Intelligent medicine最新文献

Background

Methods

Results

Conclusion

Background

The reproducibility of positron emission tomography (PET) radiomics features is affected by several factors, such as scanning equipment, drug metabolism time and reconstruction algorithm. We aimed to explore the role of 3D local binary pattern (LBP)-based texture in increasing the accuracy and reproducibility of PET radiomics for predicting pelvic lymph node metastasis (PLNM) in patients with cervical cancer.

Methods

We retrospectively analysed data from 177 patients with cervical squamous cell carcinoma. They underwent ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG)whole-body PET/computed tomography (PET/CT), followed by pelvic ¹⁸F-FDG PET/magnetic resonance imaging (PET/MR). We selected reproducible and informative PET radiomics features using Lin's concordance correlation coefficient, least absolute shrinkage and selection operator algorithm, and established 4 models, PET/CT, PET/CT-fusion, PET/MR and PET/MR-fusion, using the logistic regression algorithm. We performed receiver operating characteristic (ROC) curve analysis to evaluate the models in the training data set (65 patients who underwent radical hysterectomy and pelvic lymph node dissection) and test data set (112 patients who received concurrent chemoradiotherapy or no treatment). The DeLong test was used for pairwise comparison of the ROC curves among the models.

Results

The distribution of age, squamous cell carcinoma (SCC), International Federation of Gynaecology and Obstetrics stage and PLNM between the training and test data sets were different (P < 0.05). The LBP-transformed radiomics features (50/379) had higher reproducibility than the original radiomics features (9/107). Accuracy of each model in predicting PLNM was as follows: training data set: PET/CT = PET/CT-fusion = PET/MR-fusion (0.848) and test data set: PET/CT = PET/CT-fusion (0.985) > PET/MR = PET/MR-fusion (0.954). There was no statistical difference between the ROC curve of PET/CT and PET/MR models in both data sets (P > 0.05).

Conclusions

The LBP-transformed radiomics features based on PET images could increase the accuracy and reproducibility of PET radiomics in predicting pelvic lymph node metastasis in cervical cancer to allow the model to be generalised for clinical use across multiple centres.

Objective

Segmentation of medical images is a crucial process in various image analysis applications. Automated segmentation methods excel in accuracy when compared to manual segmentation in the context of medical image analysis. One of the essential phases in the quantitative analysis of the brain is automated brain tissue segmentation using clinically obtained magnetic resonance imaging (MRI) data. It allows for precise quantitative examination of the brain, which aids in diagnosis, identification, and classification of disorders. Consequently, the efficacy of the segmentation approach is crucial to disease diagnosis and treatment planning.

Methods

This study presented a hybrid optimization method for segmenting brain tissue in clinical MRI scans using a fractional Henry horse herd gas optimization-based Shepard convolutional neural network (FrHHGO-based ShCNN). To segment the clinical brain MRI images into white matter (WM), grey matter (GM), and cerebrospinal fluid (CSF) tissues, the proposed framework was evaluated on the Lifespan Human Connectome Projects (HCP) database. The hybrid optimization algorithm, FrHHGO, integrates the fractional Henry gas optimization (FHGO) and horse herd optimization (HHO) algorithms. Training required 30 min, whereas testing and segmentation of brain tissues from an unseen image required an average of 12 s.

Results

Compared to the results obtained with no refinements, the Skull stripping refinement showed significant improvement. As the method included a preprocessing stage, it was flexible enough to enhance image quality, allowing for better results even with low-resolution input. Maximum precision of 93.2%, recall of 91.5%, Dice score of 91.1%, and F1-score of 90.5% were achieved using the proposed FrHHGO-based ShCNN, which was superior to all other approaches.

Conclusion

The proposed method may outperform existing state-of-the-art methodologies in qualitative and quantitative measurements across a wide range of medical modalities. It might demonstrate its potential for real-life clinical application.

Background

Bacterial vaginosis is a polymicrobial syndrome in which the homeostasis exerted by the Latobacillus species that protect the vaginal mucosa has been lost. This study explored the data balancing process with the intention of improving the quality of association rules. The article aimed to balance the unbalanced multiclass dataset to improve association rule creation.

Methods

A dataset with 201 observations and 58 variables was analyzed. A preconstructed dataset was used. The authors collected the data between August 2016 and October 2018 in Tabasco, Mexico. The study population comprised sexually active women ages 18 to 50 who underwent gynecological inspection at the infectious and metabolic diseases research laboratory at the Universidad Juarez Autonoma de Tabasco. To determine the best $k$-value, the random-forest algorithm was used and the balancing was performed with the synthetic minority over-sampling technique (SMOTE), random over-sampling examples (ROSE), and adaptive syntetic sampling approach for imbalanced learning (ADASYN) algorithms. The Apriori algorithm created the rules and to select rules with statistical significance, the is.redundant(), is.significant(), and is.maximal() functions and quality metric Fisher’s exact tes were used. The biological validation was carried out by the expert (bacteriologist).

Results

The ADASYN algorithm at $K=9$ the out of the bag (OOB) error was zero, this was the best $K$-values. In the balancing process the ADASYN algorithm show best the performance. From the dataset balanced with ADASYN, the apriori algorithm created the association rules and the selection with the quality metric Fisher’s exact test, and the biological validation reported 13 rules. Gram - bacteria Atopobium vaginae, Gardnerella vaginalis, Megasphaera filotipo 1, Mycoplasma hominis and Ureaplasma parvum were detected by the apriori algorithm from the balanced dataset.

Conclusion

Balancing may improve the creation of association rules to efficiently model the bacteria that cause bacterial vaginosis.

Background

Precise estimation of current and future comorbidities of patients with cardiovascular disease is an important factor in prioritizing continuous physiological monitoring and new therapies. Machine learning (ML) models have shown satisfactory performance in short-term mortality prediction in patients with heart disease, whereas their utility in long-term predictions is limited. This study aimed to investigate the performance of tree-based ML models on long-term mortality prediction and effect of two recently introduced biomarkers on long-term mortality.

Methods

This study used publicly available data from the Collaboration Center of Health Information Application at the Ministry of Health and Welfare, Taiwan, China. The collected data were from patients admitted to the cardiac care unit for acute myocardial infarction (AMI) between November 2003 and September 2004. We collected and analyzed mortality data up to December 2018. Medical records were used to gather demographic and clinical data, including age, gender, body mass index, percutaneous coronary intervention status, and comorbidities such as hypertension, dyslipidemia, ST-segment elevation myocardial infarction, and non-ST-segment elevation myocardial infarction. Using the data, collected from 139 patients with AMI, from medical and demographic records as well as two recently introduced biomarkers, brachial pre-ejection period (bPEP) and brachial ejection time (bET), we investigated the performance of advanced ensemble tree-based ML algorithms (random forest, AdaBoost, and XGBoost) to predict all-cause mortality within 14 years. A nested cross-validation was performed to evaluate and compare the performance of our developed models precisely with that of the conventional logistic regression (LR) as the baseline method.

Results

The developed ML models achieved significantly better performance compared to the baseline LR (C-Statistic, 0.80 for random forest, 0.79 for AdaBoost, and 0.78 for XGBoost, vs. 0.77 for LR) (P_RF < 0.001, P_AdaBoost < 0.001, and P_XGBoost < 0.05). Adding bPEP and bET to our feature set significantly improved the performance of the algorithm, leading to an absolute increase in C-statistic of up to 0.03 (C-statistic, 0.83 for random forest, 0.82 for AdaBoost, and 0.80 for XGBoost, vs. 0.74 for LR) (P_RF < 0.001, P_AdaBoost < 0.001, P_XGBoost < 0.05).

Conclusion

The study indicates that incorporating new biomarkers into advanced ML models may significantly improve long-term mortality prediction in patients with cardiovascular diseases. This advancement may enable better treatment prioritization for high-risk individuals.

Neuropsychological disorders (e.g., dementia, epilepsy, brain cancer, autism, stroke, and multiple sclerosis) adversely affect the quality of life of patients and their families; moreover, in some instances, they may lead to loss of life. The primary aim was to evaluate and compare the use of machine learning in neuropsychological research in contrast to traditional approaches such as through case studies. This was achieved by referring to earlier studies on this subject. This article presented the use of support vector machines (SVMs) and convolutional neural networks (CNN) for detecting and predicting neuropsychological diseases, such as dementia and Alzheimer's disease. Challenges in using these models include data availability, quality, variability, model interpretability, and validation. Experimental findings have demonstrated the potential of these models in this field. It has been shown that SVM models are robust and efficient in processing and classifying data, particularly in neuroimaging applications, such as magnetic resonance imaging (MRI). CNNs have excelled in handling visual input; thus, they have been used in neuroimaging segregation, recognition, and classification, with applications in brain tumor segmentation, radiation therapy, robotic neurosurgery, and disease prediction. Future research will explore asymmetric differences among left- and right-handed patients, incorporate longitudinal studies, and utilize larger sample sizes. The use of machine learning models has the potential to revolutionize the diagnosis and treatment of neuropsychological diseases, allowing for early detection and intervention. This approach could offer significant advantages to healthcare, such as cost-effective diagnosis and treatment, to help save lives and preserve the quality of life of patients.

Objective

Technological advances have led to drastic changes in daily life, and particularly healthcare, while traditional diagnosis methods are being replaced by technology-oriented models and paper-based patient healthcare records with digital files. Using the latest technology and data mining techniques, we aimed to develop an automated clinical decision support system (CDSS), to improve patient prognoses and healthcare delivery. Our proposed approach placed a strong emphasis on improvements that meet patient, parent, and physician expectations. We developed a flexible framework to identify hepatitis, dermatological conditions, hepatic disease, and autism in adults and provide results to patients as recommendations. The novelty of this CDSS lies in its integration of rough set theory (RST) and machine learning (ML) techniques to improve clinical decision-making accuracy and effectiveness.

Methods

Data were collected through various web-based resources. Standard preprocessing techniques were applied to encode categorical features, conduct min-max scaling, and remove null and duplicate entries. The most prevalent feature in the class and standard deviation were used to fill missing categorical and continuous feature values, respectively. A rough set approach was applied as feature selection, to remove highly redundant and irrelevant elements. Then, various ML techniques, including K nearest neighbors (KNN), linear support vector machine (LSVM), radial basis function support vector machine (RBF SVM), decision tree (DT), random forest (RF), and Naive Bayes (NB), were employed to analyze four publicly available benchmark medical datasets of different types from the UCI repository and Kaggle. The model was implemented in Python, and various validity metrics, including precision, recall, F1-score, and root mean square error (RMSE), applied to measure its performance.

Results

Features were selected using an RST approach and examined by RF analysis and important features of hepatitis, dermatology conditions, hepatic disease, and autism determined by RST and RF exhibited 92.85 %, 90.90 %, 100 %, and 80 % similarity, respectively. Selected features were stored as electronic health records and various ML classifiers, such as KNN, LSVM, RBF SVM, DT, RF, and NB, applied to classify patients with hepatitis, dermatology conditions, hepatic disease, and autism. In the last phase, the performance of proposed classifiers was compared with that of existing state-of-the-art methods, using various validity measures. RF was found to be the best approach for adult screening of: hepatitis with accuracy 88.66 %, precision 74.46 %, recall 75.17 %, F1-score 74.81 %, and RMSE value 0.244; dermatology conditions with accuracy 97.29 %, precision 96.96 %, recall 96.96 %, F1-score 96.96 %, and RMSE value, 0.173; hepatic disease, with accuracy 91.58 %, precision 81.76 %, recall 81.82 %, F1-Score 81.79 %, and RMS