## Internet of Things (IoT) System and Field Sensors for Exercise Intensity Measurements

### Abstract

Although exercise training according to individual peak aerobic capacity ( o2peak) has been recommended at all ages, sensors available in the field are limited. The most popular sensors in the field are pedometers, but they cannot be used to monitor exercise intensity. Instead, although heart rate (HR) monitors are broadly available in the field to estimate exercise intensity, HR responses to exercise vary by individual according to physical fitness and environmental conditions, which hinders the precise measurement of energy expenditure. These issues make it difficult for exercise physiologists to collaborate with geneticists, nutritionists, and clinicians using the internet of things (IoT). To conquer these problems, we have developed a device that is equipped with a triaxial accelerometer and a barometer to measure energy expenditure during interval walking training (IWT) in the field with inclines. IWT is a training regimen to repeat fast and slow walking for 3 min each, equivalent to greater than 70% and approximately 40% of individual o2peak, respectively. Additionally, we developed an IoT system that enables users to receive instructions from trainers according to their walking records even if they live far away. Since the system is available at low cost with minimum personnel, we can investigate any factors affecting the adherence to and effects of IWT in a large population for a long period. This system was also used to verify any effects of nutritional supplements during IWT and to examine the value of applying IWT to clinical medicine. © 2020 American Physiological Society. Compr Physiol 10:1207‐1240, 2020.

 Figure 1. Our experimental strategy. We first conducted the walking program by instructing middle‐aged and older individuals to walk 10,000 steps per day. However, physical fitness did not increase after training, and adherence to the program was very low. On the other hand, standard exercise training tailored to an individual's fitness level is generally performed at a gymnasium using machines but is costly and of limited availability. To extend individualized training beyond the gymnasium for middle‐aged and older individuals, we developed a graded walking test to determine peak aerobic capacity ($V˙$o2peak) for walking using an accelerometer 125. Based on each subject's results, we then instructed them to perform interval walking training, which was remotely supervised via the internet. Our hypothesis is that an exercise training system composed of interval walking training and the IoT can increase physical fitness in older people. Reused, with permission, from Masuki S, et al., 2017 97. Figure 2. Comparisons between peak aerobic capacity ($V˙$o2peak) values determined by accelerometry (JD Mate) during the graded walking test (W$V˙$o2peak) and by respiratory gas analysis during the graded cycling test (C$V˙$o2peak) 125. Figure 3. Interval walking training and the e‐health link named “e‐Health Promotion System.” Participants visit a health care institution, a drug store, or a local community office every 2 weeks to transfer their walking records from the tracking device, JD Mate, to a central server computer via the internet. The server computer then provides them with a trend graph of their records along with advice automatically generated by the server. Based on this report, the staff, nurses, dietitians, or trainers give them individualized training advice. By anonymizing and combining the DNA data stored in a separate offline computer and the clinical data stored in the central server computer, we have begun to explore the genomic variations explaining inter‐individual variations in training response. The data from the research may be used to revise the e‐Health Promotion System to develop an algorithm to predict the effects of interval walking training on physical fitness and the indices of lifestyle‐related diseases in individuals with different physical and genetic characteristics. An e‐Key is used to limit a person's access to the database (DB). Squares on the internet circle indicate the firewall function. Solid arrows indicate online communications between users and the server via the internet, and dotted arrows indicate offline communications. Reused, with permission, from Nose H, et al., 2009 130. Figure 4. Percentage changes in isometric knee extension (FEXT) and flexion forces (FFLX) (A) and peak aerobic capacity by the graded walking test (W$V˙$o2peak) (B) after training in three groups: no‐walking training (no‐WT; males = 9, females = 37, total = 46); moderate‐intensity continuous walking (WCNT; males = 8, females = 43, total = 51); and high‐intensity interval walking (WINT; males = 11, females = 31, total = 42). *P < 0.05, **P < 0.01, ***P < 0.001: significant differences from the pretraining values. ††P < 0.01, †††P < 0.001: significant differences from the corresponding values in no‐WT. #P < 0.05. ##P < 0.01, ###P < 0.001: significant differences from the corresponding values in WCNT. Reused, with permission, from Nemoto K, et al., 2007 125. Figure 5. Changes in systolic (SBP) and diastolic blood pressures (DBP) at rest after training. †Significant differences from the corresponding values in no‐WT at the level of P < 0.05. The number of subjects and other symbols are the same as in Figure 4. Reused, with permission, from Nemoto K, et al., 2007 125. Figure 6. Total lifestyle‐related disease (LSD) score and peak aerobic capacity for walking ($V˙$o2peak) before and after interval walking training in males (A) and females (B). When the subjects were divided equally into three groups according to $V˙$o2peak, the score was lower in higher $V˙$o2peak groups. After interval walking training for 4 months, the score decreased as $V˙$o2peak increased in every group. *Versus before training at P < 0.05, #versus Middle group at P < 0.05, †versus High group at P < 0.05. Reused, with permission, from Morikawa M, et al., 2011 114. Figure 7. Relationship between total lifestyle‐related disease (LSD) score and peak aerobic capacity ($V˙$o2peak) in the low (circle), middle (square), and high $V˙$o2peak groups (triangle). The mean and SE bars for 66 males and for 156 females in each group are presented. The open and closed symbols indicate the values before and after interval walking (IWT), respectively, and data pairs are shown with arrows. Total LSD score (y) was significantly correlated with $V˙$o2peak (x) in males (R2 = 0.71, P < 0.001) and females (R2 = 0.94, P < 0.001) according to the mean values for 66 males and 156 females in each group before and after IWT. *Versus before training at P < 0.05, #versus middle group at P < 0.05, †versus high group at P < 0.05, before and after training, respectively. Reused, with permission, from Morikawa M, et al., 2011 114. Figure 8. Fast, slow, and total walking time over 5 months versus changes in peak aerobic capacity for walking (Δ$V˙$o2peak, A) and in lifestyle‐related disease score (ΔLSD score, B) after training in 679 middle‐aged and older subjects. Values are means ± SE. The number of subjects that were included in each bin of training achievements was also presented in panel (A). *Significant differences from pretraining values, P < 0.05. Reused, with permission, from Masuki S, et al., 2019 98. Figure 9. One representative gene, NFKB2, influenced by the IWT. (A) The genome‐wide methylation microarray assay was performed with microarray slides containing individual oligonucleotide probes (1–17). The ratio of methylation levels post‐/pre‐IWT is shown at sites corresponding to each probe. Open and closed bars indicate control and exercise groups, respectively. (B) The location of 17 probes used in a genome‐wide microarray assay to evaluate the methylation level of the NFKB2 promoter. (C‐E) Regression analysis of methylation status of the NFKB2 gene promoter region versus total energy consumption divided by body weight. The vertical axis shows the ratio value of the post‐IWT methylation/pre‐IWT methylation in the promoter region for probes 9 (C), 10 (D), and 12 (E). The correlation between methylation and energy consumption was analyzed using the Spearman nonparametric test. TSS, transcription start site; ○, nonexercise group; Δ, exercise group. Reused, with permission, from Zhang Y, et al., 2015 195. Figure 10. Subjects with type 2 diabetes were randomized to a control group (CON, white bars), continuous walking training group (CWT, striated bars), or interval walking training group (IWT, black bars). Intervention‐induced changes in glycemic control were assessed by examining post–preintervention changes in the following variables: mean 48‐h continuous glucose monitoring (CGM) glucose (A), minimum 48‐h CGM glucose (B), and maximum 48‐h CGM glucose (C). Data are presented as mean Δ values (post–preintervention values) ± SE. Significant differences from the pre‐intervention values, *P < 0.05 and **P < 0.01; significant differences between groups (a connecting line between bars), **P < 0.01. Reused, with permission, from Karstoft K, et al., 2013 71. Figure 11. Adherence to prescribed walking days (APWD, A) and fast walking time (APFWT, B) over the 22‐month training period versus change in lifestyle‐related disease score from baseline to 22 months (ΔLSD score). First, we calculated the LSD scores based on the Japanese and US healthcare guidelines 96. We assigned one point when a value met any of the following four criteria: (i) BMI ≥25 kg/m2; (ii) systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥85 mmHg; (iii) triglycerides ≥150 mg/dL, high‐density lipoprotein cholesterol less than 40 mg/dL or low‐density lipoprotein cholesterol ≥130 mg/dL; and (iv) blood glucose ≥110 mg/dL. The maximum total score was 4 points. Subjects were then pooled according to their ΔLSD score from baseline to 22 months as follows: ≥1 (men = 30; women = 74), 0 (men = 93; women = 270), −1 (men = 50; women = 128), −2 (men = 19; women = 25), and ≤ −3 (men = 4; women = 3). Significant differences in adherence from those with the worst ΔLSD score (≥1), *P < 0.05 and **P < 0.01. APWD was calculated as the number of walking days completed divided by the total number of walking days prescribed for each month (4 days/week), and then these monthly ratios were averaged over 22 months. APFWT was calculated as the fast walking time completed divided by the total fast walking time prescribed for each month (60 min/week), but time in excess of 60 min/week was regarded as 100%, and then these monthly ratios were averaged over 22 months. Reused, with permission, from Masuki S, et al., 2015 96. Figure 12. Adherence to prescribed walking days (APWD, A) and fast walking time (APFWT, B) over the 22‐month training period versus change in peak aerobic capacity for walking from baseline to 22 months (Δ$V˙$o2peak). Subjects were pooled according to their Δ$V˙$o2peak: ≤−2.5 (men = 8; women = 39), −2.4 to 0 (men = 46; women = 116), 0.1 to 2.5 (men = 52; women = 111), 2.6 to 5.0 (men = 52; women = 135), and greater than 5.0 mL/kg/min (men = 38; women = 99). Significant differences in adherence from those with the lowest Δ$V˙$o2peak (≤−2.5 mL/kg/min), **P < 0.01 and ***P < 0.001. APWD and APFWT determinations were the same as those described in Figure 11. Reused, with permission, from Masuki S, et al., 2015 96. Figure 13. Monthly adherence rate over the 22‐month training period in men according to the presence of the 334 allele of RS3 and C/T alleles of rs1042615 after adjustment for possible covariates. RS3 and rs1042615 denote a microsatellite and single‐nucleotide polymorphism in AVPR1A, respectively. Group 1, zero 334 alleles and CC or CT genotype (n = 89); group 2, one or two 334 alleles and CC/CT (n = 47); group 3, zero 334 alleles and TT (n = 31); and group 4, one or two 334 alleles and TT (n = 29). Adherence rate was calculated as the number of walking days completed divided by the total number of walking days prescribed for the month (4 days/week) and then adjusted for baseline BMI and smoking status using multiple regression analysis. §Group 1 to 3 versus 4. Reused, with permission, from Masuki S, et al., 2015 96. Figure 14. Percentage changes after training in hamstring muscle tissue area and isometric knee flexion force. Data are presented as the mean ± SE. CNT, interval walking training group; NUT, interval walking training plus macronutrient supplement group. *P < 0.05. Reused, with permission, from Okazaki K, et al., 2013 136. Figure 15. Changes in plasma volume (ΔPV′, A), plasma albumin content (ΔAlbcont′, B), and hemoglobin A1c (ΔHbA1c′, C) from baselines before starting additional 5‐month IWT. These values are corrected by ANCOVA with baseline plasma glucose concentration as a covariate. The means and SE bars are presented for 12 and 14 subjects in the CHO and the carbohydrate + whey protein (Pro‐CHO) groups, respectively. ##Versus baseline at P < 0.01. *Versus the CHO group at P < 0.05. Reused, with permission, from Uchida K, et al., 2018 186. Figure 16. Percentage changes after training in muscle strength (A) and methylation of the NFKB1 and NFKB2 promoter regions as assessed by pyrosequencing (B). The data were adjusted for pretraining values by ANCOVA. The mean and SE bars are presented for 12, 12, and 13 subjects in the IWT control (CNT), IWT + low‐dose (LD) and IWT + high‐dose milk product intake (HD) groups, respectively. **Significant differences from pretraining values, P < 0.01. Significant differences from the CNT group, †P < 0.05 and †††P < 0.001. Significant differences from the LD group, ‡P < 0.05 and ‡‡P < 0.01. (A) Average percent changes in isometric knee extension (ΔFEXT) and flexion (ΔFFLX) forces are presented. (B) Average percentage changes in CpG sites 1 to 7 for NFKB1 (upper) and average percentage changes in CpG sites 1 to 6 for NFKB2 (lower) are presented. Reused, with permission, from Masuki S, et al., 2017 100. Figure 17. Oxygen consumption rate ($V˙$o2), carbon dioxide production rate ($V˙$co2), ventilation volume ($V˙$E) (A), and plasma lactate concentration ([Lac−]p) (B) responses during graded cycling exercise under placebo (PLC; left) and ALA + SFC supplement intake conditions (right). The average value per minute is presented from rest to the highest workload of 75 W, at which all subjects could maintain the rhythm. Open symbols, before supplement intake; solid symbols, after supplement intake. Values are means ± SE of 10 subjects. *P < 0.05, **P < 0.01, and ***P < 0.001 versus before supplement intake. Reused, with permission, from Masuki S, et al., 2016 99. Figure 18. Training days (A), training impulse (B), and training time (C) at total, fast, and slow walking during the supplement intake period. Values are means ± SE of nine subjects. *P < 0.05 and ***P < 0.001 between the PLC and ALA + SFC trials. Reused, with permission, from Masuki S, et al., 2016 99. Figure 19. Montgomery‐Åsberg Depression Rating Scale (MADRS) sum score in the PLC trial (left) and the ALA + SFC trial (right) before and after a supplement intake period. Values are the means ± SE for nine subjects. **P < 0.01 versus before the supplement intake period. Reused, with permission, from Suzuki H, et al., 2018 175. Figure 20. Relationship between the oxygen consumption rate ($V˙$o2, mL/kg/min) and the vector magnitude of triaxial accelerations (VM) during graded walking tests on land and in a pool (water) before IWT. The values of 31 subjects from the WG and the LG in the land trial and the values of 16 subjects in the water trial from the WG for the last 30 s of the rest, slow, moderate, and fastest walking periods (3 min each) were pooled. Regression lines (solid lines) and area for 95% confidence limits (dotted lines) are shown. Reused, with permission, from Handa S, et al., 2016 51. Figure 21. Fast‐walking exercise intensity measured on the last walking day of every week during training. The means and SE bars include the results of 16 subjects in the WG and 15 subjects in the LG. Significant differences in $V˙$o2 between the WG and the LG, †P < 0.05 and ††P < 0.01; ##significant interactive effect of [training weeks × groups] on $V˙$o2 for the fast walking at P < 0.01. Reused, with permission, from Handa S, et al., 2016 51. Figure 22. Isometric knee extension (FEXT) (A) and flexion (FFLX) (B) forces before and after training, presented as the means and SE bars including the results from 16 and 15 subjects in the WG and the LG, respectively. Significant differences versus the pretraining values for each group, ***P < 0.001; significant interactive effects of [before and after training × groups] on FEXT and FFLX, #P < 0.05 and ##P < 0.01, respectively. Reused, with permission, from Handa S, et al., 2016 51. Figure 23. The relationship between the hourly frequency of starting IWT versus atmospheric temperature (Ta) above 25 °C during summer vacation. The hourly frequency was negatively correlated with Ta (r = −0.81, P = 0.008). Each symbol indicates the mean value for 20 subjects. CD, critical difference in Ta is shown on the x‐axis, and hourly frequency of starting IWT shown is on the y‐axis at the level of P < 0.05 using Fisher's LSD test. Reused, with permission, from Tanabe A, et al., 2018 179. Figure 24. An example of the contents of a smartphone application to motivate participants to continue IWT. (1) Recognition of progress: Based on individual walking and health records returned to each individual by the server computer, participants can recognize their achievements and the effects of training, which encourages them to continue IWT with the confidence that their efforts are being rewarded. (2) Comparison to others: If participants recognize an individual's training achievement in the up‐to‐date histograms made from all participants' data, then this knowledge stimulates their competitive spirit. (3) Community promotion: If participants see an up‐to‐date ranking of the training achievements in their community, then they may encourage colleagues in their own community to elevate the ranking.