Introduction Radiotherapy for breast cancer must deliver a sufficient dose to the targets while limiting the dose to the heart, lungs, and contralateral breast. Treatment plans are evaluated on many dose-volume metrics at once, and the balance between target coverage and organ sparing depends on the patient's setup and on the radiotherapy technique. We investigate the relationships among doses to targets, organs, and patient and treatment factors, and examine whether principal component analysis (PCA) can display these relationships in a small factorial planning dataset. Methods PCA was applied to the data from a quasi-experimental two-level factorial treatment planning study of breast cancer adjuvant radiotherapy. The quasi-experiment assessed all combinations of five variables in eight patients: tumor laterality (left vs. right), surgery (mastectomy vs. breast-conserving), target volume (with vs. without lymph nodes), patient's setup position (supine vs. prone) (each patient underwent simulation in the two positions), and radiation delivery (conventional radiotherapy vs. tomotherapy) (both plans were computed in each setup position), which resulted in 32 distinct combinations of the variables. Results PCA showed that tumor left laterality and prone setup were associated with higher heart and contralateral breast doses; nodal irradiation and conventional radiotherapy were associated with higher lung doses, whereas post-mastectomy radiation contributed to high doses to both heart and lung. Tomotherapy was associated with reduced lung doses, better target coverage, slightly reduced heart doses, and increased contralateral breast doses. Conclusion PCA on a factorial design identified correlations between treatment factors and dosimetric outcomes with a minimal number of patients, which might be important in exploratory research, reducing the initial burden of accruing patients.

Accurate and timely identification of borehole deviation is essential in drilling, as unexpected trajectory shifts serve as key indicators of localised geomechanical failure, formation anisotropy, and wellbore integrity degradation. This study presents a systematic machine learning (ML) framework for classifying borehole deviation as Normal or deviated, utilising seven petrophysical and geomechanical well-log features from a 301-sample single-well dataset. Five supervised classifiers, K-Nearest Neighbours (KNN), Logistic Regression (LR), Gaussian Naive Bayes (GNB), Multilayer Perceptron (MLP), and Histogram Gradient Boosting (HGB) were evaluated. Preprocessing incorporated standard scaling, median imputation, and class imbalance mitigation using balanced class weights and the Synthetic Minority Oversampling Technique (SMOTE). To eliminate spatial data leakage, models were evaluated within a 10-fold depth-blocked cross-validation framework across five random seeds, yielding 95% confidence intervals for all performance metrics. The HGB model achieved the highest macro-averaged F1-score of 97.88% (±2.05% CI) and an out-of-fold cross-validation accuracy of 98.00%, outperforming KNN (94.39%), MLP (86.04%), GNB (81.04%), and LR (77.03%). Permutation-based feature importance highlighted Delta T, Gamma-ray, and Resistivity as the primary indicators of deviation risk, aligning with geomechanical principles. These findings demonstrate that Histogram Gradient Boosting with leakage-safe preprocessing provides a robust offline proof-of-concept for assessing wellbore trajectory quality before field validation.

This exploratory retrospective analysis investigated whether clinical and radiomic features could provide prognostic information regarding 6-month outcomes following regenerative periodontal treatment. Thirty-nine defects were classified using the Composite Outcome Measure as successful (COM1) or non-successful (COM2-4). Patient- and defect-related variables were obtained from clinical records, while radiomic features were extracted from baseline periapical radiographs using a standardized trapezoidal region of interest. After preprocessing, 29 radiomic features were retained. Univariate analysis, redundancy assessment, and multivariable logistic regression identified independent predictors, which were then combined in supervised machine learning models evaluated by stratified 5-fold cross-validation. Twenty-one defects (53.8%) achieved COM1, whereas 18 (46.2%) were classified as non-successful. Older age, higher sphericity, and lower intensity histogram 90th percentile values were independently associated with non-successful outcomes. The final logistic regression model showed good discrimination (AUC = 0.84; accuracy = 0.74; sensitivity = 0.72; specificity = 0.76). These findings suggest that integrating clinical and radiomic variables may support early prediction of regenerative outcomes, although external validation is required.

Efficient and accurate assessment of the severity of areca palm (Areca catechu L.) yellow leaf disease (YLD) is essential for early prevention, precision pesticide application, and the sector's long-term viability. However, accurate detection of YLD from UAV imagery remains a significant challenge owing to severe canopy occlusion, subtle early-stage lesions, intricate background clutter, and minimal inter-class variance. Given these challenges, we introduce an optimized YOLOv10s model for YLD detection. First, a non-parametric SimAM attention mechanism is incorporated into the backbone to engineer the C2f_SimAM module; this design optimizes feature extraction within lesion areas while successfully mitigating intricate background noise. Second, the Histogram Transformer Block (HTB) is embedded into the deep feature extraction stage, which strengthens multi-scale feature extraction via dynamic-range histogram self-attention, thereby advancing the network's capability in capturing diminutive targets. Finally, we incorporate a Soft-NMS-GIoU strategy to alleviate localization bias and reduce missed detections under dense occlusion. The upgraded YOLOv10s outperforms the baseline by 2.4 percentage points, reaching 92.9% mAP0.5. Precision and recall reach 88.7% and 86.8%, respectively, at a frame rate of 88.350 FPS. The confusion between healthy and mild YLD categories is significantly reduced, with their respective recall rates improved by 3% and 7%. Field deployment in areca palm plantations demonstrates the model's proficiency in pinpointing disease hotspots and generating visualization maps, confirming its practicality and reliability. The methodology developed in this work offers a robust technical foundation for the large-scale automated monitoring and precise management of YLD.

Population-based screening in Australia for breast cancer using mammography has delivered key outcomes in reducing deaths, but such programs are usually not present in Vietnam. Breast cancer is common in both countries, with Vietnamese women having high breast density and Vietnam having low radiology expertise. This paper investigated the performance of two state-of-the-art Artificial Intelligence (AI) models for cancer detection on Datasets 1 and 2 (Vietnam and Australian mammographic cases) and determined whether the breast density in the two datasets with different vendors affected the AI performance. Both datasets consisted of the same number of mammographic cases (865 malignant; 865 normal). Mammographic images were enhanced using the contrast-limited adaptive histogram equalization algorithm. Transfer learning of the Globally-aware Multiple Instance Classifier (GMIC) and Global-Local Activation Maps (GLAM) AI models was conducted on enhanced images in the two national datasets. The breast density was classified as four levels (A, B, C, and D) by three radiologists. We combined Levels A and B into 0%-50%, and Levels C and D into 50%-100% breast density. The performance of these two AI models, both with and without transfer learning, was evaluated on the two datasets using McNemar's test. The GMIC in the pre-trained and transfer learning modes outperformed the GLAM in terms of specificity and sensitivity in the two datasets. The performance of the GMIC and GLAM with transfer learning in Dataset 2 was better, with 86.7% specificity and 88.2% sensitivity, outperforming their transfer learning model with 80.4% specificity and 81.2% sensitivity in Dataset 1. The specificity and sensitivity of the two AI models (p-values <0.05) were significantly improved using transfer learning in the two datasets. The specificity and sensitivity of GMIC with transfer learning (p-values <0.05) on the 0%-50% and 50%-100% breast density in Dataset 2 were significantly higher than those in Dataset 1, respectively.

A central challenge in molecular biology is understanding how amino acid substitutions modulate various features of protein function and stability. To illuminate the complexities of this relationship, high-throughput (HTP) assays are increasingly used to assess site-saturating mutagenesis libraries. A common downstream analysis is to average the set of twenty outcomes at each amino acid position for comparison with structural and evolutionary features. Average values clearly identify positions that tolerate most substitutions (neutral positions) and positions where most substitutions abolish activity (toggle positions). However, average values conceal the existence of rheostat positions, where different amino acid substitutions sample a wide range of outcomes. To quantitatively identify rheostat positions, we previously developed a histogram-based analysis that we here expand by: (i) incorporating new position classes observed in experimental studies of rheostat positions; (ii) formalizing a hierarchy of class assignments; (iii) refining error-based identification of neutral positions; and (iv) statistically assessing the robustness of class assignments to changes in experimental and computational parameters. RheoScale 2.0 is implemented in Excel and newly implemented in Python for facile integration with existing HTP pipelines; all parameters are customizable. Example analyses are shown for three HTP datasets of the SARS-CoV-2 papain-like protease. Results illustrate two aspects that influence interpretation of HTP data: First, position assignments (and substitution outcomes) depend highly on the measured feature. Second, many protein positions play multiple roles in the sequence-structure-function relationship. The recognition of varied position roles will advance understanding of pathogen evolution, protein engineering, and variant interpretation for personalized medicine.

Stereotactic radiosurgery delivers an ablative single-fraction dose to small intracranial targets, and the reported dose depends on the calculation algorithm. This exploratory case series quantified the change in reported dosimetry when plans clinically finalised with the Anisotropic Analytical Algorithm (AAA) were recalculated with Acuros XB (AXB) in dose-to-medium mode. Seven consecutive brain metastases in six patients, treated with 6 MV flattening filter-free volumetric modulated arc therapy, were recalculated on the same CT dataset using identical arc geometry, control points, multileaf collimator sequences and monitor units, without re-optimisation or renormalisation. Dose-volume histogram parameters were compared for the planning target volume (PTV), normal brain and organs at risk, and plan quality with the RTOG prescription-isodose-to-target-volume (PITV) ratio, the Paddick conformity index (PCI) and the gradient index (GI). Paired comparisons used exact two-sided Wilcoxon signed-rank tests and are interpreted as exploratory. AXB reported lower PTV dose metrics than AAA for the same deliverable plans, with mean D95% falling from 21.66 to 21.10 Gy, D98% from 21.19 to 20.58 Gy and Dmean from 24.58 to 24.13 Gy (all p = 0.016). The mean paired D98% difference was 3.00 ± 0.80% of the prescription dose. The PITV ratio fell from 1.19 to 1.10 and GI rose from 3.23 to 3.43 (both p = 0.016), whereas PCI did not differ (0.79 versus 0.82, p = 0.188). Normal-brain V12Gy and V10Gy were unchanged. Routine electronic portal imaging device dosimetry of the delivered plans gave a mean 2%/2 mm gamma passing rate of 98.5 ± 1.7%, documenting deliverability rather than validating either algorithm. These findings support careful review of target-dose metrics when an established AAA workflow is transitioned to AXB. The cohort is small and no independent reference dosimetry was available, so the study establishes neither clinical superiority nor an effect on tumour control or toxicity.

Conventional monocular polarimetric imaging in turbid underwater environments suffers from severe particulate interference, leading to unreliable polarimetric parameter estimation and loss of image detail. This paper proposes AM-PDC, a multi-view polarimetric restoration framework based on sequential 5×5 spatial sampling using a single translated polarimetric camera. Feature alignment and pixel-level median fusion suppress stochastic noise, while multi-view variance characteristics enable automatic background extraction for estimating the background polarization direction. Backscatter suppression is performed through orthogonal polarization decomposition guided by the estimated background polarization direction, followed by contrast-truncated histogram stretching (CTHS) and contrast-limited adaptive histogram equalization (CLAHE) applied to the luminance channel. Experimental results demonstrate that AM-PDC recovers structural and textural information across varying turbidity levels and achieves improved structural preservation, visual quality, and detail enhancement under the evaluated medium-to-high turbidity conditions.

In petrochemical environments, dynamic illumination fluctuations severely degrade inspection image quality, leading to inaccurate instrument readings. To address this issue, a non-uniform illumination correction method is proposed that integrates improved K-means clustering with an adaptive 2D Gamma function. The clustering step adaptively segments illumination levels via histogram peak-valley extraction. The results then guide Gaussian scale selection and adaptive weighting for illumination estimation. Finally, the adaptive 2D Gamma function is dynamically optimized via parameter estimation and monotonicity constraints. Extensive experiments under varying illumination conditions demonstrate the robustness of the proposed method, maintaining entropy values above 7.05, average gradients ranging from 4.05 to 8.79, and NIQE scores as low as 3.21. The proposed method effectively corrects non-uniform illumination to facilitate subsequent instrument reading tasks.