Quantifying Human Biological Age: A Machine Learning Approach

Quantifying human biological age is an important and difficult challenge. Different biomarkers and numerous approaches have been studied for biological age prediction, each with its advantages and limitations. In this work, we first introduce a new anthropometric measure (called Surface-based Body Shape Index, SBSI) that accounts for both body shape and body size, and evaluate its performance as a predictor of all-cause mortality. We analyzed data from the National Health and Human Nutrition Examination Survey (NHANES). Based on the analysis, we introduce a new body shape index constructed from four important anthropometric determinants of body shape and body size: body surface area (BSA), vertical trunk circumference (VTC), height (H) and waist circumference (WC). SBSI is generally linear with age, and increases with increasing mortality, when compared with other popular anthropometric indices of body shape. We then investigate whether human body shape can be exploited for reliable age estimation for adult humans. We introduce a new multi-stage approach, based on human body measurements. Specifically, we develop an eigen body shape model, and use this to perform body shape clustering. Each cluster contains individuals with similar body shapes as captured by the eigen body shape model. First, we perform initial age estimation based on the body shape model. This initial estimate is then used to assign the subject into a probable age group. The second stage of estimation is then performed by using a specific estimation model as determined by the age group and body shape model. We then apply information from the neighborhood context to further improve estimation accuracy and stability. Experimental results show that, with appropriate modeling, human body shape can be used in human age estimation. We obtain a mean absolute error (MAE) of 5.90 years on the NHANES dataset, using 10-fold cross-validation. We then study the question of whether blood biomarkers can be used for reliable biological age estimation. We propose a new biological age estimation method, and investigate the performance of the new method against popular biological age estimation methods. We introduce a centroid based approach, using the notion of age neighborhoods. Specifically, we develop a model, based on which we compute biological age using blood biomarkers. Compared with current popular methods for biological age prediction, our results show that, the proposed age neighborhood model is robust, and results in improved performance in human biological age prediction. Furthermore, we investigate whether human locomotor activity can be exploited for biological age estimation for adult humans. We introduce an approach based on deep convolutional long short term memory (ConvLSTM) to predict biological age, using human physical activity as recorded by a wearable device. We consider five deep biological age estimation models including the proposed approach and compare their performance on the NHANES physical activity dataset. Results on mortality hazard prediction using both the Cox proportionality hazard model and Kaplan-Meier curves each show that the proposed method for estimating biological age outperforms other state-of-the-art approaches. This work has significant implications in various fields, such as health assessment, forensic science, biometrics, security, and in vaccination and immunization when the true age of the subject is unknown. Our work also has implications in combining wearable sensors and deep learning techniques for improved health monitoring, for instance, in a mobile health environment