Communications Medicine

IntroductionThyroid nodules with indeterminate cytology (Bethesda III and IV) present a diagnostic challenge, as conventional cytology offers limited predictive value and can lead to unnecessary surgeries. Recently, validated molecular classifiers have been developed with the aim of improving the stratification of the risk of malignancy in these nodules and optimizing clinical decision-making. Objectives To evaluate and compare the diagnostic yield of validated commercial molecular systems, including ThyroSeq and Afirma, versus conventional cytology in Bethesda III and IV thyroid nodules, using the result of postsurgical histopathology as a reference. MethodA structured review of prospective studies, randomized controlled trials, retrospective cohorts, and meta-analyses that analyzed the performance of commercial molecular classifiers in Bethesda III and IV nodules was conducted. We included studies that reported sensitivity, specificity, positive and negative predictive value, and that used postoperative histopathology as a reference standard. The sample volume of individual studies ranges from several hundred to more than six thousand nodules using pooled analyses. ResultsThe selected studies show that molecular classifiers such as ThyroSeq v3 and Afirma GSC achieve a high sensitivity and negative predictive value ([≥]94% and [≥]96%, respectively), outperforming conventional cytology. Specificity and positive predictive value show greater variability between studies and clinical settings. The use of these classifiers has made it possible to reduce the number of unnecessary surgeries on benign nodules. ConclusionsThe available evidence supports that validated molecular classifiers increase diagnostic accuracy in thyroid nodules with indeterminate cytology, reduce unnecessary surgical interventions, and improve clinical decision-making compared to conventional cytology, using histopathology as a standard reference.

Background: Olfactory dysfunction is a recognised but poorly characterised comorbidity of Primary Ciliary Dyskinesia (PCD). No prior systematic review has synthesised its prevalence or clinical correlates. Methodology: A PRISMA compliant systematic review and meta-analysis was conducted. Five databases were searched to February 2026. Observational studies reporting olfactory function in confirmed PCD were included. Risk of Bias was assessed using the Newcastle-Ottawa Scale. A random-effects meta-analysis using the Freeman-Tukey double arcsine transformation was performed to calculate pooled prevalence with 95% confidence intervals (CI) and prediction intervals (PI). Results: Twelve studies (n=865) were included. Overall pooled prevalence of olfactory dysfunction was 43.4% (95% CI 25.2-62.5%; 95% PI 0.1-99.0%). Objective psychophysical testing yielded a significantly higher pooled prevalence of 66.1% (95% CI 55.5-76.0%; 95% PI 38.4-88.9%) compared to patient-reported outcome measures (30.5%; 95% CI 11.4-54.0%). Older age, greater sinonasal disease burden, and specific ciliary ultrastructural defects were associated with worse olfactory function. A striking discordance between objective dysfunction and subjective awareness was observed across multiple studies. Conclusions: Olfactory dysfunction is highly prevalent in PCD and substantially under-recognised by patients. Routine objective olfactory screening should be integrated into standard multidisciplinary PCD care.

Background: Chronic rhinosinusitis (CRS) is a heterogeneous inflammatory airway disease associated with impaired mucociliary clearance and persistent inflammation. While prior work has focused on inflammatory and molecular pathways, the physicochemical properties of mucus itself remain poorly characterized. This study aimed to define compositional and biophysical features of CRS mucus that may contribute to dysfunction. Methods: A prospective cross-sectional study was conducted in 15 adults undergoing endoscopic sinus surgery (11 CRS, 4 controls). Mucus was collected from the middle meatus. Hydration was measured by lyophilization. Ionic composition was quantified using mass spectrometry. Viscoelasticity was assessed via oscillatory shear rheology. Total protein, total carbohydrate, sialic acid (Sia) and fucose (Fuc) content were quantified using enzymatic and chemical assays. Statistical comparisons were performed using nonparametric tests. Results: CRS mucus exhibited significantly higher Ca2+; and Mg2+; concentrations (approximately two-fold; p<0.05) and increased variability in hydration and ion content compared to controls. Rheology showed greater heterogeneity and a non-significant trend toward increased viscoelasticity in CRS. Total protein and carbohydrate content were not significantly different; however, the carbohydrate-to-protein ratio was significantly reduced in CRS (p=0.04). Sia content and Sia-to-carbohydrate ratio were significantly elevated in CRS (p=0.04 and p=0.002), particularly in CRS with nasal polyps. Fuc content did not differ between groups. Conclusions: CRS mucus demonstrates coordinated alterations in ionic composition and glycosylation, characterized by increased cation content, hypersialylation, and reduced carbohydrate-to-protein ratios. These changes may contribute to altered mucus properties and impaired mucociliary clearance, highlighting mucus composition as a potential therapeutic target in CRS.

IMPORTANCEDifferentiated thyroid cancer (DTC) is commonly diagnosed in women of child-bearing age, but whether pregnancy influences the prognosis of DTC remained controversial. OBJECTIVEThis systematic review and meta-analysis aimed to summarize and appraise the existing evidence of the impact of pregnancy on the prognosis of patients previously treated for DTC. DATA SOURCESWe searched PubMed, Embase, Web of Science, Cochrane, and Scopus until February 2023. STUDY SELECTIONStudies of patients diagnosed and treated for DTC before pregnancy reporting the recurrence/progression condition of DTC were included. Case reports and studies failing to identify the time of diagnosis or initial treatment were excluded. DATA EXTRACTION AND SYNTHESISMeta-analyses were conducted according to MOOSE guideline. Data extraction was conducted by two independent investigators with a standard form. Pooled effect estimates were calculated in a random-effects model. MAIN OUTCOMES AND MEASURESDTC recurrence/progression and the type of recurrence/progression (structural or biochemical). RESULTSAmong the 10 included studies (n = 625), 4 (n = 143) of them compared the pregnancy group with the non-pregnancy group while the remaining 6 (n = 482) only included the pregnant patients. The pooled proportion of recurrence/progression in all pregnant patients was 13% (95% CI, 6%, 25%). Compared with the non-pregnancy group, the pooled odds ratio of recurrence/progression in the pregnancy group was 0.75 (95% CI, 0.45, 1.23). Two included studies focused on patients with distant metastasis and also did not observe difference in disease recurrence/progression between the pregnancy group and the non-pregnancy group [OR, 0.51 (95% CI, 0.14-1.87)]. Six included studies also reported response to therapy status prior to pregnancy, and the pooled proportion for recurrence/progression in pregnant DTC patients with excellent response (n=287), indeterminate response (n=44), biochemical incomplete response (n=41) and structural incomplete response (n=70) was 0.00 (95% CI, 0.00-0.86), 0.09 (95% CI, 0.00-0.99), 0.20 (95% CI, 0.06-0.46) and 0.45 (95% CI, 0.17-0.76), respectively. There was a trend for an increasingly higher risk of recurrence/progression from excellent, indeterminate, biochemical incomplete to structural incomplete response to therapy (P<0.05). CONCLUSIONS AND RELEVANCEPregnancy appears to have a minimal impact on the prognosis of DTC with initial treatment. Clinicians may pay more attention to the progression of DTC among pregnant women with biochemical and/or structural persistence. Key PointsO_ST_ABSQuestionC_ST_ABSDoes subsequent pregnancy has an impact on the prognosis of patients previously treated for differentiated thyroid cancer (DTC)? FindingsIn this systematic review and meta-analysis of 10 studies including 625 patients previously treated for DTC and underwent pregnancy subsequently, pregnancy might have a minimal impact on DTC recurrence/progression. Patients with biochemical and/or structural incomplete response to DTC treatment prior to pregnancy appears to have a higher risk of DTC recurrence/progression compared to those with excellent or indeterminate response. MeaningThough pregnancy appears to have little influence on the prognosis of patients previously treated for DTC, patients with biochemical and/or structural persistence should be more carefully monitored during pregnancy.

Background & AimsLiver cancer primarily develops in patients with chronic liver disease (CLD), yet most cases are diagnosed at an advanced stage with poor prognosis. While clinical surveillance of patients with CLD generates extensive longitudinal data, its unstructured free-text nature hinders large-scale research. To unlock this real-world evidence, we developed a scalable framework using open-source Large Language Models (LLMs) to transform unstructured clinical text into structured data. MethodsWe conducted a multi-stage evaluation of LLM-based extraction from multi-source clinical documentation of liver transplant recipients. A calibration set comprising 507 reports (414 radiology, 65 pathology, and 28 liver transplant assessment reports) from 30 patients was manually annotated to benchmark four open-source LLMs (Llama 3.1 8B, Llama 3.3 70B, Open-BioLLM 70B, DeepSeek R1 8B) against a regular expression baseline across 73 tasks. To ensure structured outputs, we compared constrained decoding (Guidance and Ollama packages) against unconstrained prompting across 5,590 prompt-output pairs. The finalised pipeline was then applied to the full cohort of 835 patients transplanted in our centre over the past decade. ResultsAmong the models tested, Llama 3.3 70B performed best, exceeding 90% accuracy on 59/73 tasks, outperforming both a medically fine-tuned model (OpenBioLLM 70B) and a smaller variant (Llama 3.1 8B). Constrained decoding achieved >99.9% format adherence, far surpassing unconstrained prompting (87.4%). Applied to the full cohort, the pipeline successfully analysed 22,493 reports to generate 37,125 datapoints (45 variables, 835 patients) without manual annotation. Further analysis confirmed known liver cancer risk factors (male sex, viral hepatitis, smoking, diabetes), and allowed for reconstruction of longitudinal disease timelines. ConclusionsThis work provides a scalable blueprint for transforming real-world clinical free-text into structured formats, paving the way for accelerated, data-driven research into complex pre-cancerous diseases like CLD.

This study mapped the age- and region-specific risks of eye diseases in the Brazilian population, evaluating temporal trends and geographical inequalities in access to healthcare. Secondary data from DATASUS, covering the 27 Brazilian federative units from 2010 to 2024, were used, employing hierarchical negative binomial regression. A significant national increase in hospital admission rates was observed during the studied period, with increases of 160.8% for retinopathy, 126.4% for eye and appendage diseases, and 122.8% for glaucoma. State-level heterogeneity was extreme, with variations spanning from -93.1% to +3588% for glaucoma, for example. Even so, regional disparities were observed throughout the period; the South region reported an average 43.2% higher than the national average for retinopathies, and the Southeast 28.5% higher for eye and adnexal diseases, while the North region reported the lowest rates. Projections up to 2036 predict a further national increase of up to +377.0% for retinopathies, with interventions covering more than an order of magnitude. In addition to the temporal projection, rates in state, age, and year components on a logarithmic scale with calibrated uncertainty were verified. Out-of-sample tests show that the chosen modeling outperforms the last observed value maintenance method and naive linear extrapolation in all three diseases considered. Thus, the escalating, age-driven burden of ophthalmological diseases and profound geographic disparities highlight an urgent need to decentralize specialized care and target resource allocation within the public health system.

Suspected colorectal cancer (CRC) referrals based on non-specific symptoms currently lead to large numbers of patients being referred for invasive investigations and poor yield in cancer detection. Secondary care diagnostics, particularly endoscopy, struggle to meet the ever-increasing demand and patients face lengthy waits from the point of referral. Here we propose a blood test utilising high-throughput Raman spectroscopy and machine learning as an accurate triage tool. We present results from the first mixed methods clinical validation study of its kind, evaluating the ability of the test to perform in its target population of primary care patients, and its acceptability to those administering and receiving the test. The test was able to accurately rule out cancer with a negative predictive value of 98.0%. This performance could reduce the number of invasive diagnostic procedures in the cohort by at least 47%. Collectively, our findings promote a novel, non-invasive solution to triage CRC referrals with potential to reduce patient anxiety, accelerate access to treatment and improve outcomes.

Medication adherence is critical for effective management of chronic diseases and reducing healthcare burdens. Statins, commonly prescribed for cardiovascular disease prevention, require sustained, lifelong adherence, yet maintaining long-term adherence remains a significant challenge. Here, we analysed longitudinal electronic health records from over one million statin users in Finland and Italy to characterise adherence trajectories and their determinants. Using functional data analysis, we identified five distinct adherence patterns, with consistently high adherence being the most prevalent across both populations. Younger age, socioeconomic vulnerability, and statin use for primary prevention were consistently associated with a higher risk of declining adherence over time. Sex differences were observed in Italy but not in Finland, where divorced status and health-related educational background were also associated with declining adherence. Despite differences in healthcare systems, several determinants of adherence were consistent across countries. These findings highlight shared behavioural factors underlying long-term statin use and suggest that population-level interventions tailored to patient subgroups defined by adherence patterns may help support sustained medication adherence.

ObjectiveDevelop a neighborhood-level framework using machine learning and causal inference to identify socioeconomic and behavioral drivers of Type 2 diabetes for targeted public health interventions. Materials and MethodsData from 1,149 Census Tracts in Toronto were integrated, linking demographic, health, and marginalization indices. Seven machine learning models classified neighborhoods with high diabetes prevalence. Feature engineering mitigated skewness and correlation, while Causal Forests estimated the Conditional Average Treatment Effect (CATE,{tau} ) for predictors such as work stress, smoking, and mental health. ResultsPredictive models achieved over 90% recall and high AUC metrics on both test and external validation datasets. Key predictors included obesity, overweight status, physical activity, and log-transformed median age. Causal analysis further indicated that elevated work stress ({tau} = 0.312) and daily smoking ({tau} = 0.155) increased diabetes risk, while stronger mental health ({tau} {approx} -1.1) was protective. DiscussionWhile genetic and clinical factors often dominate the conversation on diabetes, data is often restricted to confirmed diagnoses or not readily available for prevalence analyses. Our study shows how neighborhood contexts, including walkability, stress levels, and socioeconomic differences, help drive rising disease rates. We integrated machine learning classifiers with causal inference to examine how interventions, such as active transportation and adjusted work stress, could shift diabetes risk. ConclusionThis integrated method offers a blueprint for precision public health by clarifying how modifiable neighborhood factors affect diabetes risk. It can help tailor interventions to community needs and is applicable to other areas facing similar chronic disease challenges.

Background: Juvenile nasopharyngeal angiofibroma (JNA) is a rare locally aggressive vascular sinonasal tumor that primarily affects adolescent males. Despite advances in endoscopic surgery and preoperative embolization, JNA can be associated with major operative bleeding risk and clinically meaningful recurrence, while non-surgical treatment options remain limited. Methods: To define the cellular programs underlying JNA vascularity, we performed single-cell RNA sequencing of JNA tumors (n=2), tumor-adjacent mucosa, and control sinonasal tissue. We analyzed cell composition, differential gene expression, pathway enrichment, and cell-cell communication, followed by Drug2cell-based mapping of transcriptional states to candidate therapeutic targets. Results: JNA contained an expanded fibrovascular compartment composed of endothelial cells, fibroblasts, pericytes, vascular smooth muscle cells, and neural crest-like cells. Neural crest-like cells were enriched in JNA but showed relatively limited transcriptional differences from tumor-adjacent tissue. By contrast, endothelial cells demonstrated the strongest disease-associated remodeling, with enrichment of angiogenesis, extracellular matrix organization, hypoxia response, and cell migration pathways. Endothelial cells also showed downregulation of adaptive immune signaling pathways, suggesting reduced immune engagement within the tumor microenvironment. Intercellular communication analyses revealed dense endothelial-stromal signaling across the JNA fibrovascular network. Drug2cell analysis nominated VEGF/VEGFR signaling as a candidate therapeutic vulnerability, with VEGFR-targeting agents predicted to act primarily on vascular and lymphatic endothelial populations. Conclusions: JNA is organized around an angiogenesis-dominant fibrovascular program driven by endothelial-centered signaling. These data support further investigation of VEGF/VEGFR-directed therapy as a potential adjunctive strategy for patients with recurrent, unresectable, or surgically high-risk JNA.

Sensitive and specific blood biomarkers for detection of cancer are highly warranted. To discover such biomarkers, we measured plasma levels of 165 proteins and 244 metabolites in 818 patients with colorectal, lung or ovarian carcinoma at diagnosis, 119 patients with non-malignant conditions of the corresponding organs, and 1,129 healthy individuals. We performed exhaustive search over all cut-off values of the ROC statistic and identified composite biomarkers with diagnostic performance significantly superior to benchmark FDA approved blood tests in clinical use for detection of cancer. We found biomarkers composed of 2-4 proteins separating cases of each tumor type from healthy controls with ROC AUC in the range 0.89 to 0.98. These biomarkers also separated cases of each tumor type from the other two (ROC AUC 0.82-0.88). For lung and ovarian cancers, we identified biomarkers distinguishing cases with intermediate and high from those with low tumor stages. These biomarkers for cancer detection and stage can find use in early detection, staging and differential diagnosis of common tumor types.

Hepatoblastoma (HB) is the most common pediatric liver cancer and represents a major clinical challenge, due to the lack of effective therapies for advanced stages and disease relapse. In this work, we use the results of a previously HB-tailored agent-based model of the immune system to investigate whether model-derived variables can be of use in the prediction of patients' outcomes. To this aim, we apply factor analysis to the results of a simulated cohort of HB patients, to identify combinations of key immunological variables able to discriminate disease outcomes in the simulator, and we then assess the coherence of such predictions with independent results of differential expression and enrichment analyses on HB transcriptomics. Our analysis proposes that the ability of immune cells, particularly natural killer and CD8+ cytotoxic T cells, to recognize tumor-associated antigens and exert cytotoxic activity is essential for disease control following treatment.

Background: Artificial intelligence (AI) systems for glaucoma diagnosis and prognostication from visual fields (VF) are under active development, yet do not audit for vertical-meridian-respecting field loss - known sequelae of stroke, hemorrhage, and neoplasm. We developed a self-supervised encoder of automated perimetry that learns anatomically interpretable VF structure without labels, and evaluated its capacity to identify suspected neurologic VF patterns in an independent public glaucoma dataset. Methods: We pretrained a 128-dimensional masked autoencoder on 23,223 unlabeled Humphrey VFs (patient-grouped training split of 28,943 fields from 3,871 patients; UWHVF, all-comers perimetry), using monocular pattern-deviation input. A supervised linear classifier over vertical-midline latent dimensions was trained on per-eye expert neurological/non-neurological labels and assessed under hard-negative cross-validation, with specificity evaluated on 100 held-out, structurally separated UWHVF controls. External evaluation used the Harvard-Glaucoma Fairness dataset (Harvard-GF; 3,300 patients with paired VF and optical coherence tomography [OCT] from a single academic center), which contributed no data at any training stage. Results: Masked reconstruction recovered structure concordant with retinal neuroanatomy: 50 of 128 latent dimensions emerged spatially specialized, versus 23 for the total-deviation encoder. The classifier achieved cross-validated balanced accuracy 0.78 (95% CI, 0.75-0.82) and AUC 0.85 (95% CI, 0.82-0.89), with no false positives among the 100 held-out controls. Applied to Harvard-GF without fine-tuning, it identified a top-20 of 1,748 glaucoma-labeled patients (1.1%) with morphology inconsistent with glaucoma; all 20 were positive on the rule-based Neurological Hemifield Test (mean score 62.4), and OCT showed preserved superior (Cohen d = +0.68; P < .001) and inferior (d = +0.63; P = .003) retinal nerve fiber layer versus severity-matched controls. Conclusions: A self-supervised VF encoder learned anatomically interpretable visual field structure from unlabeled data and identified suspected neurological cases in a curated glaucoma dataset, with expert, rule-based, and OCT corroboration. Visual field datasets used to train glaucoma AI may benefit from neurological screening before model training; the encoder reported here supports such audits and provides a foundation for neuro-ophthalmic AI beyond fundus photography and OCT.

During the COVID-19 pandemic, effective public policy interventions have been crucial in combating virus transmission, sparking extensive debate on crisis management strategies and emphasizing the necessity for reliable models to inform governmental decisions, particularly at the local level. Leveraging disaggregated socio-demographic microdata, including social determinants, age-specific strata, and mobility patterns, we design a comprehensive network model of Catalonias population and, through numerical simulation, assess its response to the outbreak of COVID-19 over the two-year period 2020-21. Our findings underscore the critical importance of timely implementation of broad non-pharmaceutical measures and effective vaccination campaigns in curbing virus spread; in addition, the identification of high-risk groups and their corresponding maps of connections within the network paves the way for tailored and more impactful interventions.

As of early 2026, over 115 million US adults (more than 1 in 3) have prediabetes, a condition with an annual conversion rate of 5%-10% to type 2 diabetes. Total diabetes (diagnosed and undiagnosed) affects approximately 40.1 million Americans, or 12% of the population, with roughly 1.5 million new cases diagnosed annually. Continuous Glucose Monitoring (CGM) provides real-time, 24/7 insights into glycemic variability, detecting dangerous highs, lows, and trends that HbA1c (a 3-month average) misses. It enables, for instance, identification of nocturnal hypoglycemia or postprandial spikes, enhancing personalized, actionable treatment decisions and improving safety. The Artificial Intelligence Ready and Exploratory Atlas for Diabetes Insights (AI-READI) dataset was produced by the National Institutes of Health (NIH) Common Fund Data Ecosystem (CFDE) Bridge2AI program. This dataset offers a rich resource for diabetes research, providing comprehensive biosensor data from over 1,067 participants. However, like many medical datasets, AI-READI contains label inaccuracies due to self-reported health surveys and static HbA1c indicators, which can undermine model effectiveness. We developed a strong classification framework using Convolutional-Bidirectional Long Short-Term Memory (Conv+BiLSTM) to analyze and accurately classify glycemic health states from continuous glucose monitoring time-series data. Our aim was to establish and correct any misclassified labels through hybrid unsupervised-supervised learning methods and validated our results with expert-in-the-loop clinical review. We analyzed 784 participants from the AI-READI dataset, which represented four health states: healthy, prediabetes lifestyle controlled, oral medication, and insulin-dependent. Based on recommendations from the literature and our own expertise, we sought to compare the self-provided "healthy" group labels with a cluster-agnostic, CGM-defined healthy (CGM-H) reference derived from the CGM metrics using K-means clustering (K=6) on standardized CGM summary features to identify CGM-H participants and then applied XGBoost-based iterative label refinement. We identified a misclassification rate of 56.9% (161/283) in the initially labeled "healthy" group. After eight iterations of XGBoost refinement with dual-criterion relabeling ([&ge;]80% probability + unanimous out-of-fold voting), the cleaned dataset increased CGM-H participants from 122 to 195 for binary classification. Next, we developed a Conv+BiLSTM model combining Convolutional layers (32, 64 filters) for local temporal feature extraction with Bidirectional LSTM layers (64, 32 units) for sequence modeling, using time-series engineered features including rolling statistics, glucose derivatives, and circadian rhythm encoding. Class imbalance was addressed with per-class weighting, and 5-fold stratified cross-validation estimated generalization performance, computing a global decision threshold (0.374) by maximizing Youdens J statistic on concatenated out-of-fold predictions. Additionally, we analyzed heart rate, activity level, and stress and sleep data and validated it against CGM data. The Conv+BiLSTM model achieved ROC-AUC {approx} 0.932 on the held-out test set and 0.907 {+/-} 0.026 in cross-validation, with well-calibrated predictions (Expected Calibration Error = 0.075, temperature scaling T = 1.00). A 3-tier confidence-based decision system achieved 82% detection rate with only 6% OGTT burden, enabling actionable clinical recommendations. This hybrid approach addressed label noise while achieving high discrimination. This framework demonstrates potential for real-time glycemic state monitoring and early intervention in diabetes progression.

Background: Olfactory dysfunction (OD) in children remains underdiagnosed and poorly characterised. Despite its known impacts on nutrition, quality of life, safety awareness, and psychosocial development, no standardised diagnostic or management pathway currently exists for paediatric OD. This study aimed to characterise global practice patterns and identify diagnostic and therapeutic challenges unique to paediatric care. Methodology/Principal: A 44-item cross-sectional online survey was distributed to a verified international network of paediatric otolaryngologists across 36 countries via a closed professional platform. The survey assessed five domains: diagnostic practices, management protocols, technology and innovation, education and training, and barriers to effective care. Regional grouping was used to facilitate meaningful statistical comparisons. Categorical variables were evaluated using chi-square tests, with odds ratios and 95% confidence intervals reported for significant findings. Results: Of 351 potential participants, 167 responded (47.6% response rate). Most respondents (83%) reported seeing children with OD, yet 95% saw fewer than ten such patients annually. Psychophysical testing was never performed by 54.8% of respondents, while 88.4% routinely ordered cross-sectional imaging. Testing frequency increased significantly with patient age (Cochran's Q p<0.001). The most common barriers to objective testing were insufficient training (44.3%), time constraints (29.9%), and funding limitations (28.1%). Multidisciplinary collaboration was negligible. Significant regional variation was observed across most practice domains. Conclusions: Paediatric OD care is characterised by functional underinvestigation, fragmented multidisciplinary collaboration, and systemic educational gaps. These findings support urgent development of standardised clinical guidelines, age-appropriate validated assessment tools, and formal interdisciplinary care pathways.

BackgroundSinonasal malignancies frequently present with symptoms overlapping chronic inflammatory conditions such as chronic rhinosinusitis (CRS), complicating early detection and delaying treatment. A fast, scalable, non-invasive approach capable of resolving immune and epithelial cell states across inflammatory and malignant disease from routine nasal swabs could substantially improve clinical screening, leading to the initiation of appropriate treatment. MethodsWe developed a deep learning-enabled single-cell morpholomic framework using the REM-I platform to generate a reference atlas of >641K cell brightfield images from purified immune cell populations. This reference atlas was applied to >2.5 million images obtained from nasal swabs spanning a clinical spectrum of health, CRS, and sinonasal carcinoma. Embeddings were integrated using dimensionality reduction for differential feature testing and comparative feature enrichment across disease states. FindingsAcross the disease continuum, sinonasal carcinoma samples exhibited distinct immune remodeling, including increased myeloid-like cell abundance and elevated small dark pixel intensity consistent with enhanced granulocyte activity. Basophil/NK-enriched clusters contained tumor-associated cells with deep learning-derived morphologic signatures not observed in CRS or healthy samples. Tumor-associated epithelial cells were significantly smaller and displayed disease-specific morpholomic patterns distinct from chronic inflammation. ConclusionsThis study establishes a deep learning-enabled single-cell morpholomic atlas of nasal swabs spanning healthy epithelium, chronic inflammation and sinonasal malignancies. Morpholomic cytology reveals reproducible immune and epithelial states associated with inflammatory and malignant disease and provides a scalable, non-invasive framework for cellular stratification in sinonasal pathology, supporting future applications in early point-of-care diagnostics.

How human respiratory physiology and inhaled airflow therein proceed to impact transmission of SARS-CoV-2, leading to the initial infection, is an open question. An answer can help determine the susceptibility of an individual on exposure to a COVID-2019 carrier and can also quantify the still-unknown infectious dose for the disease. Combining computational fluid mechanics-based tracking of respiratory transport in anatomic domains with sputum assessment data from hospitalized COVID-19 patients and earlier measurements of ejecta size distribution during regular speech - this study shows that the regional deposition of virus-laden inhaled droplets at the initial nasopharyngeal infection sites, located in the upper airway, peaks over the droplet size range of 2.5 - 19 {micro}; and reveals that the number of virions that can potentially establish the infection is, at most, of[O] (102).

Multimodal diseases are those in which affected individuals can be divided into subtypes (or data modes); for instance, mild vs. severe, based on (unknown) modifiers of disease severity. Studies have shown that despite the inclusion of a large number of subjects, the causal role of the microbiome in human diseases remains uncertain. The role of the microbiome in multimodal diseases has been studied in animals; however, findings are often deemed irreproducible, or unreasonably biased, with pathogenic roles in 95% of reports. As a solution to repeatability, investigators have been told to seek funds to increase the number of human-microbiome donors (N) to increase the reproducibility of animal studies (doi:10.1016/j.cell.2019.12.025). Herein, through simulations, we illustrate that increasing N will not uniformly/universally enable the identification of consistent statistical differences (patterns of analytical irreproducibility), due to random sampling from a population with ample variability in disease and the presence of disease data subtypes (or modes). We also found that studies do not use cluster statistics when needed (97.4%, 37/38, 95%CI=86.5,99.5), and that scientists who increased N, concurrently reduced the number of mice/donor (y=-0.21x, R2=0.24; and vice versa), indicating that statistically, scientists replace the disease variance in mice by the variance of human disease. Instead of assuming that increasing N will solve reproducibility and identify clinically-predictive findings on causality, we propose the visualization of data distribution using kernel-density-violin plots (rarely used in rodent studies; 0%, 0/38, 95%CI=6.9e-18,9.1) to identify disease data subtypes to self-correct, guide and promote the personalized investigation of disease subtype mechanisms. HighlightsO_LIMultimodal diseases are those in which affected individuals can be divided into subtypes (or data modes); for instance, mild vs. severe, based on (unknown) modifiers of disease severity. C_LIO_LIThe role of the microbiome in multimodal diseases has been studied in animals; however, findings are often deemed irreproducible, or unreasonably biased, with pathogenic roles in 95% of reports. C_LIO_LIAs a solution to repeatably, investigators have been told to seek funds to increase the number of human-microbiome donors (N) to increase the reproducibility of animal studies. C_LIO_LIHerein, we illustrate that although increasing N could help identify statistical effects (patterns of analytical irreproducibility), clinically-relevant information will not always be identified. C_LIO_LIDepending on which diseases need to be compared, random sampling alone leads to reproducible patterns of analytical irreproducibility in multimodal disease simulations. C_LIO_LIInstead of solely increasing N, we illustrate how disease multimodality could be understood, visualized and used to guide the study of diseases by selecting and focusing on disease modes. C_LI

Evaluating thyroid nodules to rule out malignancy is a very common clinical task. Image-based clinical and machine learning risk stratification schemas rely on the presence of thyroid nodule high-risk sonographic features. However, this approach is less suitable for diagnosing malignant thyroid nodules with a benign appearance on ultrasound. In this study, we developed thyroid cancer polygenic risk scoring (PRS) to complement deep learning analysis of ultrasound images. When the output of the deep learning model was combined with thyroid cancer PRS and genetic ancestry estimates, the area under the receiver operating characteristic curve (AUROC) of the benign vs. malignant thyroid nodule classifier increased from 0.83 to 0.89 (DeLong, p-value = 0.007). The combined deep learning and genetic classifier achieved a clinically relevant sensitivity of 0.95, 95 CI [0.88-0.99], specificity of 0.63 [0.55-0.70], and positive and negative predictive values of 0.47 [0.41-0.58] and 0.97 [0.92-0.99], respectively. An improved AUROC was consistent in ancestry-stratified analysis in Europeans (0.83 and 0.87 for deep-learning and deep learning combined with PRS classifiers, respectively). An elevated PRS was associated with a greater risk of thyroid cancer structural disease recurrence (ordinal logistic regression, p-value = 0.002). This study demonstrates that augmenting ultrasound image analysis with PRS improves diagnostic accuracy, paving the way for developing the next generation of clinical risk stratification algorithms incorporating inherited risk for developing thyroid malignancy.