44%, precision of 98.03%, recall of 99.00%, F1-score of 98.52%, and specificity of 97.82%.ConclusionThis method offers a promising solution for the non-invasive and real-time detection of brain tumors, taking advantage of the electromagnetic properties of brain tissue and the capabilities of AI to address the limitations of current diagnostic methods, such as MRI and CT scans.

PMID:40208040 | DOI:10.1177/09287329251325740