Recent years have seen a significant rise in the development and implementation of advanced artificial intelligence (AI) and machine learning (ML) tools and algorithms in health care settings and medicine.1,2 For instance, a large language model (LLM), is a form of AI trained extensively on vast volumes of natural language texts, capable of generating human-like conversations.3 Although still controversial,4 preliminary evidence strongly indicates that these conversational AI models, if not yet surpassing, exhibit performance levels at least comparable to human capabilities, particularly in medical applications. For instance, they can generate more empathetic responses to patient queries than physicians5 and even provide patient diagnoses as accurately as a trained doctor.6 However, despite their great promise, the full potential of AI and ML remains largely underutilized in physical activity (PA) and sedentary behavior research. Here, we highlight the capabilities of AI and ML as alternative methods and approaches possessing the complexity required to further advance the study of these naturally complex behaviors.
How Can AI and ML Contribute to PA and Sedentary Behavior Research?
To date, AI and ML techniques have been predominately used for improved measurement of PA and sedentary behaviors from data collected by wearable activity monitors.7,8 However, an infrequently explored aspect of ML in PA and sedentary behavior research is its capacity for hypothesis generation without predefined notions, generating novel, and data-driven hypotheses. This capability has proven useful, notably in studies focused on identifying and studying PA and sedentary behavior profiles. Several studies utilizing ML techniques to establish profiles of PA and sedentary behaviors—often accommodating a wide range of variables describing the duration, patterns, and temporal aspects of these behaviors9,10—have successfully facilitated the identification of previously unnoticed movement behavior patterns9–11 (eg, active couch potatoes11). ML techniques have also been widely used for researching factors linked to physical activity, particularly to create hierarchical models predicting and associated with physical activity and sedentary behaviors.12–14
AI and ML algorithms can now detect disease from medical images at the expert level.15 These promising advances in prediction accuracy, coupled with the proliferation of wearable devices used to measure high-quality PA and sedentary behavior data,7,8 further inspire the utilization of AI and ML approaches. The digital, noninvasive signals obtained from wearable activity monitors encapsulate traits and phenotypes16 capable of predicting numerous health indicators and outcomes. These movement behavior traits and phenotypes potentially enable the early identification of future health risks and diseases, providing an unprecedented opportunity to initiate timely interventions to facilitate disease prevention. For example, ML models trained using accelerometry data could predict Parkinson’s disease years before clinical diagnosis, outperforming other established biomarkers (genetics, lifestyle, etc) in predicting Parkinson disease in the general population.17 Emerging evidence suggests ML from wearable movement-tracking data can predict mortality risks18 and anxiety disorder symptoms,19 among other health indicators.20,21
AI and ML offer opportunities to advance and accelerate precision medicine.22 Precision medicine utilizes AI and ML for tailored and precise interventions based on individualized characteristics, genetics, lifestyles, and specific health indicators22—an opportunity that could be used to transcend the current one-size-fits-all recommendations for PA and sedentary behaviors.23 In parallel disciplines like nutrition, precision nutrition24 has already been widely recognized as an emerging area of research, utilizing AI and ML to formulate highly personalized recommendations. This recognition stems from the understanding that people may respond differently to specific foods and nutrients. A similar assumption holds for PA and sedentary behaviors—people have different capacities, inclinations, and opportunities to be physically active.
Now, the Question Is as Follows: Why Have AI and ML Not Yet Been Widely Used in Research on PA and Sedentary Behaviors?
Implementing AI and ML techniques is typically more time-consuming than performing other forms of statistical analyses and may also require substantial computing power. In addition, while AI and ML show promise in advancing medical and health research, the ethical issues surrounding these techniques can be complex.25 For instance, individuals whose data are utilized must have the right to determine ownership of their data, specify where it should be stored, and dictate its permissible uses. Hence, the complexity of AI and ML makes the details of these techniques difficult to translate into humanly understandable terms. Existing literature has also highlighted ethical concerns regarding the utilization of AI and ML algorithms in medical and health research. These include the potential exacerbation of existing biases and inequalities within health care systems.26
AI and ML have 2 notable limitations, which are significant when utilizing these tools and techniques. First, AI and ML algorithms seek patterns in the data or learn the best patterns for predicting an outcome without assuming any prior knowledge.27 This can introduce a conflict between how humans and machines perceive these patterns. Second, AI and ML tools have limited capabilities for identifying and understanding cause–effect relationships between variables, which are crucial aspects in studying PA and sedentary behaviors. Causal ML is an emerging research field aiming to overcome these limitations.28 Similarly, another active research field is designing and developing explainable AI and interpretable ML learning approaches,27 which may mitigate such limitations. While still in their early stages, such efforts may eventually lead to increased adoption of AI and ML tools and algorithms in medicine, including in the research field of PA and sedentary behavior.
Conclusions
Most enduring challenges in the research field of PA and sedentary behavior, such as personalized recommendations, research on health consequences, and the investigation of correlates and determinants of physical inactivity and sedentary lifestyle, require more innovative tools and approaches to accelerate research progress. Integrating ML and AI tools and algorithms, renowned for their computation power and capacity to handle complex data is, therefore, likely to further advance research progress. As advanced AI and ML methodologies continue to evolve, leveraging these techniques to analyze both PA and sedentary behavior data is likely to yield a deeper understanding of these behaviors.
References
- 1.↑
Haug CJ, Drazen JM. Artificial intelligence and machine learning in clinical medicine, 2023. N Engl J Med. 2023;388(13):1201–1208.
- 2.↑
Rajpurkar P, Chen E, Banerjee O, Topol EJ. AI in health and medicine. Nat Med. 2022;28(1):31–38. doi:
- 3.↑
Thirunavukarasu AJ, Ting DSJ, Elangovan K, Gutierrez L, Tan TF, Ting DSW. Large language models in medicine. Nat Med. 2023;29(8):1930–1940. doi:
- 4.↑
Seitz L, Bekmeier-Feuerhahn S, Gohil K. Can we trust a chatbot like a physician? A qualitative study on understanding the emergence of trust toward diagnostic chatbots. Int J Hum Comput Stud. 2022;165:102848. doi:
- 5.↑
Ayers JW, Poliak A, Dredze M, et al. Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum. JAMA Intern Med. 2023;183(6):589–596. doi:
- 6.↑
ten Berg H, van Bakel B, van de Wouw L, et al. ChatGPT and generating a differential diagnosis early in an emergency department presentation. Ann Emerg Med. 2024;83(1):83–86. doi:
- 7.↑
Farrahi V, Niemelä M, Kangas M, Korpelainen R, Jämsä T. Calibration and validation of accelerometer-based activity monitors: a systematic review of machine-learning approaches. Gait Posture. 2019;68:285–299. doi:
- 8.↑
Narayanan A, Desai F, Stewart T, Duncan S, Mackay L. Application of raw accelerometer data and machine-learning techniques to characterize human movement behavior: a systematic scoping review. J Phys Act Health. 2020;17(3):360–383.
- 9.↑
Farrahi V, Kangas M, Kiviniemi A, Puukka K, Korpelainen R, Jämsä T. Accumulation patterns of sedentary time and breaks and their association with cardiometabolic health markers in adults. Scand J Med Sci Sports. 2021;31(7):1489–1507. doi:
- 10.↑
Niemelä M, Kangas M, Farrahi V, et al. Intensity and temporal patterns of physical activity and cardiovascular disease risk in midlife. Prev Med. 2019;124:33–41. doi:
- 11.↑
Farrahi V, Rostami M, Dumuid D, et al. Joint profiles of sedentary time and physical activity in adults and their associations with cardiometabolic health. Med Sci Sports Exerc. 2022;54(12):2118–2128. doi:
- 12.↑
Farrahi V, Niemelä M, Kärmeniemi M, et al. Correlates of physical activity behavior in adults: a data mining approach. Int J Behav Nutr Phys Act. 2020;17(1):94. doi:
- 13.
Biswas A, Chen C, Dobson KG, et al. Identifying the sociodemographic and work-related factors related to workers’ daily physical activity using a decision tree approach. BMC Public Health. 2023;23(1):1853.
- 14.↑
Ezeugwu VE, Mandhane PJ, Hammam N, et al. Influence of neighborhood characteristics and weather on movement behaviors at age 3 and 5 years in a longitudinal birth cohort. J Phys Act Health. 2021;18(5):571–579. doi:
- 15.↑
Esteva A, Kuprel B, Novoa RA, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115–118. doi:
- 16.↑
Jeong H, Jeong YW, Park Y, Kim K, Park J, Kang DR. Applications of deep learning methods in digital biomarker research using noninvasive sensing data. Digit Health. 2022;8:20552076221136640. doi:
- 17.↑
Schalkamp AK, Peall KJ, Harrison NA, Sandor C. Wearable movement-tracking data identify Parkinson’s disease years before clinical diagnosis. Nat Med. 2023;29(8):2048–2056. doi:
- 18.↑
Chen M, Landré B, Marques-Vidal P, et al. Identification of physical activity and sedentary behaviour dimensions that predict mortality risk in older adults: development of a machine learning model in the Whitehall II accelerometer sub-study and external validation in the CoLaus study. eClinicalMedicine. 2023;55:101773. doi:
- 19.↑
Jacobson NC, Lekkas D, Huang R, Thomas N. Deep learning paired with wearable passive sensing data predicts deterioration in anxiety disorder symptoms across 17–18 years. J Affect Disord. 2021;282:104–111. doi:
- 20.↑
Park C, Mishra R, Golledge J, Najafi B. Digital biomarkers of physical frailty and frailty phenotypes using sensor-based physical activity and machine learning. Sensors. 2021;21(16):5289. doi:
- 21.↑
De Vos M, Prince J, Buchanan T, FitzGerald JJ, Antoniades CA. Discriminating progressive supranuclear palsy from Parkinson’s disease using wearable technology and machine learning. Gait Posture. 2020;77:257–263.
- 22.↑
König IR, Fuchs O, Hansen G, von Mutius E, Kopp MV. What is precision medicine? Eur Respir J. 2017;50(4):1700391. doi:
- 23.↑
Bevington F, Piercy KL, Olscamp K, Hilfiker SW, Fisher DG, Barnett EY. The move your way campaign: encouraging contemplators and families to meet the recommendations from the physical activity guidelines for Americans. J Phys Act Health. 2020;17(4):397–403. doi:
- 24.↑
Rodgers GP, Collins FS. Precision nutrition—the answer to “what to eat to stay healthy.” JAMA. 2020;324(8):735–736. doi:
- 25.↑
Martinho A, Kroesen M, Chorus C. A healthy debate: exploring the views of medical doctors on the ethics of artificial intelligence. Artif Intell Med. 2021;121:102190.
- 26.↑
Anom BY. Ethics of Big Data and artificial intelligence in medicine. Ethics Med Public Health. 2020;15:100568.
- 27.↑
Confalonieri R, Coba L, Wagner B, Besold TR. A historical perspective of explainable Artificial Intelligence. WIREs Data Min Knowl Discov. 2021;11(1):e1391. doi:
- 28.↑
Sanchez P, Voisey JP, Xia T, Watson HI, O’Neil AQ, Tsaftaris SA. Causal machine learning for healthcare and precision medicine. R Soc Open Sci. 2022;9(8):220638.
