Special Articles on Evolution of Lifestyle-supporting Estimation Services
Immunity Estimation AI for Immunity Self-care Using Smartphone Logs

Healthcare Behavior Change Immunity

Shota Kobayashi, Takafumi Yamauchi and Satoshi Hiyama
X-Tech Development Department

Abstract
After the COVID-19 pandemic and other events, people's interest in and awareness of immunity has increased. Although the importance of enhancing immunity is generally known, it is difficult to assess immunity in daily life because immunity is assessed by saliva and blood tests. Focusing on the relationship between lifestyle habits and immunity, NTT DOCOMO has developed an immunity estimation AI using smartphone logs that can express lifestyle habits. This technology is expected to lead to behavioral changes for encouraging immunity self-care by enabling the identification of improving or declining trends in immunity from daily smartphone use and by providing the immunity estimating AI to suggest lifestyle improvements for individuals.

01. Introduction

  • NTT DOCOMO is focusing on businesses in the healthcare and medical ...

    Open

    NTT DOCOMO is focusing on businesses in the healthcare and medical fields to realize a society in which everyone can maintain and improve their health and is developing “d-Healthcare [1]” and “Kenko-mileage [2]” (Health mileage) as services to support health management and health promotion. By analyzing smartphone logs that can be obtained from a large customer base and customer contact points, NTT DOCOMO has built technology to estimate the health status and disease risk of users on a daily basis with little burden and naturally guide them to health actions and is implementing this functionality in its services. So far, NTT DOCOMO has developed and deployed AIs to estimate people's health status from smartphone logs, such as stress level estimation AI [3], frailty estimation AI [4], and AI estimation of habits that raise blood pressure [5]. We are currently further developing technology to estimate various health conditions to promote health actions.

    After the recent COVID-19 pandemic and other events, people of all generations are becoming increasingly concerned and aware of their immunity [6]. It has been suggested that increasing mucosal immunity in the oral cavity may contribute to the prevention of COVID-19 infections [7][8]. It is thus important to enhance one's own immunity on a daily basis to prevent infection with viruses and bacteria. However, while the importance of enhancing immunity is known, it is difficult to assess immunity in daily life because immunity is assessed by saliva and blood tests.

    Based on the relationship between immunity and lifestyle habits such as daily exercise and sleep, and on the hypothesis that immunity is related to weather conditions, NTT DOCOMO has developed an immunity estimation AI (hereafter referred to as “this AI”) using smartphone logs that can express lifestyle habits. With this AI, it is possible to determine whether a user's immunity is improving or declining based on daily smartphone use. Furthermore, by proposing lifestyle improvements to enhance immunity, it is possible to encourage users to change their behaviors for self-care of their immunity.

    This article describes the level of secretory Immunoglobulin A (sIgA) in saliva as an indicator for evaluating immunity, an overview of this AI, and its functional implementation in commercial services.

    • 02. A Key Immunity Index

  • Having experienced the COVID-19 pandemic, it can be said that ...

    Open

    Having experienced the COVID-19 pandemic, it can be said that immunity has gained widespread interest in terms of preventing the acquisition of infectious diseases. The sIgA level is an indicator of immunity functioning to prevent infectious diseases caused by viruses and bacteria. sIgA is an antibody that prevents various pathogens from entering the body [9]. For example, studies focusing on sIgA have shown that when the sIgA level is low, the risk of contracting upper respiratory infections (including the common cold) increases [10].

    sIgA can be quantitatively evaluated by saliva testing, which has less bodily invasiveness*1 than immunological tests that involve blood sampling. As described in this article, in the development of this AI, it is necessary to measure immunity indices continuously and multiple times, and it is relatively easy to collect correct values if the sIgA level is measured.

    sIgA has also been reported as being closely related to lifestyle. For example, a study [11] that grouped elderly people by average daily number of steps and investigated the relationship with sIgA secretion showed that sIgA secretion was highest in the group with an average daily number of steps of approximately 7,000 steps (Figure 1(a)). In a study investigating the relationship between sleep duration and sIgA secretion [12], it was shown that sIgA secretion was higher in the optimal sleep group with 6 to 8 hours of sleep compared to the short sleep group with 5 hours or less and the long sleep group with 9 hours or more (Fig. 1(b)). These studies indicate that sIgA secretion varies with average number of steps and sleep duration.

    In addition to this, we can also focus on the relationship with weather conditions. In particular, passive stress due to weather conditions changes may alter the sIgA level [13][14].

    NTT DOCOMO adopted the sIgA level as an evaluation index for this AI, focusing on the ease of quantitative evaluation of sIgA and its relationship to lifestyle and weather conditions.

    Figure 1 Previous studies showing the relationship between lifestyle and sIgA
    1. Invasiveness: The degree to which a stimulus or action produces a change in the organism. High invasiveness increases the burden and pain on the organism.

    • 03. Overview of Immunity Estimation AI

  • 3.1 Construction of Immunity Estimation AI

    Open

    An overview of this AI is shown in Figure 2. Smartphone logs related to lifestyle habits and sIgA levels, the data set for the construction of this AI, were collected continuously for approximately one month with consent from approximately 160 men and women in their 20s to 60s. Smartphone logs that were assumed to be related to lifestyle habits, including sleep, walking, location information, and usage of all installed applications, were collected. At the same time, attribute information and information related to the weather conditions in users' areas of residence were also collected.

    After data collection, sIgA levels and smartphone logs were analyzed in NTT DOCOMO's data analysis environment.

    The collected sIgA level data were analyzed for time-series variation and labeled to indicate an increase or decrease, and an objective variable*2 was created. Individual differences were observed in the absolute values of the sIgA levels measured here and in the percentage of variation. Therefore, there may be individual differences in how far sIgA levels can be raised. In self-care of immunity, it is important to capture relative changes in immunity in individuals to improve lifestyle and prevent infectious diseases, etc. Therefore, we decided to evaluate the relative variation in sIgA levels from individual to individual.

    In addition, several explanatory variables*3 were created from collected smartphone logs and weather information that were thought to be related to the variation in sIgA levels. These explanatory variables refer to lifestyle and environmental factors that may be associated with variation of immunity, for example, length of sleep, regularity of lifestyle, and air temperature differences.

    This AI was constructed by learning the relationship between the created objective variables and explanatory variables through machine learning*4. The estimation performance of the estimation model was evaluated, and both sensitivity*5 and specificity*6 were approximately 0.7. The relationship between explanatory variables indicating lifestyle habits, such as number of steps and sleep duration, and variation in sIgA levels was comparable to known relationships, as shown in the literature [11][12]. In the development process of this AI, Dr. Kazuhiro Shimizu of the Japan Sport Council supervised the development from an immunological perspective.

    After the construction of this AI, even without information on the sIgA level available through saliva sample collection, estimation can be performed simply by providing the smartphone logs and weather information needed to create the explanatory variables. Specifically, information indicating whether the sIgA level, an indicator of immunity, is improving or declining under lifestyle habits and environmental factors derived from smartphone logs and weather information can be obtained from this AI.

    Figure 2 Overview of immunity estimation AI

    3.2 Feedback for Behavioral Changes

    This AI outputs information about improving or declining trends in immunity based on smartphone logs and weather information. This enables users to understand trends in the fluctuations in their immunity in their normal daily life. However, understanding fluctuations in immunity alone does not lead to immunity self-care to counteract viral and other infections. Information on how to improve immunity is also needed.

    In this regard, in addition to outputting information on whether immunity is trending up or down, this AI also offers suggestions for improving immunity in the actual conditions of the user's lifestyle. Specifically, explainable AI*7 technology is used to identify the explanatory variables that are influencing the decline in immunity for each user. Among these explanatory variables, items that can be improved by the user's actions are extracted, and lifestyle habits based on the explanatory variables that contribute most to the decline in immunity are presented to the user. This will encourage behavioral change toward immunity self-care. Voluntary immunity self-care can be expected as users improve these lifestyle habits (Figure 3).

    Figure 3 Feedback image of the immunity estimation AI
    1. Objective variables: The data that the machine learning (see *4) model trying to predict (e.g., house prices or product sales).
    2. Explanatory variables: In machine learning (see *4) models, data used to predict an object variable (e.g., area or number of rooms in a house).
    3. Machine leaning: A mechanism allowing a computer to learn the relationship between inputs (explanatory variables) and outputs (objective variables) through statistical processing of example data.
    4. Sensitivity: In this article, it presents the percentage of data in which sIgA levels are estimated to be decreasing by an estimation model among the data in which sIgA levels actually tend to decrease.
    5. Specificity: In this article, it presents the percentage of data in which sIgA levels are estimated to be increasing by an estimation model among the data in which sIgA levels actually tend to increase.
    6. Explainable AI: Also called XAI (Explainable AI). Technology that provides human-interpretable reasons or rationales for AI output.

    • 04. Providing Functionality to Commercial Services

  • On May 8, 2023, COVID-19, which has had a major impact on people's ...

    Open

    On May 8, 2023, COVID-19, which has had a major impact on people's lives for several years since early 2020, became a Category V infectious disease under the Act on the Prevention of Infectious Diseases and Medical Care for Patients with Infectious Diseases (Infectious Diseases Control Law) in Japan [15]. With this transition, municipalities are now required to use their own independent judgment in taking measures to prevent the transmission of COVID-19. NTT DOCOMO has been developing “Kenko-mileage”, a health management and health promotion service for local residents of municipalities and employees of companies, and has implemented this AI functionality in this service, which was launched commercially on October 30, 2023 [16]. This is expected to help in the fight against infectious disease after the transition to a Category V infectious disease.

    This commercial offering utilizes the HealthTech platform built and operated by NTT DOCOMO. The HealthTech platform aggregates NTT DOCOMO's AIs that estimate health status and lifestyle habits, such as this AI, the AI estimation of habits that raise blood pressure, and the frailty estimation AI. Through Application Programming Interface (API) linkage*8 with this platform, NTT DOCOMO can provide its services and business partners with the estimation AI functions on the HealthTech platform (Figure 4). There are two ways of providing estimation AI functions on the HealthTech platform to business partners: through services and apps provided by NTT DOCOMO, and through direct API linkage with business partners' own services and apps (Figure 5). As mentioned above, NTT DOCOMO has already provided this AI functionality to the “Kenko-mileage” service and has started to provide it to “d-Healthcare” service from December 2024 [17]. In addition, as a business partner, NTT DOCOMO has begun providing this AI functionality to Bellsystem24 Holdings, Inc.'s “Zutool” (Zutsu (headache in Japanese) + tool) health management application [18] and plans to provide the functionality through “Kenko-mileage” to KAGOME CO., LTD. [19] and Sustainable Pavilion 2025 Inc. [20].

    In addition to providing value to municipalities in the form of countermeasures against infectious diseases, this AI has the potential to provide value in the form of improved Quality of Life (QoL)*9 by preventing the incidence of infectious diseases, as well as attracting users to the immune-boosting food products market. In this way, we believe that there are possibilities to collaborate with various business partners.

    Figure 4 Providing estimation AI to business partners through the HealthTech platform, Figure 5 HealthTech platform offerings
    1. API linkage: Linkage of different programs or software through a predefined interface.
    2. QoL: A concept that refers to quality of life and evaluates the richness and comfort of life, including health, happiness, financial situation, and social satisfaction.

    • 05. Conclusion

  • This article described the sIgA level as an index for ...

    Open

    This article described the sIgA level as an index for evaluating immunity, an overview of this AI, and the implementation of this functionality in commercial services. In the future, we will accumulate evidence on this AI by further improving the accuracy of its model, verifying whether it led users to behavioral changes oriented toward immune self-care, and effects in terms of reducing medical and nursing care costs for municipalities. In addition, by providing technology to business partners through the HealthTech platform and conducting demonstration experiments, we will collect evidence of value provision to industries outside the healthcare domain and promote the possibility of providing value not only to municipalities and their residents in the area of immunity care, but also to other industries.

  • REFERENCES

    Open
    1. [1] NTT DOCOMO: “d-Healthcare” (in Japanese).
      New windowhttps://health.docomo.ne.jp/
    2. [2] NTT DOCOMO: “Kenko-mileage” (in Japanese).
      New windowhttps://kenko-mileage.jp/
    3. [3] N. Yamamoto et al.: “Smartphone Log-based Stress and Cognitive Performance Estimation Technology,” NTT DOCOMO Technical Journal, Vol. 22, No. 4, pp. 34–42, Apr. 2021.
      PDFhttps://www.docomo.ne.jp/english/binary/pdf/corporate/technology/rd/technical_journal/bn/vol22_4/vol22_4_005en.pdf (PDF format:1,862KB)
    4. [4] T. Yamauchi et al.: “Frailty Estimation AI that Uses Smartphone Logs to Reduce the Risk of Needing Aged Care,” NTT DOCOMO Technical Journal, Vol. 24, No. 4, Jun. 2023.
      https://www.docomo.ne.jp/english/corporate/technology/rd/technical_journal/bn/vol24_4/005.html
    5. [5] Y. Ikezoe et al.: “AI Estimation of Habits that Raise Blood Pressure Using Smartphone Logs,” NTT DOCOMO Technical Journal, Vol. 25, No. 4, Apr. 2024.
      https://www.docomo.ne.jp/english/corporate/technology/rd/technical_journal/bn/vol25_4/003.html
    6. [6] Kirin Holdings Company, Limited.: “Survey of Attitudes about ‘Immunity’,” Dec. 2021 (in Japanese).
      New windowhttps://www.kirinholdings.com/jp/newsroom/release/2021/1214_01.html
    7. [7] K. Hamuro, H. Saito, T. Saito and N. Kohda: “Identification of antigens recognized by salivary IgA using microbial protein microarrays,” Bioscience of Microbiota, Food and Health, Vol. 41, No. 4, pp. 177–184, 2022.
      New windowhttps://doi.org/10.12938/bmfh.2022-014
    8. [8] K. Tsukinoki, T. Yamamoto, K. Handa, M. Iwamiya, J. Saruta, S. Ino and T. Sakurai: “Detection of cross-reactive immunoglobulin A against the severe acute respiratory syndrome-coronavirus-2 spike 1 subunit in saliva,” PLOS ONE, Vol. 16, No. 11, Nov. 2021.
      New windowhttps://doi.org/10.1371/journal.pone.0249979
    9. [9] P. Brandtzaeg, I. N. Farstad, F.-E. Johansen, H. C. Morton, I. N. Norderhaug and T. Yamanaka: “The B-cell system of human mucosae and exocrine glands,” WILEY Online Library, Immunological Reviews, Vol. 171, No. 1, pp. 45–87, Oct. 1999.
      New windowhttps://doi.org/10.1111/j.1600-065x.1999.tb01342.x
    10. [10] V. Neville, M. Gleeson and J. P. Folland: “Salivary IgA as a Risk Factor for Upper Respiratory Infections in Elite Professional Athletes,” Medicine & Science in Sports & Exercise, Vol. 40, No. 7, pp. 1228–1236, Jul. 2008.
      New windowhttps://doi.org/10.1249/mss.0b013e31816be9c3
    11. [11] K. Shimizu, F. Kimura, T. Akimoto, T. Akama, S. Kuno and I. Kono: “Effect of Free-Living Daily Physical Activity on Salivary Secretory IgA in Elderly,” Medicine & Science in Sports & Exercise, Vol. 39, No. 4, pp. 593–598, Apr. 2007.
      New windowhttps://doi.org/10.1249/mss.0b013e318031306d
    12. [12] H. Okamura, A. Tsuda, J. Yajima, S. Horiuchi and T. Matsuishi: “Sleeping time relates to perceived health and psychoneuroimmunological responses,” Japanese Journal of Behavioral Medicine, Vol. 15, No. 1, pp. 33–40, 2010 (in Japanese).
      New windowhttps://doi.org/10.11331/jjbm.15.33
    13. [13] M. Mimura and T. Ikaga: “Human response to the indoor environment under fluctuating temperature,” Science and Technology for the Built Environment, Vol. 22, No. 6, pp. 820–830, Jun. 2016.
      New windowhttps://doi.org/10.1080/23744731.2016.1184550
    14. [14] S. Fujihara and M. Yogo: “Time series changes in salivary secretory immunoglobulin A to acute stress,” Japanese Journal of Physiological Psychology and Psychophysiology, Vol. 29, No. 3, pp. 193–203, Dec. 2011 (in Japanese).
      New windowhttps://doi.org/10.5674/jjppp.1115oa
    15. [15] Ministry of Health, Labour, and Welfare: “Response to COVID 19 (Novel Coronavirus) after the classification change,” 2023.
      New windowhttps://www.mhlw.go.jp/stf/covid-19/kenkou-iryousoudan_00006.html
    16. [16] NTT DOCOMO Press Release: “Development of ‘Immunity Estimation AI’ to Estimate and Predict Changes in Immunity Based on Data Stored in Smartphones,” May 2023 (in Japanese).
      PDFhttps://www.docomo.ne.jp/binary/pdf/info/news_release/topics_230508_00.pdf (PDF format:931KB)
    17. [17] NTT DOCOMO Press Release: “Subscription Fees for ‘d-Healthcare’ Revised and New Health Scoring Feature Added,” Sep. 2024 (in Japanese).
      https://www.docomo.ne.jp/info/news_release/2024/09/25_01.html
    18. [18] NTT DOCOMO Press Release: “New Functionality Using Immunity Estimation AI added to ‘Zutool for Sugotoku’ in ‘Sugotoku Contents’,” Dec. 2024 (in Japanese).
      PDFhttps://www.docomo.ne.jp/binary/pdf/info/news_release/topics_241203_00.pdf (PDF format:684KB)
    19. [19] NTT DOCOMO Press Release: “DOCOMO to Exhibit the Use Case ‘Wakuwaku! Colorful VR Game with Vegetable Consumption Estimation!,’ Jointly Developed by KAGOME and NTT DOCOMO and Adopted by the Ministry of Economy, Trade and Industry's ‘FY2023 Supplementary Project for Acceleration of Personal Health Record (PHR) Social Implementation,’ at Expo 2025 Osaka, Kansai, Japan in April 2025 - Contributing to Healthcare Services in Japan with the Power of Vegetables -,” Oct. 2024 (in Japanese).
      PDFhttps://www.docomo.ne.jp/binary/pdf/info/news_release/topics_241008_00.pdf (PDF format:816KB)
    20. [20] NTT DOCOMO Press Release: “DOCOMO Announces Overview of the Demonstration Use Case ‘Talkable PHR: Health Chat with Yourself’ in the Ministry of Economy, Trade and Industry's ‘FY2023 Supplementary Project for Acceleration of PHR Social Implementation’ to be Exhibited at Expo 2025 Osaka, Kansai, Japan - A Society where People can be Healthy Naturally through PHR -,” Oct. 2024 (in Japanese).
      PDFhttps://www.docomo.ne.jp/binary/pdf/info/news_release/topics_241008_01.pdf (PDF format:928KB)

VOL.26 NO.4

Go to top of the page