AUC Of ROC Curve Research Articles

Enhancing Diagnostic Accuracy of Fresh Vertebral Compression Fractures with Deep Learning Models.

Retrospective study. The study aimed to develop and authenticated a deep learning model based on X-ray images to accurately diagnose fresh thoracolumbar vertebral compression fractures. In clinical practice, diagnosing fresh vertebral compression fractures often requires MRI. However, due to the scarcity of MRI resources and the high time and economic costs involved, some patients may not receive timely diagnosis and treatment. Using a deep learning model combined with X-rays for diagnostic assistance could potentially serve as an alternative to MRI. In this study, the main collection included X-ray images suspected of thoracolumbar vertebral compression fractures from the municipal shared database between December 2012 and February 2024. Deep learning models were constructed using frameworks of EfficientNet, MobileNet, and MnasNet respectively. We conducted a preliminary evaluation of the deep learning model using the validation set. The diagnostic performance of the models was evaluated using metrics such as AUC value, accuracy, sensitivity, specificity, F1 score, precision, and ROC curve. Finally, the deep learning models were compared with evaluations from two spine surgeons of different experience levels on the control set. This study included a total of 3025 lateral X-ray images from 2224 patients. The dataset was divided into a training set of 2388 cases, a validation set of 482 cases, and a control set of 155 cases. In the validation set, the three groups of DL models had accuracies of 83.0%, 82.4%, and 82.2%, respectively. The AUC values were 0.861, 0.852, and 0.865, respectively. In the control set, the accuracies of the three groups of DL models were 78.1%, 78.1%, and 80.7%, respectively, all higher than spinal surgeons and significantly higher than junior spine surgeon. This study developed deep learning models for detecting fresh vertebral compression fractures, demonstrating high accuracy.

3D modelling-based left atrium geometry analysis to predict atrial fibrillation in patients with cryptogenic stroke

Abstract Introduction Atrial fibrillation (AF) is a high prevalence condition associated with high morbidity and mortality. Its early diagnosis allows to initiate several prophylactic treatments that have proved to improve its prognosis. 3D modeling techniques allow to design accurate representations of cardiac cavities from diverse source images. Purpose We evaluated the association between new geometrical measurements obtained by 3D echocardiographic models of the left atrium (LA) and the presence of underlying paroxysmal AF in patients with cryptogenic stroke (CS). Methods 3D LA volumes from a sample of consecutive patients aged ≥60 years old and admitted to our Stroke Unit with the diagnosis of CS from April 2019 to November 2020 were analyzed. 15 days ECG monitoring was performed after discharge to detect underlying paroxysmal AF. All 3D echo volumes were analyzed by a 3D modeling open-source application. From this analysis several LA geometry parameters were obtained: sphericity, elongation, flatness, major axis length, minor axis length, maximum 2D diameter (max2Ddiam) and surface-volume ratio (S/V). Functional atrial parameters (3D left atrial ejection fraction, LAEF) were also determined. Results 36 patients were analyzed. Mean age was 78 ± 7 years old. 19% were diabetic, 58% had high blood pressure and 14% of patients had history of ischemic cardiomyopathy. Mean body-mass index was 26. These baseline characteristics were equally distributed between groups. Baseline proBNP was significantly higher in patients developing atrial fibrillation (949 vs 275 pg/mL; p= 0,014.). Maximum 2D diameter (max2Ddiam) and Surface-volume ratio (S/V) were the best predictors of AF (AUC of ROC curves: 0.69; 0.49 – 0.88 and 0.63; 0.42 – 0.84, respectively). An optimal cut-off point of 39 mm was selected for max2Ddiam and of 0.20 for S/V. These values were included in a multivariate model adjusted by baseline proBNP. A Max2Ddiam <39 mm and a S/V ≥0.20 were both significantly associated with the presence of paroxysmal AF (OR 6.96; p= 0.048 and OR 12.22; p = 0.043, respectively). Conclusions New geometrical atrial parameters can be obtained from echo 3D models and might be able to predict AF in patients with CS. Further analysis and larger samples are required to better understand these findings.TableFigure

Artificial intelligence to assist the echocardiographic identification of transthyretin cardiac amyloidosis

Abstract Introduction Transthyretin cardiac amyloidosis (ATTR-CM) is a severe disease, and treatments are most effective when initiated at early stage. Echocardiography is the main tool for screening patients with suspected ATTR-CM or populations at high risk of ATTR-CM. Suspicion of ATTR-CM triggers a dedicated confirmatory evaluation. Echocardiographic diagnosis is however difficult to implement in a non-expert cardiologist workflow. An algorithm to assist the identification of potential cases would be helpful to assist clinicians. Objective To develop an algorithm using artificial intelligence (AI) to assist the identification of potential ATTR-CM cases using routine transthoracic echocardiography. Methods In this non-interventional, monocentric case-control study, a dedicated database, comprising confirmed ATTR-CM patients and a control group, was built. We trained and validated a deep learning model based on Convolutional Neural Network from transthoracic echocardiography video loops. Results A total of 4394 patients were included in the database: 208 in the ATTR-CM group (76% males, mean age 69±10 years old, 90% hereditary ATTR-CM with NAC I score in 77%) and 4186 in the control group (59% male, mean age 62±16 years). After training the model to identify potential ATTR-CM, the ROC curve AUC was 0.97, accuracy was 94.7%, precision was 87.3%, and recall 84.4%. After selecting a subgroup of 130 pairs of matched ATTR-CM/controls based on age, sex, and wall thickness, the algorithm performance remained robust (AUC 0.99, accuracy 99.4%, precision 100%, recall 98.3%). Conclusion and Perspectives: Diagnosis of ATTR-CM by echocardiography can be assisted by an algorithm using AI. The algorithm remains effective in distinguishing ATTR-CM from other causes of left ventricular hypertrophy. External validation of this tool is underway.ROC curveAssisted classification

Cutting-edge approaches for judging surface water dynamics in semi-arid environments: Integrating landsat 8 OLI/TIRS and HYDROSAM model

This research addresses the critical need to assess surface water dynamics in semi-arid regions of Andhra Pradesh, India, by using advanced Spectral Indices for hydrological applications and methodologies. It aims to answer how remote sensing data from Landsat 8 OLI/TIRS can be effectively used to monitor and classify surface water bodies. The study applies indices such as Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Normalized Difference Thermal Index (NDTI), and Water Ratio Index (WRI). Additionally, the Weighted Composite Index (WCI) and Normalized Composite Index (NCI) are integrated with Principal Component Analysis (PCA) for multi-criteria decision making. The HYDROSAM model successfully classifies and maps various categories of surface water bodies, including non-water features (53.62%), urban water zones (37.89%), seasonal water bodies (4.32%), transitional zones (2.17%), permanent water bodies (1.05%), and river bodies (0.95%). The resultant map was validated using the AUC-ROC curve, achieving an AUC of 0.820, indicating a high level of accuracy. This methodology provides a nuanced understanding of water resource distribution and availability in the region. The findings demonstrate the robustness and reliability of the HYDROSAM model in accurately assessing surface water characteristics, thereby providing critical insights for informed water resource management, strategic land-use planning, and effective ecological conservation. This innovative methodology not only fosters sustainable water resource management in semi-arid regions but also sets a precedent for advancing research in similar ecosystems globally.

9303 Evaluation of The Consistency of Thyroid Indicators Tests in Minimally Invasive Peripheral Blood and Venous Blood

Abstract Disclosure: Z. Zhu: None. Y. Huang: None. Z. Cao: None. L. Zhang: None. D. Wang: None. S. Wang: None. J. Zhang: None. Y. Gao: None. Y. Zhang: None. Background The detection of thyroid indicators is of high demand in clinical and epidemiological screenings. Compared with existing venous blood tests, peripheral fingertip blood tests are less invasive and have a lower blood collection volume, which is the basis for point-of-care tests of thyroid indicators. Objective To explore the feasibility of minimally invasive and micro blood collection tests for thyroid disease (TD) patients by evaluating the consistency of thyroid indicators and TD diagnostic efficiency between fingertip blood tests and venous blood tests. Method 63 subjects with or without thyroid dysfunction were consecutively recruited. All subjects underwent serum, plasma and fingertip blood (same day collection) measurements for TSH and TT4 detection via the chemiluminescence method and for TPOAb detection via the electrochemiluminescence immunoassay method. The fingertip blood was centrifuged and diluted after EDTA anticoagulation. Bland–Altman plot, Passing–Bablok regression, and weighted kappa test were used for consistency test; AUC of ROC curve, sensitivity and specificity were used for authenticity evaluation among the three collection and testing methods. Results 43 subjects with euthyroidism and 20 subjects with thyroid dysfunction were included; including 29 (46.03%) females, the average age was 50.98±14.76 years. Bland-Altman plots showed more than 90% of the subjects had TSH and TT4 differences within the 95% limits of agreement between peripheral blood (PB) and venous blood (including serum and plasma). Passing–Bablok regression analyses revealed strong correlations between TSH levels in PB and serum (r=0.888) and plasma (r=0.894) (p<0.05); for TT4, there was a strong correlation between PB and serum (r=0.746), and there was also a moderate correlation between PB and plasma (r=0.664) (p<0.05 for both). Weighted kappa tests were performed for low, normal and high TSH values, indicating strong consistency between PB and serum and between PB and plasma, with coefficients of 0.901 and 0.881, respectively (P<0.001); for TT4, the coefficients between PB and serum, PB and plasma were 0.541 and 0.585 (P<0.001), indicating moderate consistency. Regardless of TPOAb positivity or negativity, the consistency among three blood sample sources was 100%. For euthyroidism (n=43), hyperthyroidism (n=6), subclinical hyperthyroidism (n=9), hypothyroidism (n=1), and subclinical hypothyroidism (n=4), the diagnostic sensitivities of PB were 97.06%, 83.33%, 87.50%, 100%, and 75.00%, respectively; the specificities were 94.74%, 97.87%, 95.74%, 98.11%, and 100.00%, respectively. The AUC of PB ROC curve for determining thyroid dysfunction was 0.944 (95% CI 0.867-1.000). Conclusion PB tests of thyroid indicators has high consistency with venous blood tests, offers sufficient diagnostic efficiency for TD, and is a feasible minimally invasive detection method. Presentation: 6/1/2024

7583 Evaluation of The Consistency of Thyroid Indicators Tests in Minimally Invasive Peripheral Blood and Venous Blood

Abstract Disclosure: Z. Zhu: None. Y. Huang: None. Z. Cao: None. L. Zhang: None. D. Wang: None. S. Wang: None. J. Zhang: None. Y. Gao: None. Y. Zhang: None. Background: The detection of thyroid indicators is of high demand in clinical and epidemiological screenings. Compared with existing venous blood tests, peripheral fingertip blood tests are less invasive and have a lower blood collection volume, which is the basis for point-of-care tests of thyroid indicators. Objective To explore the feasibility of minimally invasive and micro blood collection tests for thyroid disease (TD) patients by evaluating the consistency of thyroid indicators and TD diagnostic efficiency between fingertip blood tests and venous blood tests. Method 63 subjects with or without thyroid dysfunction were consecutively recruited. All subjects underwent serum, plasma and fingertip blood (same day collection) measurements for TSH and TT4 detection via the chemiluminescence method and for TPOAb detection via the electrochemiluminescence immunoassay method. The fingertip blood was centrifuged and diluted after EDTA anticoagulation. Bland–Altman plot, Passing–Bablok regression, and weighted kappa test were used for consistency test; AUC of ROC curve, sensitivity and specificity were used for authenticity evaluation among the three collection and testing methods. Results 43 subjects with euthyroidism and 20 subjects with thyroid dysfunction were included; including 29 (46.03%) females, the average age was 50.98±14.76 years. Bland-Altman plots showed more than 90% of the subjects had TSH and TT4 differences within the 95% limits of agreement between peripheral blood (PB) and venous blood (including serum and plasma). Passing–Bablok regression analyses revealed strong correlations between TSH levels in PB and serum (r=0.888) and plasma (r=0.894) (p<0.05); for TT4, there was a strong correlation between PB and serum (r=0.746), and there was also a moderate correlation between PB and plasma (r=0.664) (p<0.05 for both). Weighted kappa tests were performed for low, normal and high TSH values, indicating strong consistency between PB and serum and between PB and plasma, with coefficients of 0.901 and 0.881, respectively (P<0.001); for TT4, the coefficients between PB and serum, PB and plasma were 0.541 and 0.585 (P<0.001), indicating moderate consistency. Regardless of TPOAb positivity or negativity, the consistency among three blood sample sources was 100%. For euthyroidism (n=43), hyperthyroidism (n=6), subclinical hyperthyroidism (n=9), hypothyroidism (n=1), and subclinical hypothyroidism (n=4), the diagnostic sensitivities of PB were 97.06%, 83.33%, 87.50%, 100%, and 75.00%, respectively; the specificities were 94.74%, 97.87%, 95.74%, 98.11%, and 100.00%, respectively. The AUC of PB ROC curve for determining thyroid dysfunction was 0.944 (95% CI 0.867-1.000). Conclusion PB tests of thyroid indicators has high consistency with venous blood tests, offers sufficient diagnostic efficiency for TD, and is a feasible minimally invasive detection method. Presentation: 6/1/2024

7771 Surveillance Role in Differentiated Thyroid Cancer Through Ultrasensitive Thyroglobulin IRMA with High Negative Prediction Value

Abstract Disclosure: J. Jang: None. H. Kim: None. S. Ha: None. K. Jung: None. G. Jung: None. S. Cho: None. D. Park: None. G. Cheon: None. Y. Park: None. TSH-stimulated serum thyroglobulin (s-Tg) is a well-established marker for monitoring differentiated thyroid cancer (DTC) following remnant ablation therapy. The 2nd-generation immunoradiometric assay for Tg (ultrasensitive Tg, uTg) was developed to detect Tg with higher sensitivity compared with the 1st generation (conventional Tg, cTg). This study aimed to compare efficacy of both methods, cTg and uTg, to determine whether unstimulated Tg (us-Tg) could predict s-Tg levels. Paired serum samples were collected from 274 DTC patients before (us-Tg) and after (s-Tg) TSH stimulation for 1st radioiodine therapy after surgery (1st Tg) or both 1st therapy and 2nd therapy/diagnostic test (1st and 2nd Tg, respectively). Both cTg (Dynotest Tg-plus, BRAHMS Gmbh, Germany, LOD 0.1 ng/mL and functional sensitivity (FS) 0.2 ng/mL) and uTg (RIAKEY, Shinjin Medics, South Korea; LOD 0.01 ng/mL and FS 0.06 ng/mL) were measured. For analysis, samples with inadequate TSH stimulation (<30 μIU/mL) or with anti-thyroglobulin antibody levels > 60 U/mL were excluded. Ultimately, the included samples consisted of 296 for us-Tg and 352 for s-Tg.. Among these, paired samples for us-Tg and s-Tg were available in 260 cases (189 for 1st Tg and 71 for 2nd Tg). Comparing all Tg values, cTg and uTg levels had a correlation (Intraclass correlation coefficient r=0.752, P < 0.001), with no significant difference between us-Tg (r=0.803, P < 0.001) and s-Tg (r= 0.752, P < 0.001). The optimal cut-off level for predicting s-cTg at 1 ng/mL was 0.12 ng/mL in uTg or 0.16 ng/mL in cTg tests using AUC-ROC curve analysis. Considering the FS for cTg, diagnostic performance was assessed with a cut-off level of 0.2 ng/mL for cTg and 0.16 ng/mL for uTg. For s-cTg positivity as ≥ 1 ng/mL, the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of uTg were 72.1%, 59.8%. 52.0%, and 78.0%, respectively, while for cTg, they were 33.3%, 95.3%, 78.4%, and 73.9%, respectively. Similar trends were observed for s-cTg ≥ 2 ng/mL (69.0%, 71.4%, 35.0%, and 91.2% for uTg, respectively; 49.0%, 93.8%, 64.9%, and 88.7% for cTg, respectively). In essence, uTg demonstrated higher sensitivity than cTg, while having lower specificity and PPV, with a similar NPV. Of the 71 patients with paired 2nd Tg levels, 7 (9.8%) had biochemical disease (BCD) by s-cTg (1 ng/mL), whereas only 1 (1.4%) showed BCD using us-cTg ≥ 0.2 ng/mL. Conversely, 40 (56.0%) were classified as BCD using us-uTg ≥ 0.16 ng/mL. During the 6.48 years of follow-up, 4 patients developed structural recurrence; all had positive us-Tg levels with uTg tests, while only 1 patient demonstrated positivity with cTg tests. Although, the ultrasensitive Tg is sensitive in predicting positive stimulated Tg and recurrence, the specificity is very low resulting high frequency of BCD cases without clinical significance. Therefore, the clinical utility of uTg appears to be limited. Presentation: 6/2/2024

B-350 The application of MALDI-TOF imaging to biomarker discovery in a mouse model of papillomavirus associated tumorigenesis

Abstract Background Human papillomavirus (HPV) can cause infection and dysplasia in cutaneous epithelium of immunodeficient patients. An animal model has been developed to study HPV-associated tumorigenesis using Nude mouse (Foxn1nu) and mouse papillomavirus (MmuPV1). This study applies MALDI-TOF imaging to compare the expression pattern of small molecules between viral-infected dysplastic mouse skin and mock-infected control tissue. Three sets of molecules showing distinct patterns are identified in correlation with the upregulation of an enzyme in related biosynthetic pathway. Methods Nude mice were wounded on tail and inoculated with MmPV1 suspension (1×109 viral DNA genomes per infected site) or PBS 24 hours later for treatment and control group, respectively. Tail skins were harvested 4 weeks post-infection and processed for cryosectioning. Tissue sections were analyzed using timsTOF flex (Bruker) under negative ion mode with 9-aminoacridine as the matrix. Viral infection (MmuPV1-L1 antigen) and enzyme expression (3'-phosphoadenosine 5'-phosphosulfate synthetase 1, PAPSS1) were assessed by immunofluorescence on adjacent sections. Results Three peaks demonstrated distinct distribution patterns in mouse skin. One peak at m/Z = 465 was identified as cholesterol sulfate (CS) by tandem MS and comparing with the standard. CS was enriched in keratinocytes, hair follicles, and blood vessels in normal skins, and it increased in dysplastic skins with enhanced angiogenesis. The combined peaks at m/Z = 327 & 329 were tentatively identified as docosahexaenoic acid (DHA) and docosapentaenoic acid (DPA). These long chain polyunsaturated fatty acids (LC-PUFAs) were present in the dermis as foci partially co-localized with CS, and they also increased in virus-infected skins. The peak at m/Z = 426 may represent a combination of ADP and APS. Although these two isobars cannot be separated by tandem MS, APS was revealed by the presence of sulfur isotopes at expected natural abundance (34S and 32S). They were present at a low level in the epidermis of normal skins but significantly increased after virus infection. APS and ADP were closely associated with the proliferating cells, and the signal intensity also increased with the degree of tissue dysplasia. As a single factor, APS & ADP demonstrated a fair differential power for MmuPV1-induced dysplastic skin and normal control (AUC of ROC curve = 0.75). The bifunctional enzyme PAPSS1 catalyzes the conversion of ATP to 3’-phosphonato-5’-adenyl sulfate (PAPS) via APS as an intermediate. Interestingly, immunofluorescence showed the expression of PAPSS1 also increased significantly with the severity of tissue dysplasia and MmuPV1 infection. Conclusions MALDI-TOF imaging reveals three sets of molecules with distinct patterns in mouse skin, and their levels are increased during MmuPV1-induced epithelium dysplasia. As a precursor of the common sulfate donor (PAPS), increased APS and ADP in the epidermis of virus-infected skins indicate an upregulation of sulfation pathway and adenine metabolism. This finding is supported by enhanced expression of enzyme PAPSS1 and an increased sulfation reaction product, CS. In addition, DHA and DPA are also increased in the dermis of virus-infected skins, implying a disturbance in LC-PUFA metabolism. This study demonstrates combined use of MALDI-TOF imaging and immunostaining in the search of novel biomarkers of HPV-associated tumorigenesis.

Prediction of hemorrhage rate after tonsil surgery in children based on random forest model

Objective:Hemorrhage after tonsil surgery in children is a serious and potentially life-threatening complication. The purpose of this study was to establish a risk warning model for hemorrhage after tonsil surgery in children through a national multi-center retrospective study, providing a basis for hierarchical management after tonsil surgery in children. Methods:Stratified sampling was performed on 8 854 children who underwent tonsillectomy under general anesthesia from 15 research centers in different provinces from January 15, 2022 to May 15, 2023. The sample size of this study was 2 724 cases, including 1 096 males and 1 628 females. Children were divided into bleeding and non-bleeding groups according to whether or not they had bleeding after surgery. The random forest algorithm was used to build a risk warning model. By continuously exploring the optimized model, the accuracy of predicting the postoperative bleeding rate of tonsils in children was improved, and the prediction effectiveness of the model was verified by ten-fold cross-validation. Results:Among 2 724 children, 117 had postoperative bleeding after tonsillectomy, with a bleeding rate of 4.30%. The model constructed by the random forest algorithm for the training set was verified in the test set, and the obtained prediction accuracy was 98.72%, the recall rate was 78.95%, and the area under the ROC curve AUC was 0.96. Conclusion:Although the recall rate of the random forest model needs to be improved, the overall accuracy is quite excellent. It can effectively avoid misjudging positive cases as negative cases. It is a useful tool that can be used to predict the postoperative bleeding rate of tonsils and clinical medical decision-making, laying a good foundation for subsequent optimization and improvement.

The use of a clinical calculator to determine the rate of development of chronic kidney disease in elderly patients with type 2 diabetes mellitus

Aim – to develop an applied prognostic calculator for the rate of progression of CKD in elderly patients with type 2 diabetes, which makes it possible to identify a group of high rate of GFR reduction in conditions of routine outpatient admission. Material and methods. 69 clinical indicators were studied, the interrelationships and significance of differences in parameters in the groups identified according to the original diagnostic parameter, the glomerular filtration rate reduction index with a threshold value of 3.83 ml/min/1.73 m2 per year, above which the rate of progression of CKD was considered high. Results. By using regression analysis, significant factors for the prognostic calculator were identified: the duration of diabetes and insulin therapy, concomitant diagnosis of obesity, pulse in the ankle artery, severe stage of polyneuropathy, risk group IV of hypertension, treatment with sulfonylureas, the number of antihypertensive drugs taken. When evaluating the information capacity and predictive ability of the calculator, the area under the AUC ROC curve was 0.89 (0.80; 0.99) p 0.001, which characterizes the quality of the diagnostic technique as high. Conclusion. The presented calculator gives the doctor the opportunity to identify a group of patients with the risk of rapid progression of CKD directly on an outpatient basis.

Enhancing alzheimer’s diagnosis through optimized brain lesion classification in MRI with attention-driven grid feature fusion

This paper explores cognitive interface technology, aiming to tackle current challenges and shed light on the prospects of brain-computer interfaces (BCIs). It provides a comprehensive examination of their transformative impact on medical technology and patient well-being. Specifically, this study contributes to addressing challenges in classifying brain lesion images arising from the complex nature of lesions and limitations of traditional deep learning approaches. It introduces advanced feature fusion models that leverage deep learning algorithms, including the African vulture optimization (AVO) algorithm. These models integrate informative features from multiple pre-trained networks and employ innovative fusion techniques, including the attention-driven grid feature fusion (ADGFF) model. The ADGFF model incorporates an attention mechanism based on the optimized weights obtained using AVO. The objective is to improve the overall accuracy by providing fine-grained control over different regions of interest in the input image through a grid-based technique. This grid-based technique divides the image into vertical and horizontal grids, simplifying the exemplar feature generation process without compromising performance. Experimental results demonstrate that the proposed feature fusion strategies consistently outperform individual pre-trained models in terms of accuracy, sensitivity, specificity, and F1-score. The optimized feature fusion strategies, particularly the GRU-ADGFF model, further enhance classification performance, outperforming CNN and RNN classifiers. The learning progress analysis shows convergence, indicating the effectiveness of the feature fusion strategies in capturing lesion patterns. AUC-ROC curves highlight the superior discriminatory capabilities of the ADGFF-AVO strategy. Five-fold cross-validation is employed to assess the performance of the proposed models, demonstrating their accuracy, and few other accuracy-based measures. The GRU-ADGFF model optimized with AVO consistently achieves high accuracy, sensitivity, and AUC values, demonstrating its effectiveness and generalization capability. The GRU-ADGFF model also outperforms the majority voting ensemble technique in terms of accuracy and discriminative ability. Additionally, execution time analysis reveals good scalability and resource utilization of the proposed models. The Friedman rank test confirms significant differences in classifier performance, with the GRU-ADGFF model emerging as the top-performing method across different feature fusion strategies and optimization algorithms.

Support Vector Machine based Data Hacking Prediction using PMU Data

As global reliance on power systems grows due to increasing energy demands and modern consumption patterns, maintaining the stability and reliability of the power grid has become crucial. Power systems are complex and nonlinear, and their operations are continuously evolving, making it difficult and expensive to ensure stability. Traditionally, power systems are designed to handle a single outage at a time. However, recent years have seen several significant blackouts, each originating from a single failure, which have been extensively reported. These reports are vital for mitigating operational risks by strengthening systems against identified high-risk scenarios. While extensive research has been conducted on these blackouts, cyber- attacks introduce a new dimension of risk. The advent of Phasor Measurement Units (PMUs) has enabled centralized monitoring of power system data, allowing for more effective fault and cyber-attack detection.This paper proposes a machine learning-based approach to detecting cyber-attacks using PMU data. Given the complexity and volume of power system data, traditional mathematical and statistical methods are challenging to implement. Instead, a Support Vector Classification (SVC) algorithm is used for binary classification, distinguishing between 'attack' and 'normal' states. The algorithm is trained on PMU data and evaluated using metrics such as the AUC-ROC curve and confusion matrix, achieving an 82% AUC- ROC score, demonstrating its effectiveness in identifying cyber- attacks.

Monitoring Breast Cancer Response to Neoadjuvant Chemotherapy Using Probability Maps Derived from Quantitative Ultrasound Parametric Images.

Neoadjuvant chemotherapy (NAC) is widely used in the treatment of breast cancer. However, to date, there are no fully reliable, non-invasive methods for monitoring NAC. In this article, we propose a new method for classifying NAC-responsive and unresponsive tumors using quantitative ultrasound. The study used ultrasound data collected from breast tumors treated with NAC. The proposed method is based on the hypothesis that areas that characterize the effect of therapy particularly well can be found. For this purpose, parametric images of texture features calculated from tumor images were converted into NAC response probability maps, and areas with a probability above 0.5 were used for classification. The results obtained after the third cycle of NAC show that the classification of tumors using the traditional method (area under the ROC curve AUC = 0.81 - 0.88) can be significantly improved thanks to the proposed new approach (AUC = 0.84-0.94). This improvement is achieved over a wide range of cutoff values (0.2-0.7), and the probability maps obtained from different quantitative parameters correlate well. The results suggest that there are tumor areas that are particularly well suited to assessing response to NAC. The proposed approach to monitoring the effects of NAC not only leads to a better classification of responses, but also may contribute to a better understanding of the microstructure of neoplastic tumors observed in an ultrasound examination.

Multidimensional-Based Prediction of Pressure Ulcers Development and Severity in Hospitalized Frail Oldest Old: A Retrospective Study.

In recent times, growing uncertainty has emerged regarding the effectiveness of standard pressure ulcer (PU) risk assessment tools, which are suspected to be no better than clinical judgment, especially in the frail and comorbid elderly population. This study aimed to identify the primary clinical predictive variables for PU development and severity in hospitalized older adults, utilizing a multidimensional frailty assessment, and compare them with the Braden scale. The population consisted of 316 patients, admitted to the Geriatric Unit and Transitional Care of San Bartolomeo Hospital in Sarzana (Italy) during the period 21/02/22-01/07/22. The collected information included both anamnestic and laboratory data. A comprehensive geriatric assessment was performed, including also anthropometric and physical performance measurements. Multivariate logistic analysis was used, both in a binary classification test and in the subsequent ordinal classification test of severity levels. The final performance of the model was assessed by ROC curve estimation and AUC comparison with the Braden scale. Within the population, 152 subjects (48%) developed PU at different levels of severity. The results showed that age, Braden scale (subscales of mobility and friction/shear), Barthel scale, Mini Nutritional Assessment, hemoglobin, and albumin are predictors associated with the development of PU (AUC 85%). The result is an improvement over the use of the Braden scale alone (AUC 75%). Regarding the identification of predictive factors for PU severity, 4AT also emerges as potentially relevant. Assessing the subject's nutritional status, physical performance, and functional autonomies enables the effective integration of the Braden scale in identifying patients most susceptible to developing PU. Our findings support the integration of a comprehensive set of methodologically robust frailty determinants into traditional risk assessment tools. This integration reflects the mutual interplay between patients' frailty, skin frailty, and PU development in very old hospitalized patients.

Abnormal lower limb posture recognition based on spatial gait feature dynamic threshold detection

Lower limb rehabilitation training often involves the use of assistive standing devices. However, elderly individuals frequently experience reduced exercise effectiveness or suffer muscle injuries when utilizing these devices. The ability to recognize abnormal lower limb postures can significantly enhance training efficiency and minimize the risk of injury. To address this, we propose a model based on dynamic threshold detection of spatial gait features to identify such abnormal postures. A human-assisted standing rehabilitation device platform was developed to build a lower limb gait depth dataset. RGB data is employed for keypoint detection, enabling the establishment of a 3D lower limb posture recognition model that extracts gait, time, spatial features, and keypoints. The predicted joint angles, stride length, and step frequency demonstrate errors of 4 %, 8 %, and 1.3 %, respectively, with an average confidence of 0.95 for 3D key points. We employed the WOA-BP neural network to develop a dynamic threshold algorithm based on gait features and propose a model for recognizing abnormal postures. Compared to other models, our model achieves a 96 % accuracy rate in recognizing abnormal postures, with a recall rate of 83 % and an F1 score of 90 %. ROC curve analysis and AUC values reveal that the WOA-BP algorithm performs farthest from the pure chance line, with the highest AUC value of 0.89, indicating its superior performance over other models. Experimental results demonstrate that this model possesses a strong capability in recognizing abnormal lower limb postures, encouraging patients to correct these postures, thereby reducing muscle injuries and improving exercise effectiveness.

Prediction of PSA Response after Dexamethasone Switch during Abiraterone Acetate + Prednisolone Treatment of Metastatic Castration-Resistant Prostate Cancer Patients.

The aim was to elaborate a predictive model to find responders for the corticosteroid switch (from prednisolone to dexamethasone) at the first prostate-specific antigen (PSA) progression (≥25% increase) during abiraterone acetate (AA) treatment of metastatic castration-resistant prostate cancer (mCRPC) patients. If PSA has decreased (≥25%) after switch, patients were considered responders. Logistic regression of 19 dichotomized parameters from routine laboratory and patients' history was used to find the best model in a cohort of 67 patients. The model was validated in another cohort of 42 patients. The model provided 92.5% and 90.5% accuracy in the testing and the validation cohorts, respectively. Overall the accuracy was 91.7%. The AUC of ROC curve was 0.92 (95% CI 0.85-0.96). After a median follow-up of 27.9 (26.3-84) months, the median AA+dexamethasone treatment duration (TD) in non-responders and responders was 4.7 (3.1-6.5) and 11.1 (8.5-12.9) months and the median overall survival (OS) was 23.2 (15.6-25.8) and 33.5 (26.1-38) months, respectively. Multivariate Cox regression revealed that responsiveness was an independent marker of TD and OS. A high accuracy model was developed for mCRPC patients in predicting cases which might benefit from the switch. For non-responders, induction of the next systemic treatment is indicated.

Developing flood mapping procedure through optimized machine learning techniques. Case study: Prahova river basin, Romania

Study regionPrahova river basin located in the central-southern region of Romania. Study focusThis study aims to assess the susceptibility to flooding by using state-of-the-art machine learning and optimization procedures. To achieve this goal, we employed ten flood-related variables as independent variables in our machine learning models. These variables include slope angle, convergence index, distance from the river, elevation, plan curvature, hydrological soil group, lithology, topographic wetness index, rainfall, and land use. We used 158 flood locations as dependent variables in the training of four hybrid models: Deep Learning Neural Network-Statistical Index (DLNN-SI), Particle Swarm Optimization-Deep Learning Neural Network-Statistical Index (PSO-DLNN-SI), Support Vector Machine-Statistical Index (SVM-SI), and Particle Swarm Optimization-Support Vector Machine-Statistical Index (PSO-SVM-SI). Utilizing the Statistical Index method, we calculated coefficients for each flood predictor class or category. New hydrological insights for the regionThe PSO-DLNN-SI model demonstrated the best performance, achieving an AUC-ROC curve of 0.952. It's worth noting that the application of the PSO algorithm significantly enhanced the model's performance. Additionally, it's crucial to highlight that approximately 25 % of the study region exhibits a high to very high susceptibility to flood events. Taking into account the very precise results of the models applied in the present study, we can state that from a hydrological point of view, the current research contributes to a better understanding of the intensity with which floods can affect the different areas of the Prahova river basin.

Understanding the value of monocyte distribution width (MDW) in acutely ill medical patients presenting to the emergency department: a prospective single center evaluation

The monocyte distribution width (MDW) has emerged as a promising biomarker for accurate and early identification of patients with potentially life-threatening infections. Here we tested the diagnostic performance of MDW in adult patients requiring hospital admission for community-acquired infections and sepsis, evaluated sources of heterogeneity in the estimates of diagnostic accuracy, and assessed the meaning of MDW in a patient population presenting to the emergency department (ED) for acute non-infectious conditions. 1925 consecutive patients were categorized into three groups: non-infection (n = 1507), infection (n = 316), and sepsis/septic shock (n = 102). Diagnostic performance for infection or sepsis of MDW alone or in combination with components of SOFA was tested using AUC of ROC curves, sensitivity, and specificity. The relationship between MDW and different pathogens as well as the impact of non-infectious conditions on MDW values were explored. For the prediction of infection, the AUC/ROC of MDW (0.84) was nearly overlapping that of procalcitonin (0.83), and C-reactive protein (0.89). Statistical optimal cut-off value for MDW was 21 for predicting infection (sensitivity 73%, specificity 82%) and 22 for predicting sepsis (sensitivity 79%, specificity 83%). The best threshold to rule out infection was MDW ≤ 17 (NPV 96.9, 95% CI 88.3–100.0), and ≤ 18 (NPV 99.5, 95% CI 98.3–100.0) to rule out sepsis. The combination of MDW with markers of organ dysfunction (creatinine, bilirubin, platelets) substantially improved the AUC (0.96 (95% CI 0.94–0.97); specificity and sensitivity of 88% and 94%, respectively). In conclusion, MDW has a good diagnostic performance in diagnosing infection and sepsis in patients presenting in ED. Its use as an infection marker even increases when combined with other markers of organ dysfunction. Understanding the impact of interactions of non-infectious conditions and comorbidities on MDW and its diagnostic accuracy requires further elucidation.

Machine learning-based model to predict severe acute kidney injury after total aortic arch replacement for acute type A aortic dissection

BackgroundSevere acute kidney injury (AKI) after total aortic arch replacement (TAAR) is related to adverse outcomes in patients with acute type A aortic dissection (ATAAD). However, the early prediction of severe AKI remains a challenge. This study aimed to develop a novel model to predict severe AKI after TAAR in ATAAD patients using machine learning algorithms. MethodsA total of 572 ATAAD patients undergoing TAAR were enrolled in this retrospective study, and randomly divided into a training set (70 %) and a validation set (30 %). Lasso regression, support vector machine-recursive feature elimination and random forest algorithms were used to screen indicators for severe AKI (defined as AKI stage III) in the training set, respectively. Then the intersection indicators were selected to construct models through artificial neural network (ANN) and logistic regression. The AUC-ROC curve was employed to ascertain the prediction efficacy of the ANN and logistic regression models. ResultsThe incidence of severe AKI after TAAR was 22.9 % among ATAAD patients. The intersection predictors identified by different machine learning algorithms were baseline serum creatinine and ICU admission variables, including serum cystatin C, procalcitonin, aspartate transaminase, platelet, lactic dehydrogenase, urine N-acetyl-β-d-glucosidase and Acute Physiology and Chronic Health Evaluation II score. The ANN model showed a higher AUC-ROC than logistic regression (0.938 vs 0.908, p < 0.05). Furthermore, the ANN model could predict 89.1 % of severe AKI cases beforehand. In the validation set, the superior performance of the ANN model was further confirmed in terms of discrimination ability (AUC = 0.916), calibration curve analysis and decision curve analysis. ConclusionThis study developed a novel and reliable clinical prediction model for severe AKI after TAAR in ATAAD patients using machine learning algorithms. Importantly, the ANN model showed a higher predictive ability for severe AKI than logistic regression.

Machine learning algorithms to predict healthcare-seeking behaviors of mothers for acute respiratory infections and their determinants among children under five in sub-Saharan Africa.

Acute respiratory infections (ARIs) are the leading cause of death in children under the age of 5 globally. Maternal healthcare-seeking behavior may help minimize mortality associated with ARIs since they make decisions about the kind and frequency of healthcare services for their children. Therefore, this study aimed to predict the absence of maternal healthcare-seeking behavior and identify its associated factors among children under the age 5 in sub-Saharan Africa (SSA) using machine learning models. The sub-Saharan African countries' demographic health survey was the source of the dataset. We used a weighted sample of 16,832 under-five children in this study. The data were processed using Python (version 3.9), and machine learning models such as extreme gradient boosting (XGB), random forest, decision tree, logistic regression, and Naïve Bayes were applied. In this study, we used evaluation metrics, including the AUC ROC curve, accuracy, precision, recall, and F-measure, to assess the performance of the predictive models. In this study, a weighted sample of 16,832 under-five children was used in the final analysis. Among the proposed machine learning models, the random forest (RF) was the best-predicted model with an accuracy of 88.89%, a precision of 89.5%, an F-measure of 83%, an AUC ROC curve of 95.8%, and a recall of 77.6% in predicting the absence of mothers' healthcare-seeking behavior for ARIs. The accuracy for Naïve Bayes was the lowest (66.41%) when compared to other proposed models. No media exposure, living in rural areas, not breastfeeding, poor wealth status, home delivery, no ANC visit, no maternal education, mothers' age group of 35-49 years, and distance to health facilities were significant predictors for the absence of mothers' healthcare-seeking behaviors for ARIs. On the other hand, undernourished children with stunting, underweight, and wasting status, diarrhea, birth size, married women, being a male or female sex child, and having a maternal occupation were significantly associated with good maternal healthcare-seeking behaviors for ARIs among under-five children. The RF model provides greater predictive power for estimating mothers' healthcare-seeking behaviors based on ARI risk factors. Machine learning could help achieve early prediction and intervention in children with high-risk ARIs. This leads to a recommendation for policy direction to reduce child mortality due to ARIs in sub-Saharan countries.