Optimizing Breast Cancer Detection: Integrating Few‐Shot and Transfer Learning for Enhanced Accuracy and Efficiency

The objective was to study the dynamics of B-type natriuretic peptide (BNP) and its relationship with hemodynamic parameters during on-pump coronary artery bypass grafting (CABG), and to evaluate the informativeness of the biomarker as a predictor of myocardial dysfunction.Materials and methods. The study involved 127 patients aged 59 [54–66.75] years with ischemic heart disease who underwent CABG. The BNP blood level was determined in the operating room at stages: I – before surgery (BNP1 ), II – at the end of surgery (BNP2 ). Hemodynamic parameters were analyzed at the same stages. Correlation analysis, logistic regression with the calculation of the odds ratio (OR) and 95% confidence interval (95% CI) and ROC analysis with the calculation of the area under the ROC curve (AUC) were used.Results. BNP1 blood level was 49 [25.6–91.6], BNP2 – 90 [47.8–140.2] pg/ml (p < 0.0001). BNP1 correlated with central venous pressure (CVP) at stage I (rho = 0.212; p = 0.017) and with pulmonary artery wedge pressure (PAWP) at stage II (rho = 0.204; p = 0.045). BNP2 correlated with PAWP at stage II (rho = 0.204; p = 0.045). BNP1 > 52.1 pg/ml was the predictor of ICU length of stay > 24 hours (OR 1.0290, 95% CI 1.0154– 1.0427, p < 0.0001, AUC 0.775), BNP1 > 71 pg/ml was the predictor of inotropic index > 5 c. u. (OR 1.0076, 95% CI 1.0015–1.0138, p = 0.014, AUC 0.705) and BNP1 > 90.8 pg/ml was the predictor of vasoactive inotropic index > 10 c. u. (OR 1.0070, 95% CI 1.0014–1.0126, p = 0.013, AUC 0.727). BNP2 > 67.5 pg/ml was the predictor of ICU length of stay > 24 hours (OR 1.0179, 95% CI 1.0073–1.0287, p < 0.0009, AUC 0.763), BNP2 > 94.3 pg/ml was the predictor of inotropic index > 5 c. u. (OR 1.0063, 95% CI 1.0010–1.0117, p = 0.020, AUC 0.713), BNP2 > 144 pg/ml was the predictor of intra-aortic balloon pumping (OR 1.0037, 95% CI 1 .0000–1.0074, p = 0.048, AUC 0.854), BNP2 > 159 pg/ml was the predictor of vasoactive inotropic index > 10 c. u. (OR 1.0072, 95% CI 1.0006–1.0139, p = 0.033, AUC 0.729) and BNP2 > 161 pg/ml was the predictor of early mortality in the ICU (OR 1.0040, 95% CI 1, 0000-1.0080, p = 0.049, AUC 0.845). Conclusion. In 78.7% of patients undergoing on-pump CABG, BNP blood level does not exceed the upper limit of normal; by the end of surgery, the biomarker level increases by 32.9 [17.7–62.0] pg/ml. Before and at the end of surgery, BNP values are weakly correlated with CVP and PAWP and do not correlate with other hemodynamic parameters. Before surgery, BNP blood level in the range of 52.1–90.8 pg/ml are predictors of ICU stay > 24 hours (AUC 0.775), inotropic scale > 5 (AUC 0.705) and vasoactive-inotropic scale > 10 c. u. (AUC 0.727). At the end of surgery, BNP > 67.5 pg/ml is associated with an ICU stay > 24 hours (AUC 0.763), and BNP > 90.4 pg/ml is associated with inotropic scale > 5 c. u. (AUC 0.713). The BNP, increased to 144.0–161.0 pg/ml, indicates severe myocardial dysfunction, including hemodynamic support with intra-aortic balloon pumping (AUC 0.854), vasoactive-inotropic scale > 10 c. u. (AUC 0.729) and the risk of early mortality in the ICU (AUC 0.845).

The use of machine learning models in consequential decision making often exacerbates societal inequity, in particular yielding disparate impact on members of marginalized groups defined by race and gender. The area under the ROC curve (AUC) is widely used to evaluate the performance of a scoring function in machine learning, but is studied in algorithmic fairness less than other performance metrics. Due to the pairwise nature of the AUC, defining an AUC-based group fairness metric is pairwise-dependent and may involve both intra-group and inter-group AUCs. Importantly, considering only one category of AUCs is not sufficient to mitigate unfairness in AUC optimization. In this paper, we propose a minimax learning and bias mitigation framework that incorporates both intra-group and inter-group AUCs while maintaining utility. Based on this Rawlsian framework, we design an efficient stochastic optimization algorithm and prove its convergence to the minimum group-level AUC. We conduct numerical experiments on both synthetic and real-world datasets to validate the effectiveness of the minimax framework and the proposed optimization algorithm.

Total prostate-specific antigen (tPSA) is flawed for prostate cancer (PCa) detection. [-2]proprostate-specific antigen (p2PSA), a molecular isoform of free PSA (fPSA), shows higher specificity compared with tPSA or percentage of free PSA (%fPSA). The prostate health index (Phi), a measure based on p2PSA and calculated as p2PSA/fPSA × √tPSA, was evaluated in a multicenter study for detecting PCa. A total of 1362 patients from 4 different study sites who had tPSA values of 1.6-8.0 μg/L (668 patients with PCa, 694 without PCa) underwent ≥10 core biopsies. Serum concentrations of tPSA, fPSA (both calibrated against a WHO reference material), and p2PSA were measured on Access2 or DxI800 analyzers (Beckman Coulter). The percentage ratio of p2PSA to fPSA (%p2PSA) and Phi were significantly higher in all PCa subcohorts (positive initial or repeat biopsy result or negative digital rectal examination) (P < 0.0001) compared with patients without PCa. Phi had the largest area under the ROC curve (AUC) (AUC = 0.74) and provided significantly better clinical performance for predicting PCa compared with %p2PSA (AUC = 0.72, P = 0.018), p2PSA (AUC = 0.63, P < 0.0001), %fPSA (AUC = 0.61) or tPSA (AUC = 0.56). Significantly higher median values of Phi were observed for patients with a Gleason score ≥7 (Phi = 60) compared with a Gleason score <7 (Phi = 53; P = 0.0018). The proportion of aggressive PCa (Gleason score ≥7) increased with the Phi score. The results of this multicenter study show that Phi, compared with tPSA or %fPSA, demonstrated superior clinical performance in detecting PCa at tPSA 1.6-8.0 μg/L (i.e., approximately 2-10 μg/L in traditional calibration) and is better able to detect aggressive PCa.

Introduction: HDL cholesterol (HDL-C) may not always represent the HDL function. This study was conducted to help predict the risk of coronary artery disease (CAD) in patients with type 2 diabetes (T2D) based on substituting paraoxonase 1 (PON1) as an important enzyme in HDL function for that of HDL-C in the atherogenic ratios LDL-C/HDL-C and log (TG/HDL-C). Methods: A total of 274 subjects undergoing diagnostic coronary angiography in this study; 92 without significant CAD and 182 with severe CAD. The hydrolysis rate of phenylacetate as a PON1 substrate was spectrophotometrically measured in kinetic mode. The area under the ROC curve (AUC) was calculated to quantify the overall discriminative capacity of the variables. Results: ROC analyses showed that LDL-C/PON1 (AUC=0.66, p=0.02) and log (TG/PON1) (AUC=0.65, p=0.019) had higher predictive powers compared with LDL-C/HDL-C (AUC=0.55, p=0.221) and log (TG/HDL-C) (AUC=0.57, p=0.121) for CAD characterization in T2D patients. Also, AUCs of LDL-C/PON1 (AUC=0.74, p=0.014) and log (TG/PON1) (AUC=0.67, p=0.057) were higher than those of LDL-C/HDL-C (AUC=0.51, p=0.469) and log (TG/HDL-C) (AUC=0.53, p=0.364) for predicting CAD in T2D patients with HbA1c ≥7%. The multiple regression analysis with adjustments for gender, age, BMI, duration of T2D, metformin and statin therapy indicated that LDL-C/PON1 [adjusted OR=1.1 (0.99,1.23), p=0.085] and log (TG/PON1) [adjusted OR=3.76 (0.94,14.89), p=0.058] were independently predictors of CAD in T2D patients. Conclusions: New cardiometabolic biomarkers using PON1 provide an incremental predictive value compared with those using HDL-C in identifying subjects at increased risk for CAD among patients with T2D and poor glycemic control. This study sought to exploit the lipoprotein-related biomarkers of CAD from a more effective perspective in T2D. Our findings may help develop clinically cost-effective and functional laboratory measurements to evaluate HDL function.

Lung cancer is considered one of the leading causes of death all across the world. Various radiology-related fields increasingly have used Computer-aided diagnosis (CAD) systems. It just has already become a part of clinical work for lung cancer detection. In this article, we proposed an Adaptive Boost-based Grid Search Optimized Random Forest (Ada-GridRF) classifier that best optimized the hyperparameters of the base random forest model to identify the malignant and non-malignant nodules from the trained CT images. Improved performance speed and reduced computational complexity were the advantages of the proposed method. The proposed methodology was compared with other hyperparameter optimization techniques and also with different conventional approaches. It even outperformed the popular state-of-the-art deep learning techniques such as transfer learning and convolutional neural network. The experimental results proved that the proposed method yielded the best performance metrics of 97.97% accuracy, 100% sensitivity, 96% specificity, 96.08% precision, 98% F1-score, 4% False positives rate, and 99.8% Area under the ROC curve (AUC). It took only 8 msec to train the model. Thus, the proposed Ada-GridRF model can aid radiologists in fast lung cancer detection.

To address the public health issue of renal failure and the global shortage of nephrologists, an AI-based system has been developed to automatically identify kidney diseases. Recent advancements in machine learning, deep learning (DL), and artificial intelligence (AI) have unlocked new possibilities in healthcare. By harnessing these technologies, we can analyze data to gain insights into symptoms and patterns, ultimately facilitating remote patient care. To create an AI-based diagnosis system for kidney disease, this paper focused on the three major categories of kidney diseases: stones, cysts, and tumors, which were collected and annotated on 12,446 computed tomography (CT) whole abdomen and urogram images. To effectively aid in the automatic identification and diagnosis of kidney diseases, a novel DL model built on the transfer-learning (TL) technology is implemented in this work. DL models are designed to focus on problems, whereas TL uses the knowledge acquired while resolving one issue to another pertinent issue. The proposed model combines multiple DL models to improve overall performance by leveraging the strengths of different architectures, ensembles can enhance accuracy, robustness, and generalization. It enhances the features extracted from MobileNet-V2, ResNet50, and EfficientNet-B0 networks using metaheuristic algorithms and bidirectional long-short-term memory (Bi-LSTM) from the CT image. MobileNetV2, ResNet50, and EfficientNet-B0 hyperparameters have been optimized using a modified grey wolf optimization (GWO) approach for better performance. The suggested model’s performance has been measured using five assessment metrics: accuracy, sensitivity, specificity, precision, and area under the ROC curve (AUC) and achieved 99.85% accuracy, 99.8% sensitivity, 99.3% specificity, 98.1% precision, and 1.0 AUC.

BackgroundIt is often assumed that rare genetic variants will improve available risk prediction scores. We aimed to estimate the added predictive ability of rare variants for risk prediction of common diseases in hypothetical scenarios.MethodsIn simulated data, we constructed risk models with an area under the ROC curve (AUC) ranging between 0.50 and 0.95, to which we added a single variant representing the cumulative frequency and effect (odds ratio, OR) of multiple rare variants. The frequency of the rare variant ranged between 0.0001 and 0.01 and the OR between 2 and 10. We assessed the resulting AUC, increment in AUC, integrated discrimination improvement (IDI), net reclassification improvement (NRI(>0.01)) and categorical NRI. The analyses were illustrated by a simulation of atrial fibrillation risk prediction based on a published clinical risk model.ResultsWe observed minimal improvement in AUC with the addition of rare variants. All measures increased with the frequency and OR of the variant, but maximum increment in AUC remained below 0.05. Increment in AUC and NRI(>0.01) decreased with higher AUC of the baseline model, whereas IDI remained constant. In the atrial fibrillation example, the maximum increment in AUC was 0.02 for a variant with frequency = 0.01 and OR = 10. IDI and NRI showed at most minimal increase for variants with frequency greater than or equal to 0.005 and OR greater than or equal to 5.ConclusionsSince rare variants are present in only a minority of affected individuals, their predictive ability is generally low at the population level. To improve the predictive ability of clinical risk models for complex diseases, genetic variants must be common and have substantial effect on disease risk.

To evaluate whether the addition of gadolinium-enhanced MRI and diffusion-weighted imaging (DWI) improves T2 sequence performance for the diagnosis of local recurrence (LR) from rectal cancer and to assess which approach is better at formulating this diagnosis among readers with different experience. Forty-three patients with suspected LR underwent pelvic MRI with T2 weighted (T2) sequences, gadolinium fat-suppressed T1 weighted sequences (post-contrast T1), and DWI sequences. Three readers (expert: G, intermediate: E, resident: V) scored the likelihood of LR on T2, T2+post-contrast T1, T2+DWI, and T2+post-contrast T1+DWI. In total, 18/43 patients had LR; on T2 images, the expert reader achieved an area under the ROC curve (AUC) of 0.916, sensitivity of 88.9%, and specificity of 76%; the intermediate reader achieved values of 0.890, 88.9%, and 48%, respectively, and the resident achieved values of 0.852, 88.9%, and 48%, respectively. DWI significantly improved the AUC value for the expert radiologist by up to 0.999 (p=0.04), while post-contrast T1 significantly improved the AUC for the resident by up to 0.950 (p=0.04). For the intermediate reader, both the T2+DWI AUC and T2+post-contrast T1 AUC were better than the T2 AUC (0.976 and 0.980, respectively), but with no statistically significant difference. No statistically significant difference was achieved by any of the three readers by comparing either the T2+DWI AUCs to the T2+post-contrast T1 AUCs or the AUCs of the two pairs of sequences to those of the combined three sequences. Furthermore, using the T2 sequences alone, all of the readers achieved a fair number of "equivocal" cases: they decreased with the addition of either DWI or post-contrast T1 sequences and, for the two less experienced readers, they decreased even more with the three combined sequences. Both DWI and T1 post-contrast MRI increased diagnostic performance for LR diagnosis compared to T2; however, no significant difference was observed by comparing the two different pairs of sequences with the three combined sequences.

The area under the ROC curve (AUC) is a widely used performance measure in machine learning. Increasingly, however, in several applications, ranging from ranking to biometric screening to medicine, performance is measured not in terms of the full area under the ROC curve but in terms of the partial area under the ROC curve between two false-positive rates. In this letter, we develop support vector algorithms for directly optimizing the partial AUC between any two false-positive rates. Our methods are based on minimizing a suitable proxy or surrogate objective for the partial AUC error. In the case of the full AUC, one can readily construct and optimize convex surrogates by expressing the performance measure as a summation of pairwise terms. The partial AUC, on the other hand, does not admit such a simple decomposable structure, making it more challenging to design and optimize (tight) convex surrogates for this measure. Our approach builds on the structural SVM framework of Joachims ( 2005 ) to design convex surrogates for partial AUC and solves the resulting optimization problem using a cutting plane solver. Unlike the full AUC, where the combinatorial optimization needed in each iteration of the cutting plane solver can be decomposed and solved efficiently, the corresponding problem for the partial AUC is harder to decompose. One of our main contributions is a polynomial time algorithm for solving the combinatorial optimization problem associated with partial AUC. We also develop an approach for optimizing a tighter nonconvex hinge loss-based surrogate for the partial AUC using difference-of-convex programming. Our experiments on a variety of real-world and benchmark tasks confirm the efficacy of the proposed methods.

Deep learning analysis of breast MRIs for prediction of occult invasive disease in ductal carcinoma in situ

The rise of targeted advertising has led to frequent privacy data leaks, as advertisers are reluctant to share information to safeguard their interests. This has resulted in isolated data islands and model heterogeneity challenges. To address these issues, we have proposed a C-means clustering algorithm based on maximum average difference to improve the evaluation of the difference in distribution between local and global parameters. Additionally, we have introduced an innovative dynamic selection algorithm that leverages knowledge distillation and weight correction to reduce the impact of model heterogeneity. Our framework was tested on various datasets and its performance was evaluated using accuracy, loss, and AUC (area under the ROC curve) metrics. Results showed that the framework outperformed other models in terms of higher accuracy, lower loss, and better AUC while requiring the same computation time. Our research aims to provide a more reliable, controllable, and secure data sharing framework to enhance the efficiency and accuracy of targeted advertising.

ObjectiveThe presence of pancreaticobiliary maljunction (PBM) in pediatric patients with congenital choledochal malformation significantly impacts clinical management and surgical decision-making. Current preoperative evaluation of PBM coexistence remains challenging in children, while intraoperative cholangiography does not consistently provide diagnostic-quality imaging. This study aims to develop machine learning-based algorithm models for detecting pancreaticobiliary maljunction (PBM) in children with congenital choledochal malformation.MethodsWe conducted a retrospective study utilizing data from patients with congenital choledochal malformation treated at our center between January 2019 and January 2024. Demographic characteristics, clinical features, and preoperative laboratory parameters were processed through rigorous data curation and feature engineering pipelines. Cases were allocated via random sampling into training (80%) and hold-out test (20%) cohorts, maintaining strict separation between training and test cohorts. Seven machine learning algorithms - Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), Light Gradient Boosting Machine (LightGBM), and K-Nearest Neighbors (KNN) - were implemented with five-fold cross-validation. An ensemble voting classifier was specifically constructed using these models. Model performance was quantified through comprehensive metrics including area under the ROC curve (AUC), sensitivity, specificity, positive/negative predictive values, accuracy, precision, recall, and F1-score. This study employed the nonparametric bootstrap method to estimate the confidence interval for the area under the receiver operating characteristic curve (AUC). SHapley Additive exPlanations (SHAP) was employed for model interpretability, with feature importance rankings determined by absolute SHAP value magnitudes.ResultsIn a cohort of 803 pediatric patients with congenital choledochal malformation, 628 (78.2%) demonstrated concurrent pancreaticobiliary maljunction. We developed a detection model incorporating 43 clinical features, with Random Forest showing optimal performance. An ensemble voting classifier integrating seven machine learning algorithms achieved enhanced discriminative performance (AUC: 0.87 (0.81, 0.92); Recall: 0.91 (0.85, 0.95); F1-score: 0.91 (0.87, 0.94)). Key features contributing to PBM detection included: laboratory markers and clinical parameters.ConclusionBy integrating preoperative clinical symptoms and laboratory parameters, machine learning algorithms demonstrated significant detection capability in identifying PBM among pediatric congenital choledochal malformation patients, with the RF model achieving superior performance metrics among all base models. The developed ensemble voting classifier provides valuable preoperative guidance for surgical planning and clinical management, enabling detection of PBM comorbidity before surgery in congenital choledochal malformation cases.

Radiation Therapy Oncology Group (RTOG) Protocol 0915: A Randomized Phase 2 Study Comparing 2 Stereotactic Body Radiation Therapy (SBRT) Schedules for Medically Inoperable Patients With Stage I Peripheral Non-Small Cell Lung Cancer

Several indicators associated with poor outcomes in older persons have been developed, but a direct comparison of their accuracy is lacking. Knowing which indicator performs better in the prediction of specific outcomes could help health care providers to choose the most suitable one. We compared the accuracy in predicting different clinically-relevant outcomes of five indicators: frailty index (FI), frailty phenotype (FP), the Health Assessment Tool (HAT), walking speed (WS), and multimorbidity. Data from the Swedish National Study on Aging and Care in Kungsholmen, an ongoing population-based study including 3363 people 60+, were used. The ability of the five indicators to predict mortality (3- and 5-year), unplanned hospitalizations (1- and 3-year), and 2+ health provider contacts (6 months prior and after assessment) was compared using the area under the ROC curves (AUC). FI, WS, and HAT showed the best accuracy in the prediction of mortality (AUC for 3-year mortality: 0.84, 0.85, 0.87 respectively; AUC for 5-year mortality: 0.84, 0.85, 0.86 respectively; all p < 0.05). Unplanned hospitalizations were better predicted by the FI (AUC: 1-year 0.73; 3-year 0.72) and HAT (AUC: 1-year 0.73; 3-year 0.71).The most accurate predictor of multiple contacts with health providers was multimorbidity (AUC: 0.67; p < 0.05). All indicators, but multimorbidity, showed higher accuracy among older individuals (75+ years). Different indicators can be used to support physicians during their decision-making process. Some of these tools may also be used to forecast future use of health-care resources, including both hospital-based services and outpatient ones .

Highlights. The article presents a novel and unique method for assessment of left ventricular systolic disfunction using electrocardiography and photoplethysmography data. This method will improve and simplify the detection of cardiovascular diseases.Aim. To evaluate left ventricular (LV) systolic function using electrocardiogram (ECG) and photoplethysmogram (PPG) signals recorded by a single-channel ECG and PPG-based monitor.Methods. The prospective study included 489 patients over 18 years old with various cardiovascular diseases. All participants underwent echocardiography to determine the main indicators of LV systolic function: LV ejection fraction (EF), LV outflow tract velocity time integral (LVOT VTI), and global longitudinal strain (GLS). Moreover, all patients underwent 1-lead ECG and PPG recording using a single-channel ECG and PPG-based monitor (CardioQvark). The obtained data were analyzed, and ROC curve analysis was performed.Results. We have identified ECG and PPG parameters associated with a decrease in LV contractile function. During the analysis, the ECG, T-wave amplitude (TA) and RonsF parameters showed the highest diagnostic accuracy. With EF below 55%, the area under the ROC curve (AUC) was 0.822, sensitivity (Se) 80%, specificity (Sp) 69% in EF below 55% in TA; in RonsF AUC was 0.743, Se 81%, Sp 77%. With EF below 40%, AUC was 0.915, Se 85%, Sp 83% in TA, and in RonsF AUC was 0.844, Se 82%, Sp 82%. Diagnostic accuracy of ECG signals in case of LVOT VTI lower than 16 cm was measured: TA (AUC 0.755, Se 82%, Sp 70%), RonsF (AUC 0.620, Se 77%, Sp 72%). PPG signals were not significantly associated with reduced EF; however, the pulse wave parameters were associated with lower LVOT VTI: in DP-B0 AUC was 0.687, Se 71%, Sp 74%. The combination of ECG and PPG signals was significantly associated with EF below 40% (RonsF * DP-SEP (AUC 0.877, Se 86%, Sp 85%). ECG and PPG signals were not associated with LV GLS.Conclusion. Assessment of LV systolic function can be performed by analyzing ECG and PPG signals recorded using a portable single-channel CardioQvark monitor.