Machine Learning in Disease Screening, Diagnosis, and Surveillance

Background: Continuously growing of HIV incidence among men who have sex with men (MSM), as well as the low rate of HIV testing of MSM in China, demonstrates a need for innovative strategies to improve the implementation of HIV prevention. The use of machine learning algorithms is an increasing tendency in disease diagnosis prediction. We aimed to develop and validate machine learning models in predicting HIV infection among MSM that can identify individuals at increased risk of HIV acquisition for transmission-reduction interventions.

Methods: We extracted data from MSM sentinel surveillance in Zhejiang province from 2018 to 2020. Univariate logistic regression was used to select significant variables in 2018–2019 data (P < 0.05). After data processing and feature selection, we divided the model development data into two groups by stratified random sampling: training data (70%) and testing data (30%). The Synthetic Minority Oversampling Technique (SMOTE) was applied to solve the problem of unbalanced data. The evaluation metrics of model performance were comprised of accuracy, precision, recall, F-measure, and the area under the receiver operating characteristic curve (AUC). Then, we explored three commonly-used machine learning algorithms to compare with logistic regression (LR), including decision tree (DT), support vector machines (SVM), and random forest (RF). Finally, the four models were validated prospectively with 2020 data from Zhejiang province.

Results: A total of 6,346 MSM were included in model development data, 372 of whom were diagnosed with HIV. In feature selection, 12 variables were selected as model predicting indicators. Compared with LR, the algorithms of DT, SVM, and RF improved the classification prediction performance in SMOTE-processed data, with the AUC of 0.778, 0.856, 0.887, and 0.942, respectively. RF was the best-performing algorithm (accuracy = 0.871, precision = 0.960, recall = 0.775, F-measure = 0.858, and AUC = 0.942). And the RF model still performed well on prospective validation (AUC = 0.846).

Conclusion: Machine learning models are substantially better than conventional LR model and RF should be considered in prediction tools of HIV infection in Chinese MSM. Further studies are needed to optimize and promote these algorithms and evaluate their impact on HIV prevention of MSM.

Many works have employed Machine Learning (ML) techniques in the detection of Diabetic Retinopathy (DR), a disease that affects the human eye. However, the accuracy of most DR detection methods still need improvement. Gray Wolf Optimization-Extreme Learning Machine (GWO-ELM) is one of the most popular ML algorithms, and can be considered as an accurate algorithm in the process of classification, but has not been used in solving DR detection. Therefore, this work aims to apply the GWO-ELM classifier and employ one of the most popular features extractions, Histogram of Oriented Gradients-Principal Component Analysis (HOG-PCA), to increase the accuracy of DR detection system. Although the HOG-PCA has been tested in many image processing domains including medical domains, it has not yet been tested in DR. The GWO-ELM can prevent overfitting, solve multi and binary classifications problems, and it performs like a kernel-based Support Vector Machine with a Neural Network structure, whilst the HOG-PCA has the ability to extract the most relevant features with low dimensionality. Therefore, the combination of the GWO-ELM classifier and HOG-PCA features might produce an effective technique for DR classification and features extraction. The proposed GWO-ELM is evaluated based on two different datasets, namely APTOS-2019 and Indian Diabetic Retinopathy Image Dataset (IDRiD), in both binary and multi-class classification. The experiment results have shown an excellent performance of the proposed GWO-ELM model where it achieved an accuracy of 96.21% for multi-class and 99.47% for binary using APTOS-2019 dataset as well as 96.15% for multi-class and 99.04% for binary using IDRiD dataset. This demonstrates that the combination of the GWO-ELM and HOG-PCA is an effective classifier for detecting DR and might be applicable in solving other image data types.

Background: Artificial intelligence-based disease prediction models have a greater potential to screen COVID-19 patients than conventional methods. However, their application has been restricted because of their underlying black-box nature.

Objective: To addressed this issue, an explainable artificial intelligence (XAI) approach was developed to screen patients for COVID-19.

Methods: A retrospective study consisting of 1,737 participants (759 COVID-19 patients and 978 controls) admitted to San Raphael Hospital (OSR) from February to May 2020 was used to construct a diagnosis model. Finally, 32 key blood test indices from 1,374 participants were used for screening patients for COVID-19. Four ensemble learning algorithms were used: random forest (RF), adaptive boosting (AdaBoost), gradient boosting decision tree (GBDT), and extreme gradient boosting (XGBoost). Feature importance from the perspective of the clinical domain and visualized interpretations were illustrated by using local interpretable model-agnostic explanations (LIME) plots.

Results: The GBDT model [area under the curve (AUC): 86.4%; 95% confidence interval (CI) 0.821–0.907] outperformed the RF model (AUC: 85.7%; 95% CI 0.813–0.902), AdaBoost model (AUC: 85.4%; 95% CI 0.810–0.899), and XGBoost model (AUC: 84.9%; 95% CI 0.803–0.894) in distinguishing patients with COVID-19 from those without. The cumulative feature importance of lactate dehydrogenase, white blood cells, and eosinophil counts was 0.145, 0.130, and 0.128, respectively.

Conclusions: Ensemble machining learning (ML) approaches, mainly GBDT and LIME plots, are efficient for screening patients with COVID-19 and might serve as a potential tool in the auxiliary diagnosis of COVID-19. Patients with higher WBC count, higher LDH level, or higher EOT count, were more likely to have COVID-19.

Survival prediction is highly valued in end-of-life care clinical practice, and patient performance status evaluation stands as a predominant component in survival prognostication. While current performance status evaluation tools are limited to their subjective nature, the advent of wearable technology enables continual recordings of patients' activity and has the potential to measure performance status objectively. We hypothesize that wristband actigraphy monitoring devices can predict in-hospital death of end-stage cancer patients during the time of their hospital admissions. The objective of this study was to train and validate a long short-term memory (LSTM) deep-learning prediction model based on activity data of wearable actigraphy devices. The study recruited 60 end-stage cancer patients in a hospice care unit, with 28 deaths and 32 discharged in stable condition at the end of their hospital stay. The standard Karnofsky Performance Status score had an overall prognostic accuracy of 0.83. The LSTM prediction model based on patients' continual actigraphy monitoring had an overall prognostic accuracy of 0.83. Furthermore, the model performance improved with longer input data length up to 48 h. In conclusion, our research suggests the potential feasibility of wristband actigraphy to predict end-of-life admission outcomes in palliative care for end-stage cancer patients.

Clinical Trial Registration: The study protocol was registered on ClinicalTrials.gov (ID: NCT04883879).