Spectroscopic data often contain artifacts or noise related to the sample characteristics, instrumental variations, or experimental design flaws. Therefore, classifying the raw data is not recommended and might lead to biased results. Nevertheless, most issues may be addressed through appropriate data pre-processing. Effective pre-processing is particularly crucial in critical applications like liquid biopsy for disease detection, where even minor performance improvements may impact patient outcomes. Unfortunately, there is no consensus regarding optimal pre-processing, complicating cross-study comparisons. This study presents a comprehensive evaluation of various pre-processing methods and their combinations to assess their influence on classification results. The goal was to identify whether some pre-processing methods are associated with higher classification outcomes and find an optimal strategy for the given data. Data from Raman optical activity and infrared and Raman spectroscopy were processed, applying tens of thousands of possible pre-processing pipelines. The resulting data were classified using three algorithms to distinguish between subjects with liver cirrhosis and those who had developed hepatocellular carcinoma. Results highlighted that some specific pre-processing methods often ranked among the best classification results, such as the Rolling Ball for correcting the baseline of Raman spectra or the Doubly Reweighted Penalized Least Squares and Mixture model in the case of Raman optical activity. On the other hand, the selection of filtering and/or normalization approach usually did not have a significant impact. Nonetheless, the pre-processing of top-scoring pipelines also depended on the classifier utilized. The best pipelines yielded an AUROC of 0.775-0.823, varying with the evaluated spectroscopic data and classifier.

• Keywords
• Chiroptical spectroscopy, Classification, Data pre-processing, Diagnostics, Liquid biopsy, Machine learning, Vibrational spectroscopy,
• MeSH
• Algorithms MeSH
• Carcinoma, Hepatocellular * diagnosis   pathology   MeSH
• Liver Cirrhosis diagnosis   pathology   MeSH
• Humans MeSH
• Least-Squares Analysis MeSH
• Liver Neoplasms * diagnosis   pathology   MeSH
• Spectrum Analysis, Raman * methods   MeSH
• Spectrophotometry, Infrared methods   MeSH
• Liquid Biopsy methods   MeSH
• Check Tag
• Humans MeSH
• Publication type
• Journal Article MeSH

OBJECTIVES: The detection and classification of oral mucosal lesions is a challenging task due to high heterogeneity and overlap in clinical appearance. Nevertheless, differentiating benign from potentially malignant lesions is essential for appropriate management. This study evaluated whether a deep learning model trained to discriminate 11 classes of oral mucosal lesions could exceed the performance of general dentists. METHODS: 4079 intraoral photographs of benign, potentially malignant and malignant oral lesions were labeled using bounding boxes and classified into 11 classes. The data were split 80:20 for training (n = 3031) and validation (n = 766), keeping an independent test set (n = 282). The YOLOv8 computer vision model was implemented for image classification and object detection. Model performance was evaluated on the test set which was also assessed by six general dentists and three specialists in oral surgery. Evaluation metrics included sensitivity, specificity, F1-score, precision, area under the receiver operating characteristic curve (AUROC), and average precision (AP) at multiple thresholds of intersection over union. RESULTS: In terms of classification, the highest F1-score (0.80) and AUROC (0.96) were observed for human papillomavirus (HPV)-related lesions, whereas the lowest F1-score (0.43) and AUROC (0.78) were obtained for keratosis. In terms of object detection, the best results were achieved for HPV-related lesions (AP25 = 0.82) and proliferative verrucous leukoplakia (AP25 = 0.80; AP50 = 0.76), while the lowest values were noted for leukoplakia (AP25 = 0.36; AP50 = 0.20). Overall, the model performed comparable to specialists (p = 0.93) and significantly better than general dentists (p < 0.01). CONCLUSION: The developed model performed as well as specialists in oral surgery, highlighting its potential as a valuable tool for oral lesion assessment. CLINICAL SIGNIFICANCE: By providing performance comparable to oral surgeons and superior to general dentists, the developed multi-class model could support the clinical evaluation of oral lesions, potentially enabling earlier diagnosis of potentially malignant disorders, enhancing patient management and improving patient prognosis.

• Keywords
• Artificial intelligence, Computer-assisted diagnosis, Deep learning, Early detection of cancer, Mouth neoplasms, Oral potentially malignant disorders (OPMDs), Squamous cell carcinoma,
• MeSH
• Deep Learning MeSH
• Humans MeSH
• Mouth Neoplasms * classification   diagnosis   pathology   diagnostic imaging   MeSH
• Mouth Diseases * classification   diagnosis   MeSH
• Leukoplakia, Oral MeSH
• ROC Curve MeSH
• Sensitivity and Specificity MeSH
• Machine Learning * MeSH
• Mouth Mucosa * pathology   diagnostic imaging   MeSH
• Dentists * MeSH
• Check Tag
• Humans MeSH
• Publication type
• Journal Article MeSH
• Comparative Study MeSH

BACKGROUND: Polygenic scores (PGSs) hold the potential to identify patients who respond favorably to specific psychiatric treatments. However, their biological interpretation remains unclear. In this study, we developed pathway-specific PGSs (PSPGSs) for lithium response and assessed their association with clinical lithium response in patients with bipolar disorder. METHODS: Using sets of genes involved in pathways affected by lithium, we developed 9 PSPGSs and evaluated their associations with lithium response in the International Consortium on Lithium Genetics (ConLi+Gen) (N = 2367), with validation in combined PsyCourse (Pathomechanisms and Signatures in the Longitudinal Course of Psychosis) (N = 105) and BipoLife (N = 102) cohorts. The association between each PSPGS and lithium response-defined both as a continuous ALDA score and a categorical outcome (good vs. poor responses)-was evaluated using regression models, with adjustment for confounders. The cutoff for a significant association was p < .05 after multiple testing correction. RESULTS: The PGSs for acetylcholine, GABA (gamma-aminobutyric acid), and mitochondria were associated with response to lithium in both categorical and continuous outcomes. However, the PGSs for calcium channel, circadian rhythm, and GSK (glycogen synthase kinase) were associated only with the continuous outcome. Each score explained 0.29% to 1.91% of the variance in the categorical and 0.30% to 1.54% of the variance in the continuous outcomes. A multivariate model combining PSPGSs that showed significant associations in the univariate analysis (combined PSPGS) increased the percentage of variance explained (R 2) to 3.71% and 3.18% for the categorical and continuous outcomes, respectively. Associations for PGSs for GABA and circadian rhythm were replicated. Patients with the highest genetic loading (10th decile) for acetylcholine variants were 3.03 times more likely (95% CI, 1.95 to 4.69) to show a good lithium response (categorical outcome) than patients with the lowest genetic loading (1st decile). CONCLUSIONS: PSPGSs achieved predictive performance comparable to the conventional genome-wide PGSs, with the added advantage of biological interpretability using a smaller list of genetic variants.

Polygenic scores (PGSs) have the potential to identify patients likely to respond to specific psychiatric treatments, but their biological interpretation remains unclear. In this study, we developed 9 pathway-specific PGSs (PSPGSs) for lithium response by aggregating genetic variants involved in pathways affected by lithium. We assessed their associations with lithium response in the International Consortium on Lithium Genetics (ConLi+Gen) (N = 2367) cohort and validated the findings in the PsyCourse (N = 105) and BipoLife (N = 102) cohorts. Clinical response to lithium treatment was significantly associated with PSPGSs for acetylcholine, GABA (gamma-aminobutyric acid), calcium channel signaling, mitochondria, circadian rhythm, and GSK pathways, with explained variance (R 2) ranging from 0.29% to 1.91%. The combined PSPGS explained up to 3.71% of the variability. Associations for GABA and circadian rhythm PGSs were successfully replicated. In a decile-based analysis, patients with the highest genetic load (10th decile) for acetylcholine pathway variants were 3.03 times more likely to respond well to lithium compared with those in the lowest decile (1st decile). PSPGSs achieved predictive performance comparable to conventional genome-wide PGSs, with better biological interpretability and a more focused set of genetic variants.

• Keywords
• Bipolar disorder, Lithium, Pharmacogenomics, Polygenic score, Psychiatry,
• Publication type
• Journal Article MeSH

AIMS: To evaluate the psychometric properties of the Adelphi Adherence Questionnaire (ADAQ©) as an adherence measure in a diverse type 2 diabetes mellitus (T2DM) population in routine clinical practice. MATERIALS AND METHODS: Data were drawn from the Adelphi T2DM Disease Specific Programme™, a survey of physicians and adults with T2DM consulting in the United States, February-May 2021. Participants completed the ADAQ and single questions on medication satisfaction and convenience. Latent variable modelling (exploratory/confirmatory factor analyses, item response theory, Mokken scaling and bifactor analyses) assessed ADAQ dimensionality and composite scoring. Differential item functioning (DIF) analyses were conducted between participants taking injectable and non-injectable therapies, and with <3 or ≥3 long-term conditions. Correlational analyses with physician-reported adherence and compliance, and patient-reported treatment satisfaction, convenience and side-effects assessed construct validity. ADAQ scores by sociodemographic and clinical factors were also described. RESULTS: Overall, 1287 people with T2DM were included in this analysis (mean age 56.7 years [standard deviation: 12.8], 54.5% [n = 702] male). Latent variable modelling indicated a unidimensional reflective model fit, with a bifactor model confirming an 11-question essentially unidimensional composite score. Negligible DIF was found between groups. Cronbach's alpha and McDonald's omega were both ≥0.90. Moderate correlations with physician-reported adherence and compliance, and patient-reported medication convenience and satisfaction support construct validity. CONCLUSIONS: The ADAQ shows strong construct validity and high internal consistency reliability within a heterogenous T2DM population with negligible DIF between sub-groups. Future work should focus on test-retest reliability and detecting change over time.

• Keywords
• ADAQ, Adelphi adherence questionnaire, MLTC, T2DM, multiple long‐term conditions, psychometric evaluation,
• MeSH
• Medication Adherence * psychology   MeSH
• Diabetes Mellitus, Type 2 * drug therapy   psychology   epidemiology   MeSH
• Adult MeSH
• Hypoglycemic Agents * therapeutic use   MeSH
• Middle Aged MeSH
• Humans MeSH
• Surveys and Questionnaires MeSH
• Psychometrics MeSH
• Reproducibility of Results MeSH
• Aged MeSH
• Patient Satisfaction MeSH
• Check Tag
• Adult MeSH
• Middle Aged MeSH
• Humans MeSH
• Male MeSH
• Aged MeSH
• Female MeSH
• Publication type
• Journal Article MeSH
• Geographicals
• United States epidemiology   MeSH
• Names of Substances
• Hypoglycemic Agents * MeSH

BACKGROUND AND OBJECTIVE: Patient-ventilator asynchronies (PVA) are associated with ventilator-induced lung injury and increased mortality. Current detection methods rely on static thresholds, extensive preprocessing, or proprietary ventilator data. This study aimed to develop and validate a fully online, real-time system that detects and classifies PVAs directly from ventilator screen data while alerting clinicians based on severity. METHODS: The SmartAlert system was developed using ventilator screen recordings from ICU patients. It extracts pressure and flow waveforms from video recordings, converts them into time-series data, and employs deep neural networks to classify asynchronies and assign alarm levels from no urgency to most urgent. A dataset of 381,280 double-breath units was independently annotated by two expert intensivists. Two deep learning models were trained: one for alarm prediction and another for asynchrony classification (ineffective triggering, double cycling, high inspiratory effort, no asynchrony). Performance was evaluated using accuracy, sensitivity, specificity, and AUC-ROC, compared to expert consensus. RESULTS: SmartAlert demonstrated strong performance for alarm level prediction (overall accuracy: 83.8 %, weighted AUC-ROC: 0.943 [95 % CI: 0.941-0.945]) and PVA classification (weighted accuracy: 89.3 %, weighted AUC-ROC: 0.951 [95 % CI: 0.950-0.953]). It showed high specificity for urgent alarms (99.9 % for level 3) and PVA types (98.5 % for ineffective triggering, 96.9 % for double cycling, 94.8 % for high inspiratory effort). CONCLUSIONS: We developed and internally validated SmartAlert, an automated system that detects PVAs, classifies severity, and alerts clinicians in real time. Its potential to reduce alarm fatigue, optimize ventilator settings, and improve patient outcomes remains to be tested in clinical trials.

• Keywords
• Deep neural networks, Mechanical ventilation, Patient-ventilator asynchrony, Real-time monitoring,
• MeSH
• Patient-Ventilator Asynchrony MeSH
• Deep Learning MeSH
• Intensive Care Units * MeSH
• Clinical Alarms MeSH
• Humans MeSH
• Ventilators, Mechanical * MeSH
• Neural Networks, Computer MeSH
• Reproducibility of Results MeSH
• ROC Curve MeSH
• Machine Learning * MeSH
• Respiration, Artificial * MeSH
• Check Tag
• Humans MeSH
• Publication type
• Journal Article MeSH

This study employed a simulation approach to model oxygen delivery in spontaneously breathing patients with chronic obstructive pulmonary disease (COPD). The Morozoff Model, originally designed for mechanically ventilated patients with a fixed fraction of inspired oxygen (FIO2), was adapted to incorporate the relationship between oxygen flow delivered through nasal cannula and FIO2, along with COPD-specific pathophysiological parameters. The effectiveness of constant and variable oxygen flow delivery was evaluated using a closed-loop control system with a Proportional-Integral-Derivative (PID) controller. The adapted Morozoff Model successfully replicated SpO2 variations observed in COPD patients, capturing desaturation patterns during rapid eye movement sleep and daily activities. Simulations showed that continuous oxygen flow was inadequate for maintaining SpO2 within the target range. Evaluating the closed-loop control system, a proportional (P) controller was found to be sufficient, with integral (I) and derivative (D) terms having negligible impact on performance for a baseline case. The proportional controller improved SpO2 regulation, increasing time within the target range (88%-92%) to 80%, compared to a maximum of 55% achieved with a constant oxygen flow system. However, as airway resistance increased compared to the baseline case, the controller's performance declined, highlighting the need for re-tuning P and potentially incorporating I and D terms to improve adaptability under varying pathophysiological parameters. In addition, more advanced control strategies, such as model-based controllers, may enhance adaptability to dynamic patient conditions. These findings support the development of adaptive oxygen delivery strategies for spontaneously breathing COPD patients.

• Publication type
• Journal Article MeSH

OBJECTIVE: We previously proposed two cell-free (cf) DNA-based scores (genome-wide Z-score and nucleosome score) as candidate non-invasive biomarkers to further improve the presurgical diagnosis of ovarian malignancy. We aimed to investigate the added value of these cfDNA-based scores in combination with the clinical and ultrasound predictors of the Assessment of Different NEoplasias in the adneXa (ADNEX) model to estimate the risk of ovarian malignancy. METHODS: In this prospective cohort study, 526 patients with an adnexal mass scheduled for surgery were recruited consecutively in three oncology referral centers. All patients underwent a transvaginal ultrasound examination, and adnexal masses were described according to the International Ovarian Tumor Analysis terms and definitions. cfDNA was extracted from preoperative plasma samples and genome-wide Z-scores and nucleosome scores were calculated. Logistic regression models were fitted for ADNEX predictors alone and after inclusion of the cfDNA-based scores. We report likelihood ratios, area under the receiver-operating-characteristics curve (AUC), sensitivity, specificity and net benefit for thresholds between 5% and 40%, to assess the diagnostic performance of the models in discriminating between benign and malignant ovarian masses. RESULTS: The study included 272 benign, 86 borderline, 36 Stage-I invasive, 113 Stage-II-IV invasive, and 19 secondary metastatic tumors. The likelihood ratios for adding the cfDNA-based scores to the ADNEX model were statistically significant (P < 0.001 for ADNEX without CA 125; P = 0.001 for ADNEX including CA 125). The accompanying increases in AUC were 0.013 when the cfDNA biomarkers were added to the ADNEX model without CA 125, and 0.003 when added to the ADNEX model including CA 125. Net benefit, sensitivity and specificity were similar for all models. The increase in net benefit at the recommended 10% threshold estimated risk of malignancy when adding the cfDNA-based scores was 0.0017 and 0.0020, respectively, for the ADNEX model without CA 125 and the ADNEX model with CA 125. According to these results, adding cfDNA markers would require at least 453 patients per additional true-positive test result at the 10% risk threshold. CONCLUSION: Although statistically significant, cfDNA-based biomarker scores have limited clinical utility in addition to established clinical and ultrasound-based ADNEX predictors for discriminating between benign and malignant ovarian masses. © 2025 International Society of Ultrasound in Obstetrics and Gynecology.

• Keywords
• ADNEX, circulating tumor DNA, diagnosis, early detection, fragmentomics, liquid biopsies, nucleosome, ovarian cancer,
• Publication type
• Journal Article MeSH

OBJECTIVES: There is a paucity of validated predictors of response to anti-tumor necrosis factor (TNF) in pediatric Crohn's disease (CD). We aimed to evaluate the predictive utility of intestinal gene expression to predict response to anti-TNF in children with CD. METHODS: We enrolled children with CD before initiating anti-TNF as part of the prospective biobank of the pediatric inflammatory bowel disease Porto group of ESPGHAN. Genes potentially associated with therapeutic response were first preselected from a systematic literature review. Ribonucleic acid was extracted and sequenced from inflamed ileal biopsies of 20 children before initiating anti-TNF (13 with steroid-free remission [SFR] at 12 months, and seven with primary nonresponse [PNR]). An external validation cohort including 22 children (21 SFR, 1 PNR) was enrolled from Germany and Canada. Using maximum relevance-minimum redundancy (mRMR) methods, we constructed a support vector machine-learning model evaluated via leave-one-out cross-validation and permutation testing. RESULTS: Of 1799 studies identified in the systematic review, 24 met the inclusion criteria, reporting on 150 genes possibly associated with anti-TNF response in children or adults. In the Porto group cohort, 30 genes were associated with treatment response, of which five (TREM1, IL23R, CCL7, IL17F, and YES1) were most frequently selected. A multivariable model of these genes achieved high predictive utility (area under receiver operating characteristic curve: 0.88 [95% confidence interval: 0.69-1.0], sensitivity/specificity/positive predictive value/negative predictive value: 92%/71%/86%/83%). The same genomic signature in external validation achieved accuracy of 82% (i.e., 18/22 samples were classified correctly, including the single PNR patient). CONCLUSION: Increased expression of five genes is associated with higher rate of anti-TNF response in pediatric CD. Prospective studies are now warranted to validate these genes as biomarkers for treatment selection.

• Keywords
• RNA expression, biologics, inflammatory bowel disease,
• Publication type
• Journal Article MeSH

Objective.One major advantage of proton therapy (PT) over conventional photon radiotherapy is reduced dose delivered to normal tissue. However, the complexity of the secondary radiation field composed of a mixture of particles with a wide energy range makes its characterization a challenging task.Approach.Measurements with a miniaturized Timepix detector were carried out in three positions out-of-field (7.4 cm, 14.1 cm, and 18.5 cm from the isocenter), inside a phantom resembling a 5 year old undergoing proton pencil beam scanning treatment for a brain tumor. Total and particle-specific deposited energy, absorbed dose, and dose equivalent in water were calculated. Results were compared with thermoluminescent detectors (TLDs) measurements and Monte Carlo (MC) simulations modelling the experimental setup.Main results.The proton absorbed dose in water normalized to the target dose, ranged from 4.8 mGy Gy-1to 65.5µGy Gy-1, while the gamma dose, which remained consistently lower, ranged between 88.4µGy Gy-1and 6.1µGy Gy-1. The measured dose equivalent varied between 6.3 mSv Gy-1and 82.3µSv Gy-1. Good agreement was observed for the two farthest-locations when comparing the absorbed dose in water estimated by the MiniPIX Timepix detector with TLD measurements and MC simulations. However, the closest position showed an overestimation for both the absorbed dose and the dose equivalent, while the farthest position exhibited an underestimation for the dose equivalent.Significance.Out-of-field dosimetry in PT is challenging due to the complexity of the secondary mixed radiation field. Multiple detectors are typically required, but many are too large for use in anthropomorphic phantoms. This study demonstrates that the MiniPIX Timepix detector can accurately determine absorbed dose, dose equivalent and particle-specific contributions (electrons/gammas, protons, and ions). Unlike passive detectors such as TLDs, it enables active measurements with high time resolution, allowing dose rates analysis. The results, validated through experimental data and MC simulations, support the detector's potential for reliable out-of-field dose assessment and improved patient safety.

• Keywords
• MiniPIX Timepix detector, Monte Carlo (MC) simulations, TOPAS, anthropomorphic phantom, out-of-field dose, proton therapy (PT), thermoluminescent detectors (TLDs),
• MeSH
• Time Factors MeSH
• Radiotherapy Dosage MeSH
• Phantoms, Imaging MeSH
• Humans MeSH
• Monte Carlo Method MeSH
• Proton Therapy * instrumentation   methods   MeSH
• Radiometry * instrumentation   MeSH
• Check Tag
• Humans MeSH
• Publication type
• Journal Article MeSH

BACKGROUND AND OBJECTIVE: Cholangiocarcinoma (CCA) is a heterogeneous neoplasm of the biliary epithelium that easily infiltrates, metastasises and recurs. Magnesium disbalance is a hallmark of CCA, with the magnesium transporter cyclin M4 (CNNM4) being a key driver of various hepatic diseases. This study aims to unravel the role of CNNM4 in the initiation and progression of CCA. DESIGN: CNNM4 protein and gene expression were assessed in vitro, in vivo and in patients with CCA. Silencing of CNNM4 was effectively achieved by using small interfering RNA (siRNA) or short hairpin RNA in CCA cell lines and GalNAc-conjugated siRNA in a transposon-based CCA mice model. The impact of CNNM4 on tumour cell proliferation, migration and invasion to the lungs was evaluated using the chicken chorioallantoic membrane model. Proteomic analysis was employed to elucidate the underlying molecular mechanisms. RESULTS: CNNM4 was upregulated in CCA samples from humans, mice and cell lines. Functional studies demonstrated that CNNM4 deficiency attenuates cell growth, chemoresistance, migration, invasion, cancer stem cell properties and Warburg effect in vitro and in vivo. Proteomic analysis identified nuclear protein 1 as an upstream regulator of CNNM4-induced ferroptosis in CCA, ultimately leading to cell death. The iron chelator deferiprone could reverse the decreased proliferation induced by CNNM4 silencing, while inhibition of the heme oxygenase-1 by zinc protoporphyrin IX affected only the growth of cells with no targeted CNNM4 inhibition, highlighting the specificity of ferroptosis in CNNM4-associated effects. CONCLUSION: This study reveals that increased CNNM4 expression drives CCA progression and malignancy and that its inhibition may be an effective therapeutic strategy to limit proliferation and metastasis in patients with CCA.

• Keywords
• CHOLANGIOCARCINOMA, GENE THERAPY, IRON METABOLISM, REACTIVE OXYGEN SPECIES,
• Publication type
• Journal Article MeSH