Machine Learning Used in Biomedical Computing and Intelligence Healthcare, Volume II

Epigenetic clock, a highly accurate age estimator based on DNA methylation (DNAm) level, is the basis for predicting mortality/morbidity and elucidating the molecular mechanism of aging, which is of great significance in forensics, justice, and social life. Herein, we integrated machine learning (ML) algorithms to construct blood epigenetic clock in Southern Han Chinese (CHS) for chronological age prediction. The correlation coefficient (r) meta-analyses of 7,084 individuals were firstly implemented to select five genes (ELOVL2, C1orf132, TRIM59, FHL2, and KLF14) from a candidate set of nine age-associated DNAm biomarkers. The DNAm-based profiles of the CHS cohort (240 blood samples differing in age from 1 to 81 years) were generated by the bisulfite targeted amplicon pyrosequencing (BTA-pseq) from 34 cytosine-phosphate-guanine sites (CpGs) of five selected genes, revealing that the methylation levels at different CpGs exhibit population specificity. Furthermore, we established and evaluated four chronological age prediction models using distinct ML algorithms: stepwise regression (SR), support vector regression (SVR-eps and SVR-nu), and random forest regression (RFR). The median absolute deviation (MAD) values increased with chronological age, especially in the 61–81 age category. No apparent gender effect was found in different ML models of the CHS cohort (all p > 0.05). The MAD values were 2.97, 2.22, 2.19, and 1.29 years for SR, SVR-eps, SVR-nu, and RFR in the CHS cohort, respectively. Eventually, compared to the MAD range of the meta cohort (2.53–5.07 years), a promising RFR model (ntree = 500 and mtry = 8) was optimized with an MAD of 1.15 years in the 1–60 age categories of the CHS cohort, which could be regarded as a robust epigenetic clock in blood for age-related issues.

Melanoma is a skin disease with a high fatality rate. Early diagnosis of melanoma can effectively increase the survival rate of patients. There are three types of dermoscopy images, malignant melanoma, benign nevis, and seborrheic keratosis, so using dermoscopy images to classify melanoma is an indispensable task in diagnosis. However, early melanoma classification works can only use the low-level information of images, so the melanoma cannot be classified efficiently; the recent deep learning methods mainly depend on a single network, although it can extract high-level features, the poor scale and type of the features limited the results of the classification. Therefore, we need an automatic classification method for melanoma, which can make full use of the rich and deep feature information of images for classification. In this study, we propose an ensemble method that can integrate different types of classification networks for melanoma classification. Specifically, we first use U-net to segment the lesion area of images to generate a lesion mask, thus resize images to focus on the lesion; then, we use five excellent classification models to classify dermoscopy images, and adding squeeze-excitation block (SE block) to models to emphasize the more informative features; finally, we use our proposed new ensemble network to integrate five different classification results. The experimental results prove the validity of our results. We test our method on the ISIC 2017 challenge dataset and obtain excellent results on multiple metrics; especially, we get 0.909 on accuracy. Our classification framework can provide an efficient and accurate way for melanoma classification using dermoscopy images, laying the foundation for early diagnosis and later treatment of melanoma.