ROC AUC Research Articles

A Learning Approach for The Identification of Network Intrusions Based on Ensemble XGBoost Classifier

The limitation of signature-based Intrusion Detection systems has given rise for the popularity of Machine learning (ML) approaches r for building such intrusion detection systems (IDSs). ML is a sub-filed of Artificial Intelligence that enables algorithms to learn from data and its applications have been widely accepted and used in many domains. To achieve a promising ML-based model that can identify attacks and intrusions in networks and the cyber space, different stages of machine learning approach like pre-processing, attribute selection, model building, hyper parameter tuning can be very important. CICIDS2017 intrusion dataset was used for all the experimentations. This study focuses on building cyber threat detection model based on the ensemble feature selection and classification method. Innovative approaches were used for the analysis and pre-processing of the dataset. Thereafter, XGboost algorithm was used for selecting relevant features from the default input attributes in each of the captures. Thereafter, the reduced features were employed in the identification of cyber intrusions. The average accuracy achieved in the 8 captures of the dataset is 98% while precision is 0.98. Also, recall is 0.98, f1-score is 0.98 while AUC ROC score is 0.99. The study concluded that XGBoost-based model was able to achieve promising results based on the proper dataset encoding, feature importance-based feature selection and tuning of the algorithm for intrusion identification.

Acoustic COVID-19 Detection Using Multiple Instance Learning.

In the COVID-19 pandemic, a rigorous testing scheme was crucial. However, tests can be time-consuming and expensive. A machine learning-based diagnostic tool for audio recordings could enable widespread testing at low costs. In order to achieve comparability between such algorithms, the DiCOVA challenge was created. It is based on the Coswara dataset offering the recording categories cough, speech, breath and vowel phonation. Recording durations vary greatly, ranging from one second to over a minute. A base model is pre-trained on random, short time intervals. Subsequently, a Multiple Instance Learning (MIL) model based on self-attention is incorporated to make collective predictions for multiple time segments within each audio recording, taking advantage of longer durations. In order to compete in the fusion category of the DiCOVA challenge, we utilize a linear regression approach among other fusion methods to combine predictions from the most successful models associated with each sound modality. The application of the MIL approach significantly improves generalizability, leading to an AUC ROC score of 86.6% in the fusion category. By incorporating previously unused data, including the sound modality 'sustained vowel phonation' and patient metadata, we were able to significantly improve our previous results reaching a score of 92.2%.

Differential gene expression of blood-based ABCA9, CNOT8, SESN1, UCP3, MAP2K1 and DDIT4 in Alzheimer’s disease

This study uncovered differential gene expression in blood to distinguish subjects with probable Alzheimer’s disease (AD) from normal elderly participants (non-demented controls, NDC). The participants were recruited via training (Phase 1) and validation cohorts (Phase 2). The changes of gene expression in blood samples from the training cohort (92 AD vs 92 NDC) were assessed using the microarray technology. The Partial Least Square Discrimination Analysis (PLSDA) was then used to develop a disease classifier algorithm (accuracy = 88.3%). Six differentially expressed genes were validated through RT-qPCR using blood samples from the validation cohort [(25 AD, 25 NDC, 12 mild cognitive impairment (MCI) and 12 vascular dementia (VaD) subjects] . The PLSDA model indicated a good separation between AD and NDC [area under the receiver operating characteristic curve (ROC AUC) = 0.88]. ABCA9, CNOT8, SESN1, UCP3, MAP2K1 and DDIT4 were found to be differentially expressed between the two groups. Validation of the panel of six genes gave an overall accuracy of 82.0% (AUC=0.86). The ABCA9 mRNA level, which was significantly (p < 0.05) lower in the AD group, correctly classified 90.9% of all subjects (AUC=0.94). This group of genes may be responsible for dysregulation of pathways related to inflammation, mitochondrial dysfunction, oxidative injury, DNA damage, apoptosis and lipid metabolism. The disease classifier algorithm discriminated probable AD from MCI and VaD at specificity of 83.3% and 75.0%, respectively. These findings warrant further validation of potential blood-based biomarkers in larger samples of clinical AD.

Sparse Representations Optimization with Coupled Bayesian Dictionary and Dictionary Classifier for Efficient Classification

Among the numerous techniques followed to learn a linear classifier through the discriminative dictionary and sparse representations learning of signals, the techniques to learn a nonparametric Bayesian classifier jointly and discriminately with the dictionary and the corresponding sparse representations have drawn considerable attention from researchers. These techniques jointly learn two sets of sparse representations, one for the training samples over the dictionary and the other for the corresponding labels over the dictionary classifier. At the prediction stage, the representations of the test samples computed over the learned dictionary do not truly represent the corresponding labels, exposing weakness in the joint learning claim of these techniques. We mitigate this problem and strengthen the joint by learning a set of weights over the dictionary to represent the training data and further optimizing the same weights over the dictionary classifier to represent the labels of the corresponding classes of the training data. Now, at the prediction stage, the representation weights of the test samples computed over the learned dictionary also represent the labels of the corresponding classes of the test samples, resulting in the accurate reconstruction of the labels of the classes by the learned dictionary classifier. Overall, a reduction in the size of the Bayesian model’s parameters also improves training time. We analytically and nonparametrically derived the posterior conditional probabilities of the model from the overall joint probability of the model using Bayes’ theorem. We used the Gibbs sampler to solve the joint probability of the model using the derived conditional probabilities, which also supports our claim of efficient optimization of the coupled/joint dictionaries and the sparse representation parameters. We demonstrated the effectiveness of our approach through experiments on the standard datasets, i.e., the Extended YaleB and AR face databases for face recognition, Caltech-101 and Fifteen Scene Category databases for categorization, and UCF sports action database for action recognition. We compared the results with the state-of-the-art methods in the area. The classification accuracies, i.e., 93.25%, 89.27%, 94.81%, 98.10%, and 95.00%, of our approach on the datasets have increases of 0.5 to 2% on average. The overall average error margin of the confidence intervals in our approach is 0.24 compared with the second-best approach, JBDC, for which it is 0.34. The AUC–ROC scores of our approach are 0.98 and 0.992, which are better than those of others, i.e., 0.960 and 0.98, respectively. Our approach is also computationally efficient.

Assessment of a deep learning model for COVID-19 classification on chest radiographs: a comparison across image acquisition techniques and clinical factors.

The purpose is to assess the performance of a pre-trained deep learning model in the task of classifying between coronavirus disease (COVID)-positive and COVID-negative patients from chest radiographs (CXRs) while considering various image acquisition parameters, clinical factors, and patient demographics. Standard and soft-tissue CXRs of 9860 patients comprised the "original dataset," consisting of training and test sets and were used to train a DenseNet-121 architecture model to classify COVID-19 using three classification algorithms: standard, soft tissue, and a combination of both types of images via feature fusion. A larger more-current test set of 5893 patients (the "current test set") was used to assess the performance of the pretrained model. The current test set contained a larger span of dates, incorporated different variants of the virus and included different immunization statuses. Model performance between the original and current test sets was evaluated using area under the receiver operating characteristic curve (ROC AUC) [95% CI]. The model achieved AUC values of 0.67 [0.65, 0.70] for cropped standard images, 0.65 [0.63, 0.67] for cropped soft-tissue images, and 0.67 [0.65, 0.69] for both types of cropped images. These were all significantly lower than the performance of the model on the original test set. Investigations regarding matching the acquisition dates between the test sets (i.e., controlling for virus variants), immunization status, disease severity, and age and sex distributions did not fully explain the discrepancy in performance. Several relevant factors were considered to determine whether differences existed in the test sets, including time period of image acquisition, vaccination status, and disease severity. The lower performance on the current test set may have occurred due to model overfitting and a lack of generalizability.

RECOVERY PROGNOSIS OF ASPECTS IN PATIENTS WITH ACUTE ANTERIOR CIRCULATION CEREBRAL INFARCTION UNDERGOING MECHANICAL THROMBECTOMY

Aim: Comparison of the prognostic value of recovery of the ASPECTS score compared with perfusion imaging in patients with acute anterior circulation cerebral infarction undergoing thrombectomy.
 Materials and methods: Retrospective cross-sectional descriptive study combined with a prospective study of 82 cases undergoing CT scan without injection, CTA, and CTP, using RAPID software to calculate core and penumbra values and mechanical thrombectomy within 6-24 hours. Good (0-2 points) and poor (3-6 points) stroke outcomes were assessed at 3 months by the modified Rankin functional recovery score (mRS). Use the ROC curve and AUC value to compare the good and bad stroke prognostic value of ASPECTS and the core, penumbra, mismatch ratio values.
 Results: From January 2022 to May 2023, 82 patients were included in the study. At 90 days, the independent outcome rate of good mRS function (0-2) was 57.3%, and poor mRS function (3-6) was 42.7%. The ROC curve showed that, among the ASPECTS, core, penumbra, and mismatch ratio indices, only ASPECTS could predict stroke outcome well (p = 0.001<0.05), the area under the curve was 0.709 with J - point cutoff index at threshold ASPECTS = 6 with sensitivity 93.6% and specificity 57.1%.
 Conclusion: ASPECTS ≥ 6 can predict good functional recovery after stroke in patients with acute anterior circulation cerebral infarction undergoing thrombectomy at the extended window.

A multiclass extreme gradient boosting model for evaluation of transcriptomic biomarkers in Alzheimer’s disease prediction

BackgroundPatients with young-onset Alzheimer's disease (AD) (before the age of 50 years old) often lack obvious imaging changes and amyloid protein deposition, which can lead to misdiagnosis with other cognitive impairments. Considering the association between immunological dysfunction and progression of neurodegenerative disease, recent research has focused on identifying blood transcriptomic signatures for precise prediction of AD. MethodsIn this study, we extracted blood biomarkers from large-scale transcriptomics to construct multiclass eXtreme Gradient Boosting models (XGBoost), and evaluated their performance in distinguishing AD from cognitive normal (CN) and mild cognitive impairment (MCI). ResultsIndependent testing with external dataset revealed that the combination of blood transcriptomic signatures achieved an area under the receiver operating characteristic curve (AUC of ROC) of 0.81 for multiclass classification (sensitivity = 0.81; specificity = 0.63), 0.83 for classification of AD vs. CN (sensitivity = 0.72; specificity = 0.73), and 0.85 for classification of AD vs. MCI (sensitivity = 0.77; specificity = 0.73). These candidate signatures were significantly enriched in 62 chromosome regions, such as Chr.19p12-19p13.3, Chr.1p22.1-1p31.1, and Chr.1q21.2-1p23.1 (adjusted p < 0.05), and significantly overrepresented by 26 transcription factors, including E2F2, FOXO3, and GATA1 (adjusted p < 0.05). Biological analysis of these signatures pointed to systemic dysregulation of immune responses, hematopoiesis, exocytosis, and neuronal support in neurodegenerative disease (adjusted p < 0.05). ConclusionsBlood transcriptomic biomarkers hold great promise in clinical use for the accurate assessment and prediction of AD.

Peptidomics Unveils Distinct Acetylation Patterns of Histone and Annexin A1 in Differentiated Thyroid Cancer.

Thyroid cancer is a common malignancy of the endocrine system. Nodules are routinely evaluated for malignancy risk by fine needle aspiration biopsy (FNAB), and in cases such as follicular lesions, differential diagnosis between benign and malignant nodules is highly uncertain. Therefore, the discovery of new biomarkers for this disease could be helpful in improving diagnostic accuracy. Thyroid nodule biopsies were subjected to a precipitation step with both the insoluble and supernatant fractions subjected to proteome and peptidome profiling. Proteomic analysis identified annexin A1 as a potential biomarker of thyroid cancer malignancy, with its levels increased in malignant samples. Also upregulated were the acetylated peptides of annexin A1, revealed by the peptidome analysis of the supernatant fraction. In addition, supernatant peptidomic analysis revealed a number of acetylated histone peptides that were significantly elevated in the malignant group, suggesting higher gene transcription activity in malignant tissue. Two of these peptides were found to be robust malignancy predictors, with an area under the receiver operating a characteristic curve (ROC AUC) above 0.95. Thus, this combination of proteomics and peptidomics analyses improved the detection of malignant lesions and also provided new evidence linking thyroid cancer development to heightened transcription activity. This study demonstrates the importance of peptidomic profiling in complementing traditional proteomics approaches.

Investigation of radiomics models for predicting biochemical recurrence of advanced prostate cancer on pretreatment MR ADC maps based on automatic image segmentation.

To develop radiomics models based on automatic segmentation of the pretreatment apparent diffusion coefficient (ADC) maps for predicting the biochemical recurrence (BCR) of advanced prostate cancer (PCa). A total of 100 cases with pathologically confirmed PCa were retrospectively included in this study. These cases were randomly divided into training (n=70) and test (n=30) datasets. Two predictive models were constructed based on the combination of age, prostate specific antigen (PSA) level, Gleason score, and clinical staging before therapy and the prostate area (Model_1) or PCa area (Model_2). Another two predictive models were constructed based on only prostate area (Model_3) or PCa area (Model_4). The area under the receiver operating characteristic curve (ROC AUC) and precision-recall (PR) curve analysis were used to analyze the models' performance. Sixty-five patients without BCR (BCR-) and 35 patients with BCR (BCR+) were confirmed. The age, PSA, volume, diameter and ADC value of the prostate and PCa were not significantly different between the BCR- and BCR+ groups or between the training and test datasets (all p>0.05). The AUCs were 0.637 (95% CI: 0.434-0.838), 0.841 (95% CI: 0.695-0.940), 0.840 (95% CI: 0.698-0.983), and 0.808 (95% CI: 0.627-0.988) for Model_1 to Model_4 in the test dataset without significant difference. The 95% bootstrap confidence intervals for the areas under the PR curve of the four models were not statistically different. The radiomics models based on automatically segmented prostate and PCa areas on the pretreatment ADC maps developed in our study can be promising in predicting BCR of advanced PCa.

NEW BIOLOGICAL MARKERS FOR A PROGNOSTIC MODEL FOR ASSESSING THE RISK OF CARDIAC FIBROSIS IN PATIENTS WITH ST-SEGMENT ELEVATION MYOCARDIAL INFARCTION

HighlightsThe developed prognostic model for assessing the risk of cardiac fibrosis in patients with STEMI with HFmrEF and HFpEF is promising from the point of view of scientific and clinical potential because similar models for predicting the risk of cardiac fibrosis in patients with index MI are not currently validated. The developed scale includes such parameters as age, LVEF, COL-1, BMI, MMP-2. The scale can be used in patients with HFmrEF and HFpEF phenotypes. Identification of patients at high risk of myocardial fibrosis will allow choosing the appropriate treatment method. Aim. To develop a prognostic model for assessing the risk of cardiac fibrosis (CF) in patients with preserved left ventricular ejection fraction (HFpEF) and mildly reduced ejection fraction (HFmrEF) a year after ST-segment elevation myocardial infarction (STEMI) based on clinical, instrumental and biochemical data.Methods. The prospective cohort study included 100 STEMI patients with HFmrEF (LVEF 40–49%) and with HFpEF (50% or more). Echo was performed in all patients on the 1st, 10–12th day and a year after onset of STEMI. Upon admission to the hospital and on the 10–12th day after the onset of the disease, the following serum biomarker levels were determined: those associated with changes in the extracellular matrix; with remodeling and fibrosis; with inflammation, and with neurohormonal activation. At the 1-year follow-up visit, 84 patients underwent contrast-enhanced MRI to assess fibrotic tissue percentage relative to healthy myocardium.Results. The distribution of patients by HFmrEF and HFpEF phenotypes during follow-up was as follows: HFmrEF on the 1st day – 27%, 10th day – 12%, after a year – 11%; HFpEF on the 1st day – 73%, 10th day – 88%, after a year – 89%. According to cardiac MRI at the follow-up visit (n = 84), the median distribution of fibrotic tissue percentage was 5 [1.5; 14]%. Subsequently, the threshold value of 5% was chosen for analysis: CF≥5% was found in 38 patients (the 1st group), whereas CF&lt;5% was noted in 46 patients (the 2nd group). When analyzing the intergroup differences in biological marker concentrations in the in-patient setting and at the annual follow-up, it was determined that the most significant differences were associated with “ST-2” (1st day) that in the “CF≥5%” group was 11.4 ng/mL higher on average compared to the “CF&lt;5%” group (p = 0.0422); “COL-1” (1st day) that in the “CF≥5%” group was 28112.3 pg/mL higher on average compared to the “CF&lt;5%” group (p = 0.0020), and “NT-proBNP” (12th day) that in the “CF&lt;5 %” group was 1.9 fmol/mL higher on average compared to the “CF≥5%” group (p = 0.0339). Certain factors (age, LVEF (12th day), collagen-1 (1st and 12th day), body mass index, matrix metalloproteinase-2 (12th day) were determined and included in the prognostic model for assessing the risk of CF a year after the STEMI (AUC ROC 0.90, Chi-square test &lt;0.0001).Conclusion. Prognostic model (scale) based on factors such as age, left ventricular ejection fraction (12th day), collagen-1 (1st and 12th day), body mass index, matrix metalloproteinase-2 (12th day) shows high prognostic power and enables identification of patients with HFmrEF and HFpEF phenotypes and at high risk of cardiac fibrosis a year after STEMI.

Smartphone-based gait assessment for multiple sclerosis

IntroductionMultiple Sclerosis causes gait alteration, even in the early stages of the disease. Traditional methods to quantify gait impairment, such as performance-based measures, lab-based motion analyses, and self-report, have limited ecological relevance. The Mon4t® app is a digital tool that uses sensors embedded in standard smartphones to measure various gait parameters. ObjectivesTo evaluate the use of Mon4t® technology in monitoring MS patients. Methods100 MS patients and age-matched healthy controls were evaluated using both a human rater and the Mon4t Clinic™ app. Three motor tasks were performed: 3m Timed up and go test (TUG), 10m TUG, and tandem walk. The digital markers were used to compare MS vs. HC, MS with EDSS=0 vs. HC, and MS with EDSS=0 vs. MS with EDSS>0. Within the MS EDSS>0 group, correlations between digital gait markers and the EDSS score were calculated. ResultsSignificant differences were found between MS patients and HC in multiple gait parameters. When comparing MS patients with minimal disability (EDSS=0) and HC: On the 3m TUG task, MS patients took longer to complete the task (mean difference 0.167seconds, p =0.034), took more steps (mean difference 1.32 steps, p =0.003), and had a weaker ML step-to-step correlation (mean difference 0.1, p = 0.001). The combination of features from the three motor tasks allowed distinguishing a nondisabled MS patient from a HC with high confidence (AUC of 85.65 on the ROC). When comparing MS patients with minimal disability (EDSS=0) to those with higher disability (EDSS>0): On the tandem walk task, patients with EDSS>0 took significantly longer to complete 10 steps than those with EDSS=0 (mean difference 4.63 seconds, p < 0.001), showed greater ML sway (mean difference 0.2, p < 0.001), and had larger angular velocity in the SI axis on average (mean difference 2.31 degrees/sec, p = 0.01). A classification model achieved 81.79 ROC AUC. In the subgroup of patients with EDSS>0, gait features significantly correlated with EDSS score in all three tasks. ConclusionThe findings demonstrate the potential of digital gait assessment to augment traditional disease monitoring and support clinical decision making. The Mon4t® app provides a convenient and ecologically relevant tool for monitoring MS patients and detecting early changes in gait impairment.

Glaucoma Prediction Models Based on Ocular and Systemic Findings

Introduction: Our aim was to explore the impact of various systemic and ocular findings on predicting the development of glaucoma. Methods: Medical records of 37,692 consecutive patients examined at a single medical center between 2001 and 2020 were analyzed using machine learning algorithms. Systemic and ocular features were included. Univariate and multivariate analyses followed by CatBoost and Light gradient-boosting machine prediction models were performed. Main outcome measures were systemic and ocular features associated with progression to glaucoma. Results: A total of 7,880 patients (mean age 54.7 ± 12.6 years, 5,520 males [70.1%]) were included in a 3-year prediction model, and 314 patients (3.98%) had a final diagnosis of glaucoma. The combined model included 185 systemic and 42 ocular findings, and reached an ROC AUC of 0.84. The associated features were intraocular pressure (48.6%), cup-to-disk ratio (22.7%), age (8.6%), mean corpuscular volume (MCV) of red blood cell trend (5.2%), urinary system disease (3.3%), MCV (2.6%), creatinine level trend (2.1%), monocyte count trend (1.7%), ergometry metabolic equivalent task score (1.7%), dyslipidemia duration (1.6%), prostate-specific antigen level (1.2%), and musculoskeletal disease duration (0.5%). The ocular prediction model reached an ROC AUC of 0.86. Additional features included were age-related macular degeneration (10.0%), anterior capsular cataract (3.3%), visual acuity (2.0%), and peripapillary atrophy (1.3%). Conclusions: Ocular and combined systemic-ocular models can strongly predict the development of glaucoma in the forthcoming 3 years. Novel progression indicators may include anterior subcapsular cataracts, urinary disorders, and complete blood test results (mainly increased MCV and monocyte count).

Predicting movies’ eudaimonic and hedonic scores: A machine learning approach using metadata, audio and visual features

In the task of modeling user preferences for movie recommender systems, recent research has demonstrated the benefits of describing movies with their eudaimonic and hedonic scores (E and H scores), which reflect the depth of their message and the level of fun experience they provide, respectively. So far, the labeling of movies with their E and H scores has been done manually using a dedicated instrument (a questionnaire), which is time-consuming. To address this issue, we propose an automatic approach for predicting E and H scores. Specifically, we collected E and H scores of 709 movies from 370 users (with a total of 3699 records), augmented this dataset with metadata, audio, and low-level and high-level visual features, and trained machine learning models for predicting the E and H scores of movies. This study investigates the use of machine learning models in predicting the E and H scores of movies using various feature sets, including audio, low-level and high-level visual features, and metadata. We compared the performance of predictive models using different combinations of features with the majority classifier as the baseline approach. The results demonstrate that our proposed machine learning-based models significantly outperform the baseline in predicting E and H scores, particularly when leveraging metadata features. Specifically, the random forest classifier achieved a 20% increase in ROC AUC compared to the baseline when predicting both the E score and the H score. These improvements were found to be statistically significant. Overall, our findings suggest that automated tools for predicting E and H scores in movies are promising alternatives to traditional questionnaire-based approaches.

Prediction of brain sex from EEG: using large-scale heterogeneous dataset for developing a highly accurate and interpretable ML model

This study presents a comprehensive examination of sex-related differences in resting-state electroencephalogram (EEG) data, leveraging two different types of machine learning models to predict an individual's sex. We utilized data from the Two Decades-Brainclinics Research Archive for Insights in Neurophysiology (TDBRAIN) EEG study, affirming that gender prediction can be attained with noteworthy accuracy. The best performing model achieved an accuracy of 85% and an ROC AUC of 89%, surpassing all prior benchmarks set using EEG data and rivaling the top-tier results derived from fMRI studies. A comparative analysis of LightGBM and Deep Convolutional Neural Network (DCNN) models revealed DCNN's superior performance, attributed to its ability to learn complex spatial-temporal patterns in the EEG data and handle large volumes of data effectively. Despite this, interpretability remained a challenge for the DCNN model. The LightGBM interpretability analysis revealed that the most important EEG features for accurate sex prediction were related to left fronto-central and parietal EEG connectivity. We also showed the role of both low (delta and theta) and high (beta and gamma) activity in the accurate sex prediction. These results, however, have to be approached with caution, because it was obtained from a dataset comprised largely of participants with various mental health conditions, which limits the generalizability of the results and necessitates further validation in future studies. . Overall, the study illuminates the potential of interpretable machine learning for sex prediction, alongside highlighting the importance of considering individual differences in prediction sex from brain activity.

Visual Features for Improving Endoscopic Bleeding Detection Using Convolutional Neural Networks.

The presented paper investigates the problem of endoscopic bleeding detection in endoscopic videos in the form of a binary image classification task. A set of definitions of high-level visual features of endoscopic bleeding is introduced, which incorporates domain knowledge from the field. The high-level features are coupled with respective feature descriptors, enabling automatic capture of the features using image processing methods. Each of the proposed feature descriptors outputs a feature activation map in the form of a grayscale image. Acquired feature maps can be appended in a straightforward way to the original color channels of the input image and passed to the input of a convolutional neural network during the training and inference steps. An experimental evaluation is conducted to compare the classification ROC AUC of feature-extended convolutional neural network models with baseline models using regular color image inputs. The advantage of feature-extended models is demonstrated for the Resnet and VGG convolutional neural network architectures.

Improving nonalcoholic fatty liver disease classification performance with latent diffusion models

Integrating deep learning with clinical expertise holds great potential for addressing healthcare challenges and empowering medical professionals with improved diagnostic tools. However, the need for annotated medical images is often an obstacle to leveraging the full power of machine learning models. Our research demonstrates that by combining synthetic images, generated using diffusion models, with real images, we can enhance nonalcoholic fatty liver disease (NAFLD) classification performance even in low-data regime settings. We evaluate the quality of the synthetic images by comparing two metrics: Inception Score (IS) and Fréchet Inception Distance (FID), computed on diffusion- and generative adversarial network (GAN)-generated images. Our results show superior performance for the diffusion-generated images, with a maximum IS score of 1.90 compared to 1.67 for GANs, and a minimum FID score of 69.45 compared to 100.05 for GANs. Utilizing a partially frozen CNN backbone (EfficientNet v1), our synthetic augmentation method achieves a maximum image-level ROC AUC of 0.904 on a NAFLD prediction task.

Evaluation of optimum classification measures used to define textbook outcome among patients undergoing curative-intent resection of gastric cancer

BackgroundTextbook outcome (TO) is a composite measure reflecting various aspects of services provided to patients with solid malignancies. We sought to evaluate the importance of various TO components previously proposed for gastric cancer.MethodsProspectively maintained electronic databases of 1,743 patients treated in two academic surgical centres were reviewed. Six candidate definitions of TO were evaluated based on their ability to accurately predict patients’ prognosis by Cox proportional hazards modelling.ResultsTO definition combining 10 measures corresponding to complete tumour resection with an uneventful postoperative course showed the best goodness of fit by achieving the lowest values of Akaike (AIC) and Bayesian (BIC) information criteria and the best predictive performance based on the highest value of c-index. The overall median survival was significantly longer for patients with than without textbook outcome (69.0 vs 20.1 months, P < 0.001). TO maintained its prognostic value in a multivariate model controlling for age, sex, comorbidities, treatment, and tumour related variables and was associated with a 39% lower risk of death (HR 0.61, 95%CI 0.51 – 0.73, P < 0.001). Nine variables identified as predictors of TO were used to develop a nomogram showing very good correlation between the predicted and actual probability of achieving TO. The AUC of ROC obtained from the nomogram was 0.752 (95% CI 0.727 to 0.781).ConclusionsA uniform definition of textbook outcome provides clinically relevant prognostic information and could be used in quality improvement programs for gastric cancer patients.

Influenza A infections: predictors of disease severity.

Influenza affects approximately 10% of the world's population annually. It is associated with high morbidity and mortality rates due to its propensity to progress to severe acute respiratory infection, leading to 10-40% of hospitalized patients needing intensive care. Characterizing the multifactorial predictors of poor prognosis is essential for developing strategies against this disease. This study aimed to identify predictors of disease severity in influenza A-infected (IFA-infected) patients and to propose a prognostic score. A retrospective cross-sectional study was conducted with 142 IFA-infected out- and inpatients treated at a tertiary hospital between 2010 and 2018. The viral subtypes, hemagglutinin mutations, viral load, IL-28B SNPs, and clinical risk factors were evaluated according to the patient's ICU admission. Multivariate analysis identified the following risk factors for disease severity: neuromuscular diseases (OR = 7.02; 95% CI = 1.18-41.75; p = 0.032), cardiovascular diseases (OR = 5.47; 95% CI = 1.96-15.27; p = 0.001), subtype (H1N1) pdm09 infection (OR = 2.29; 95% CI = 1.02-5.15; p = 0.046), and viral load (OR = 1.43; 95% CI = 1.09-1.88; p = 0.009). The prognosis score for ICU admission is based on these predictors of severity presented and ROC curve AUC = 0.812 (p < 0.0001). Our results identified viral and host predictors of disease severity in IFA-infected patients, yielding a prognostic score that had a high performance in predicting the IFA patients' ICU admission and better results than a viral load value alone. However, its implementation in health services needs to be validated in a broader population.

Performance of severity indices for admission and mortality of trauma patients in the intensive care unit: a retrospective cohort study

BackgroundLittle is known about the performance of severity indices for indicating intensive care and predicting mortality in the Intensive Care Unit (ICU) of trauma patients. This study aimed to compare the performance of severity indices to predict trauma patients’ ICU admission and mortality.MethodsA retrospective cohort study which analyzed the electronic medical records of trauma patients aged ≥ 18 years, treated at a hospital in Brazil, between 2014 and 2017. Physiological [Revised Trauma Score (RTS), New Trauma Score (NTS) and modified Rapid Emergency Medicine Score (mREMS)], anatomical [Injury Severity Score (ISS) and New Injury Severity Score (NISS)] and mixed indices [Trauma and Injury Severity Score (TRISS), New Trauma and Injury Severity Score (NTRISS), Base-deficit Injury Severity Score (BISS) and Base-deficit and New Injury Severity Score (BNISS)] were compared in analyzing the outcomes (ICU admission and mortality) using the Area Under the Receiver Operating Characteristics Curves (AUC–ROC).ResultsFrom the 747 trauma patients analyzed (52.5% female; mean age 51.5 years; 36.1% falls), 106 (14.2%) were admitted to the ICU and 6 (0.8%) died in the unit. The ISS (AUC 0.919) and NISS (AUC 0.916) had better predictive capacity for ICU admission of trauma patients. The NISS (AUC 0.949), TRISS (AUC 0.909), NTRISS (AUC 0.967), BISS (AUC 0.902) and BNISS (AUC 0.976) showed excellent performance in predicting ICU mortality.ConclusionsAnatomical indices showed excellent predictive ability for admission of trauma patients to the ICU. The NISS and the mixed indices had the best performances regarding mortality in the ICU.

Radiomic Analysis for Human Papillomavirus Assessment in Oropharyngeal Carcinoma: Lessons and Pitfalls for the Next Future

Background: Oropharyngeal Squamous Cell Carcinoma (OPSCC) is rapidly increasing due to the spread of Human Papillomavirus (HPV) infection. HPV-positive disease has unique characteristics, with better response to treatment and consequent better prognosis. HPV status is routinely assessed via p16 immunohistochemistry or HPV DNA Polymerase Chain Reaction. Radiomics is a quantitative approach to medical imaging which can overcome limitations due to its subjective interpretation and correlation with clinical data. The aim of this narrative review is to evaluate the impact of radiomic features on assessing HPV status in OPSCC patients. Methods: A narrative review was performed by synthesizing literature results from PUBMED. In the search strategy, Medical Subject Headings (MeSH) terms were used. Retrospective mono- or multicentric works assessing the correlation between radiomic features and HPV status prediction in OPSCC were included. Selected papers were in English and included studies on humans. The range of publication date was July 2015–April 2023. Results: Our research returned 23 published papers; the accuracy of radiomic models was evaluated by ROC curves and AUC values. MRI- and CT-based radiomic models proved of comparable efficacy. Also, metabolic imaging showed crucial importance in the determination of HPV status, albeit with lower AUC values. Conclusions: Radiomic features from conventional imaging can play a complementary role in the assessment of HPV status in OPSCC. Both primary tumor- and nodal-related features and multisequencing-based models demonstrated higher accuracy.