The Matthews correlation coefficient (MCC) should replace the ROC AUC as the standard metric for assessing binary classification

BioData Mining, Год журнала: 2023, Номер 16(1)

Опубликована: Фев. 17, 2023

Binary classification is a common task for which machine learning and computational statistics are used, the area under receiver operating characteristic curve (ROC AUC) has become standard metric to evaluate binary classifications in most scientific fields. The ROC true positive rate (also called sensitivity or recall) on y axis false x axis, AUC can range from 0 (worst result) 1 (perfect result). AUC, however, several flaws drawbacks. This score generated including predictions that obtained insufficient specificity, moreover it does not say anything about predictive value known as precision) nor negative (NPV) by classifier, therefore potentially generating inflated overoptimistic results. Since include alone without precision value, researcher might erroneously conclude their was successful. Furthermore, given point space identify single confusion matrix group of matrices sharing same MCC value. Indeed, (sensitivity, specificity) pair cover broad range, casts doubts reliability performance measure. In contrast, Matthews correlation coefficient (MCC) generates high its [Formula: see text] interval only if classifier scored all four basic rates matrix: sensitivity, precision, A (for example, 0.9), moreover, always corresponds vice versa. this short study, we explain why should replace statistic studies involving classification,

Язык: Английский

---

Ten quick tips for avoiding pitfalls in multi-omics data integration analyses

PLoS Computational Biology, Год журнала: 2023, Номер 19(7), С. e1011224 - e1011224

Опубликована: Июль 6, 2023

Data are the most important elements of bioinformatics: Computational analysis bioinformatics data, in fact, can help researchers infer new knowledge about biology, chemistry, biophysics, and sometimes even medicine, influencing treatments therapies for patients. Bioinformatics high-throughput biological data coming from different sources be more helpful, because each these chunks provide alternative, complementary information a specific phenomenon, similar to multiple photos same subject taken angles. In this context, integration gets pivotal role running successful study. last decades, originating proteomics, metabolomics, metagenomics, phenomics, transcriptomics, epigenomics have been labelled -omics as unique name refer them, omics has gained importance all areas. Even if is useful relevant, due its heterogeneity, it not uncommon make mistakes during phases. We therefore decided present ten quick tips perform an correctly, avoiding common we experienced or noticed published studies past. designed our guidelines beginners, by using simple language that (we hope) understood anyone, believe recommendations should into account bioinformaticians performing integration, including experts.

Язык: Английский

---

Machine learning approach using 18F-FDG-PET-radiomic features and the visibility of right ventricle 18F-FDG uptake for predicting clinical events in patients with cardiac sarcoidosis

Japanese Journal of Radiology, Год журнала: 2024, Номер 42(7), С. 744 - 752

Опубликована: Март 16, 2024

To investigate the usefulness of machine learning (ML) models using pretreatment

Язык: Английский

---

Application of Machine Learning Analyses Using Clinical and [18F]-FDG-PET/CT Radiomic Characteristics to Predict Recurrence in Patients with Breast Cancer

Molecular Imaging and Biology, Год журнала: 2023, Номер 25(5), С. 923 - 934

Опубликована: Май 16, 2023

Язык: Английский

---

Ten quick tips for electrocardiogram (ECG) signal processing

PeerJ Computer Science, Год журнала: 2024, Номер 10, С. e2295 - e2295

Опубликована: Сен. 3, 2024

The electrocardiogram (ECG) is a powerful tool to measure the electrical activity of heart, and analysis its data can be useful assess patient's health. In particular, computational data, also called ECG signal processing, reveal specific patterns or heart cycle trends which otherwise would unnoticeable by medical experts. When performing however, it easy make mistakes generate inflated, overoptimistic, misleading results, lead wrong diagnoses prognoses and, in turn, could even contribute bad decisions, damaging health patient. Therefore, avoid common practices, we present here ten guidelines follow when analyzing computationally. Our recommendations, written simple way, anyone study based on eventually better, more robust results.

Язык: Английский

---

Robust cardiac segmentation corrected with heuristics

PLoS ONE, Год журнала: 2023, Номер 18(10), С. e0293560 - e0293560

Опубликована: Окт. 27, 2023

Cardiovascular diseases related to the right side of heart, such as Pulmonary Hypertension, are some leading causes death among Mexican (and worldwide) population. To avoid invasive techniques catheterizing improving segmenting performance medical echocardiographic systems can be an option early detect right-side heart. While current imaging perform well automatically left they typically struggle cavities. This paper presents a robust cardiac segmentation algorithm based on popular U-NET architecture capable accurately four cavities with reduced training dataset. Moreover, we propose two additional steps improve quality results in our machine learning model, 1) detecting cone shapes (as it has been trained and refined multiple data sources) 2) post-processing step which refines shape contours heuristics provided by clinicians. Our demonstrate that proposed achieve accuracy comparable state-of-the-art methods datasets commonly used for this practice, compiled team. Furthermore, tested validity correction within same sequence images demonstrated its consistency manual segmentations performed

Язык: Английский

---

Deep learning in medical image analysis

Elsevier eBooks, Год журнала: 2024, Номер unknown, С. 287 - 295

Опубликована: Ноя. 22, 2024

Язык: Английский

---

Bioinformatics: A New Tool in Dentistry

European Journal of Medical and Health Research, Год журнала: 2024, Номер 2(1), С. 83 - 90

Опубликована: Янв. 1, 2024

Bioinformatics is a new speciality that focuses on using information science to solve biological problems. It deals with the collecting, storing, retrieving and analysing data from databases. has supported promoted research in field of healthcare taken it next level. can encourage dentistry by understanding underlying pathways mechanisms certain oral diseases. also help early prediction personalized treatment cancer may prove beneficial detection accurate cancer. supports developing patient care databases, image analysis X- rays, CT MRI. Diagnostic abilities will multiple databases management. Salivanomics sub-speciality bioinformatics dealing saliva knowledge base enabling global exploration relevant saliva. Incorporation AI machine learning lead immense positive outcomes personalised medicine gene therapy. This review understand tools used its role dentistry.

Язык: Английский

---

Optimizing Medical Image Analysis: Leveraging Efficient Hardware and AI Algorithms

Опубликована: Янв. 6, 2024

The surging use of medical AI algorithms and their hardware integration is transforming healthcare by improving non-invasive analysis with early disease detection, advanced segmentation, classification. However, realizing comprehensive accurate through efficient AI-based tools necessitates a fundamental requirement — extensive multimodal data for training deep learning models. Handling this volume demands significant resources, including multi-node training, to address the substantial computational requirements essential accelerating model development. Hence, challenge two-fold: Achieving high accuracy while upholding computationally inexpensive solution. To navigate challenge, we propose novel solution: lightweight predictive tool image classification developed combining Radiomics-based Random Forest MobileViT transformer, tailored mobile applications. This approach ensures enhanced reproducibility along flexibility. Our proposed method exemplified its superior performance in BraTS2021 surpassing current state-of-the-art models best AUROC 0.64 0.63 on both public private test datasets respectively. success our highlights potential hybrid diverse applications beyond

Язык: Английский

---

Machine Learning Analysis of Predictors for Inhaled Nitric Oxide Therapy Administration Time Post Congenital Heart Disease Surgery: A Single-Center Observational Study

Cureus, Год журнала: 2024, Номер unknown

Опубликована: Июль 30, 2024

Background Congenital heart disease (CHD) is a structural deformity of the present at birth. Pulmonary hypertension (PH) may arise from increased blood flow to lungs, persistent pulmonary arterial pressure elevation, or use cardiopulmonary bypass (CPB) during surgical repair. Inhaled nitric oxide (iNO) selectively reduces high in vessels without lowering systemic pressure, making it useful for treating children with postoperative PH due disease. However, reducing stopping iNO can exacerbate and hypoxemia, necessitating long-term administration careful tapering. This study aimed evaluate, using machine learning (ML), factors that predict need after open surgery CHD patients ICU, primarily management. Methods We used an ML approach establish algorithm 'patients iNO' validate its accuracy 34 pediatric who survived were discharged ICU Kagoshima University Hospital between April 2016 March 2019. All started on therapy upon admission. Overall, 16 features reflecting patient characteristics utilized needed over 168 hours analysis AutoGluon. The dataset was randomly classified into training test cohorts, comprising 80% 20% data, respectively. In cohort, model constructed important selected by decrease Gini impurity synthetic oversampling technique. testing prediction performance evaluated calculating area under receiver operating curve (AUC) accuracy. Results Among 28 five hours; among six one hours. CPB, aortic clamp time, in-out balance, lactate four most predicting achieved perfect classification AUC 1.00. also 1.00 Conclusion identified (CPB, cross-clamp lactate) are strongly associated patients. By understanding outcomes this study, we more effectively manage PH, potentially preventing recurrence thereby contributing safer

Язык: Английский

---