# Wearable Epileptic Seizure Prediction System with Machine-Learning-Based Anomaly Detection of Heart Rate Variability

^{1}

^{2}

^{3}

^{4}

^{5}

^{6}

^{7}

^{8}

^{9}

^{10}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Epileptic Seizure Prediction System

#### 2.1. System Composition

#### 2.2. HRV Analysis

_{MAD}of the i-th RRI, i.e., RRI

_{i}, is calculated with the stored RRIs (RRI

_{n}| n: 1~i-1) as

#### 2.3. Anomaly Detection Prior to Epileptic Seizure

^{2}statistics [30], which monitor the gap between sample and model data. The Q statistic is the squared distance between a sample and the subspace spanned by the principal components; that is, the Q statistic measures the dissimilarity between the sample and the model data from the viewpoint of the correlation between the variables. The T

^{2}statistic is the Mahalanobis distance between a sample and the origin in the subspace spanned by the principal components. When the T

^{2}statistic is low, the sample is close to the mean of the model data. The statistics are obtained by the following determinants

Algorithm 1 Seizure prediction algorithm | ||

1 | set | τ [0] ← 0, C [0] ← Ɲ. |

2 | while do | |

3 | Collect the newly measured t-th RRI y[t]. | |

4 | Extract and preprocess the HRV indices x[t]. | |

5 | Calculate the t-th T^{2} [t] and Q[t] from x[t] by using Equations (5) and (6). | |

6 | if | $\left\{\left({T}^{2}\left[t\right]\bigvee Q\left[t\right]CL\right)ANDC\left[t-1\right]=\u019d\right\}OR$ |

$\left\{\left({T}^{2}\left[t\right]\bigvee Q\left[t\right]\le CL\right)ANDC\left[t-1\right]=\u01a4\right\}$ | ||

7 | then | $\tau \left[t\right]=\tau \left[t-1\right]+y\left[t\right]$ |

8 | else | $\tau \left[t\right]=0$ |

9 | end if | |

10 | if | $\tau \left[t\right]\ge 10\mathrm{s}$ |

11 | then | $C\left[t\right]=\neg C\left[t-1\right]AND\tau \left[t\right]=0$ |

12 | end if | |

13 | Wait until the next RRI data y [t + 1] are measured. | |

14 | end while |

^{2}or Q statistic exceeds its control limit for more than ten seconds continuously.

^{2}statistics are defined as α% confidence limits. CLs are determined by the lower α% and upper α% of the confidence interval for samples representing the normal condition. The sensitivity and the specificity of the MSPC are controlled by α, which was set to 99%.

_{R}and Σ

_{R}were used from Fujiwara et al. [22].

## 3. Experimental Methods

#### 3.1. MSPC Model Construction

#### 3.2. Measurement Setup and Protocols

## 4. Results

#### 4.1. Patient Attribution

^{2}statistics, and RRI interpolation performed by Equations (1)–(4), then removed from the validation of the seizure prediction performance under visual inspection performed by two or more specialists certified by the Japan Epilepsy Society. The remaining data were reviewed and diagnosed by two or more specialists certified by the Japan Epilepsy Society and the start of the seizure symptoms was labeled as the seizure onset. On the other hand, the time intervals greater than 30 min from seizure onsets, sleep activity, and interictal discharges were labeled as the interictal periods by the specialists observing the video and EEG. The interictal data were used to evaluate the false positives due to the malfunction of the proposed method; they were independent from the pathological physiologic changes in the patients.

#### 4.2. Measurement Accuracy and Reliability

#### 4.3. Seizure Prediction

^{2}statistics exceed the control limit for more than 10 s, where the seizure starts at t = 0.

^{2}statistic. The one-sided sign test [38,39] of the Q and T

^{2}statistics showed that the sensitivity of our algorithm was significantly greater than and lower than the chance level, respectively (both p = 0.0065). Figure 4 and Figure 5 show examples of how the Q and T

^{2}statistics change before a seizure onset and during the interictal period, respectively. The colored band in the figures highlights the duration when $C\left[t\right]=\u01a4$ was satisfied, indicating that the statistic exceeded CL (represented with the colored horizontal line) by 10 s and was discriminated as preictal in Algorithm 1.

## 5. Discussion

^{2}statistics between the patient and healthy groups was not statistically significant (p > 0.05). These results suggest that the autonomic activity rates of the healthy subjects and the epilepsy patients are not significantly different during interictal periods, where there is no effect from the interictal discharges.

^{2}occurred while eating. Patients A and C had temporal lobe epilepsy (TLE), the severity of which may be related to abnormal heart rate regulation [49,50]. Considering preictal episodes A3 and C2, which were not predicted by the Q statistic, the interictal period should be carefully selected for appropriate control limit definition. However, the extracted interictal episodes were shorter than those in previous studies [5], because the severe exclusion criteria were defined to strictly exclude the epileptiform activity from the learning and CL tuning dataset [22]. Further work is required to determine sufficient durations of the interictal episodes for tuning of CLs and to develop a MSPC model specific to TLE patients.

^{2}statistic. According to the results of the video-EEG assessment, four of thirteen false positives occurred during a cognitive function test and seven occurred when the patients were focused on their hobbies (e.g., making plastic models), which suggests that HRV changes occurred due to the mental workload, as seen from the T

^{2}statistic. This hypothesis coincides with our previous results, which showed that the T

^{2}statistic exerted a better discrimination performance than the Q statistic in preventing drowsy driving accidents using HRV analysis and MSPC [51].

## 6. Conclusions

^{2}statistics for the MSPC model were computed in real time and there was a sensitivity of 85.7% with a false positive rate of 0.62 times/h for the Q statistic. The prototype’s real-time seizure prediction and continuous operation of the wearable system will allow applications in the real world.

## Author Contributions

## Funding

## Conflicts of Interest

## References

- Thurman, D.J.; Beghi, E.; Begley, C.E.; Berg, A.T.; Buchhalter, J.R.; Ding, D.; Hesdorffer, D.C.; Hauser, W.A.; Kazis, L.; Kobau, R.; et al. Standards for epidemiologic studies and surveillance of epilepsy. Epilepsia
**2011**, 52, 2–26. [Google Scholar] [CrossRef] - Chang, B.S.; Lowenstein, D.H. Epilepsy. N. Engl. J. Med.
**2003**, 349, 1257–1266. [Google Scholar] [CrossRef] [PubMed] - Schulze-Bonhage, A.; Sales, F.; Wagner, K.; Teotonio, R.; Carius, A.; Schelle, A.; Ihle, M. Views of patients with epilepsy on seizure prediction devices. Epilepsy Behav.
**2010**, 18, 388–396. [Google Scholar] [CrossRef] - Iasemidis, L.D. Epileptic seizure prediction and control. IEEE Trans. Biomed. Eng.
**2003**, 50, 549–558. [Google Scholar] [CrossRef] [PubMed] - Carney, P.R.; Myers, S.; Geyer, J.D. Seizure prediction: Methods. Epilepsy Behav.
**2011**, 22, S94–S101. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Kiral-Kornek, I.; Roy, S.; Nurse, E.; Mashford, B.; Karoly, P.; Carroll, T.; Payne, D.; Saha, S.; Baldassano, S.; O’Brien, T.; et al. Epileptic Seizure Prediction Using Big Data and Deep Learning: Toward a Mobile System. EBioMedicine
**2018**, 27, 103–111. [Google Scholar] [CrossRef] [Green Version] - Karoly, P.J.; Ung, H.; Grayden, D.B.; Kuhlmann, L.; Leyde, K.; Cook, M.J.; Freestone, D.R. The circadian profile of epilepsy improves seizure forecasting. Brain
**2017**, 140, 2169–2182. [Google Scholar] [CrossRef] - Zijlmans, M.; Flanagan, D.; Gotman, J. Heart rate changes and ECG abnormalities during epileptic seizures: Prevalence and definition of an objective clinical sign. Epilepsia
**2002**, 43, 847–854. [Google Scholar] [CrossRef] [Green Version] - Brotherstone, R.; McLellan, A. Parasympathetic alteration during sub-clinical seizures. Seizure
**2012**, 21, 391–398. [Google Scholar] [CrossRef] [Green Version] - Jansen, K.; Varon, C.; Van Huffel, S.; Lagae, L. Peri-ictal ECG changes in childhood epilepsy: Implications for detection systems. Epilepsy Behav.
**2013**, 29, 72–76. [Google Scholar] [CrossRef] - Galimberti, C.A.; Marchioni, E.; Barzizza, F.; Manni, R.; Sartori, I.; Tartara, A. Partial epileptic seizures of different origin variably affect cardiac rhythm. Epilepsia
**1996**, 37, 742–747. [Google Scholar] [CrossRef] [PubMed] - Leutmezer, F.; Schernthaner, C.; Lurger, S.; Pötzelberger, K.; Baumgartner, C. Electrocardiographic changes at the onset of epileptic seizures. Epilepsia
**2003**, 44, 348–354. [Google Scholar] [CrossRef] [PubMed] - Mayer, H.; Benninger, F.; Urak, L.; Plattner, B.; Geldner, J.; Feucht, M. EKG abnormalities in children and adolescents with symptomatic temporal lobe epilepsy. Neurology
**2004**, 63, 324–328. [Google Scholar] [CrossRef] [PubMed] - Kato, K.; Jin, K.; Itabashi, H.; Iwasaki, M.; Kakisaka, Y.; Aoki, M.; Nakasato, N. Earlier tachycardia onset in right than left mesial temporal lobe seizures. Neurology
**2014**, 83, 1332–1336. [Google Scholar] [CrossRef] - Malik, M.; Camm, A.J.; Bigger, J.T.; Breithardt, G.; Cerutti, S.; Cohen, R.J.; Coumel, P.; Fallen, E.L.; Kennedy, H.L.; Kleiger, R.E.; et al. Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Eur. Heart J.
**1996**, 17, 354–381. [Google Scholar] [CrossRef] [Green Version] - Quintana, D.; Alvares, G.; Heathers, J. Guidelines for Reporting Articles on Psychiatry and Heart rate variability (GRAPH): Recommendations to advance research communication. Transl. Psychiatry
**2016**, 6, e803. [Google Scholar] [CrossRef] [Green Version] - Lotufo, P.A.; Valiengo, L.; Benseñor, I.M.; Brunoni, A.R. A systematic review and meta-analysis of heart rate variability in epilepsy and antiepileptic drugs. Epilepsia
**2012**, 53, 272–282. [Google Scholar] [CrossRef] - Jeppesen, J.; Beniczky, S.; Johansen, P.; Sidenius, P.; Fuglsang-Frederiksen, A. Detection of epileptic seizures with a modified heart rate variability algorithm based on Lorenz plot. Seizure
**2015**, 24, 1–7. [Google Scholar] [CrossRef] [Green Version] - Novak, V.; Reeves, L.; Novak, P.; Low, A.; Sharbrough, W. Time-frequency mapping of R-R interval during complex partial seizures of temporal lobe origin. J. Auton. Nerv. Syst.
**1999**, 77, 195–202. [Google Scholar] [CrossRef] - Kerem, D.H.; Geva, A.B. Forecasting epilepsy from the heart rate signal. Med. Biol. Eng. Comput.
**2005**, 43, 230–239. [Google Scholar] [CrossRef] - Billeci, L.; Marino, D.; Insana, L.; Vatti, G.; Varanini, M. Patient-specific seizure prediction based on heart rate variability and recurrence quantification analysis. PLoS ONE
**2018**, 13, e0204339. [Google Scholar] [CrossRef] [PubMed] - Fujiwara, K.; Miyajima, M.; Yamakawa, T.; Abe, E.; Suzuki, Y.; Sawada, Y.; Kano, M.; Maehara, T.; Ohta, K.; Sasai-Sakuma, T.; et al. Epileptic Seizure Prediction Based on Multivariate Statistical Process Control of Heart Rate Variability Features. IEEE Trans. Biomed. Eng.
**2016**, 63, 1321–1332. [Google Scholar] [CrossRef] [PubMed] - Beniczky, S.; Ryvlin, P. Standards for testing and clinical validation of seizure detection devices. Epilepsia
**2018**, 59, 9–13. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Yamakawa, T.; Inoue, T.; Harada, M.; Tsuneda, A. Design of a CMOS heartbeat spike-pulse detection circuit integrable in an RFID tag for heart rate signal sensing. IEICE Trans. Electron.
**2007**, 90, 1336–1343. [Google Scholar] [CrossRef] - Yamakawa, T.; Matsumoto, G.; Aoki, T. A Low-Cost Long-Life RR Interval Telemeter with Automatic Gain Control for Various ECG Amplitudes. J. Adv. Res. Phys.
**2012**, 3, 011205. [Google Scholar] - Yamakawa, T.; Fujiwara, K.; Kano, M.; Miyajima, M.; Suzuki, Y.; Maehara, T.; Ohta, K.; Sasano, T.; Matsuura, M.; Matsushima, E. Development of a wearable HRV telemetry system to be operated by non-experts in daily life. In Proceedings of the 2013 Asia-Pacific Signal and Information Association Annual Summit and Conference, Kaohsiung, Taiwan, 29 October–1 November 2013. [Google Scholar]
- Nomikos, P.; Macgregor, J.F. Monitoring Batch Processes Using Multiway Principal Component Analysis. AIChE J.
**1994**, 40, 1361–1375. [Google Scholar] [CrossRef] - MacGregor, J.F.; Kourti, T. Statistical process control of multivariate processes. Control Eng. Pract.
**1995**, 3, 403–414. [Google Scholar] [CrossRef] - Kano, M.; Nakagawa, Y. Data-based process monitoring, process control, and quality improvement: Recent developments and applications in steel industry. Comput. Chem. Eng.
**2008**, 32, 12–24. [Google Scholar] [CrossRef] [Green Version] - Jackson, J.E.; Mudholkar, G.S. Control Procedures for Residuals Associated With Principal Component Analysis. Technometrics
**1979**, 21, 341–349. [Google Scholar] [CrossRef] - Lockman, J.; Fisher, R.S.; Olson, D.M. Detection of seizure-like movements using a wrist accelerometer. Epilepsy Behav.
**2011**, 20, 638–641. [Google Scholar] [CrossRef] [PubMed] - Kramer, U.; Kipervasser, S.; Shlitner, A.; Kuzniecky, R. A Novel Portable Seizure Detection Alarm System: Preliminary Results. J. Clin. Neurophysiol.
**2011**, 28, 36–38. [Google Scholar] [CrossRef] [PubMed] - Onorati, F.; Regalia, G.; Caborni, C.; Migliorini, M.; Bender, D.; Poh, M.Z.; Frazier, C.; Kovitch Thropp, E.; Mynatt, E.D.; Bidwell, J.; et al. Multicenter clinical assessment of improved wearable multimodal convulsive seizure detectors. Epilepsia
**2017**, 58, 1870–1879. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Beniczky, S.; Conradsen, I.; Wolf, P. Detection of convulsive seizures using surface electromyography. Epilepsia
**2018**, 59, 23–29. [Google Scholar] [CrossRef] [PubMed] - Sforza, E.; Jouny, C.; Ibanez, V. Cardiac activation during arousal in humans: Further evidence for hierarchy in the arousal response. Clin. Neurophysiol.
**2000**, 111, 1611–1619. [Google Scholar] [CrossRef] - Gosselin, N.; Michaud, M.; Carrier, J.; Lavigne, G.; Montplaisir, J. Age difference in heart rate changes associated with micro-arousals in humans. Clin. Neurophysiol.
**2002**, 113, 1517–1521. [Google Scholar] [CrossRef] - Bland, J.M.; Altman, D.G. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet
**1986**, 1, 307–310. [Google Scholar] [CrossRef] - Gibbons, J.D.; Chakraborti, S. Nonparametric Statistical Inference. In International Encyclopedia of Statistical Science; Lovric, M., Ed.; Springer: Berlin/Heidelberg, German, 2011; pp. 977–979. ISBN 978-3-642-04898-2. [Google Scholar]
- Snyder, D.E.; Echauz, J.; Grimes, D.B.; Litt, B. The statistics of a practical seizure warning system. J. Neural Eng.
**2008**, 5, 392–401. [Google Scholar] [CrossRef] [Green Version] - Ramgopal, S.; Thome-Souza, S.; Jackson, M.; Kadish, N.E.; Sánchez Fernández, I.; Klehm, J.; Bosl, W.; Reinsberger, C.; Schachter, S.; Loddenkemper, T. Seizure detection, seizure prediction, and closed-loop warning systems in epilepsy. Epilepsy Behav.
**2014**, 37, 291–307. [Google Scholar] [CrossRef] [Green Version] - Delamont, R.S.; Walker, M.C. Pre-ictal autonomic changes. Epilepsy Res.
**2011**, 97, 267–272. [Google Scholar] [CrossRef] - Salo, M.A.; Huikuri, H.V.; Seppänen, T. Ectopic beats in heart rate variability analysis: Effects of editing on time and frequency domain measures. Ann. Noninvasive Electrocardiol.
**2001**, 6, 5–17. [Google Scholar] [CrossRef] - Chihara, Y.; Yamakawa, T. Evaluation of a developed multichannel R-R interval telemeter and garment-type electrode. In Proceedings of the 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Berlin, Germany, 23–27 July 2019; pp. 4173–4176. [Google Scholar]
- Mormann, F.; Andrzejak, R.G.; Elger, C.E.; Lehnertz, K. Seizure prediction: The long and winding road. Brain
**2007**, 130, 314–333. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Peduzzi, P.; Concato, J.; Feinstein, A.R.; Holford, T.R. Importance of events per independent variable in proportional hazards regression analysis II. Accuracy and precision of regression estimates. J. Clin. Epidemiol.
**1995**, 48, 1503–1510. [Google Scholar] [CrossRef] - Bou Assi, E.; Nguyen, D.K.; Rihana, S.; Sawan, M. Towards accurate prediction of epileptic seizures: A review. Biomed. Signal Process. Control
**2017**, 34, 144–157. [Google Scholar] [CrossRef] - Jeppesen, J.; Fuglsang-Frederiksen, A.; Johansen, P.; Christensen, J.; Wüstenhagen, S.; Tankisi, H.; Qerama, E.; Hess, A.; Beniczky, S. Seizure detection based on heart rate variability using a wearable electrocardiography device. Epilepsia
**2019**, 60, 2105–2113. [Google Scholar] [CrossRef] [PubMed] - Pavei, J.; Heinzen, R.G.; Novakova, B.; Walz, R.; Serra, A.J.; Reuber, M.; Ponnusamy, A.; Marques, J.L.B. Early seizure detection based on cardiac autonomic regulation dynamics. Front. Physiol.
**2017**, 8, 765. [Google Scholar] [CrossRef] [Green Version] - Ansakorpi, H.; Korpelainen, J.T.; Suominen, K.; Tolonen, U.; Myllylä, V.V.; Isojärvi, J.I. Interictal cardiovascular autonomic responses in patients with temporal lobe epilepsy. Epilepsia
**2000**, 41, 42–47. [Google Scholar] [CrossRef] - Ansakorpi, H.; Korpelainen, J.T.; Huikuri, H.V.; Tolonen, U.; Myllylä, V.V.; Isojärvi, J.I.T. Heart rate dynamics in refractory and well controlled temporal lobe epilepsy. J. Neurol. Neurosurg. Psychiatry
**2002**, 72, 26–30. [Google Scholar] [CrossRef] [Green Version] - Abe, E.; Fujiwara, K.; Hiraoka, T.; Yamakawa, T.; Kano, M. Development of Drowsiness Detection Method by Integrating Heart Rate Variability Analysis and Multivariate Statistical Process Control. SICE J. Control. Meas. Syst. Integr.
**2016**, 9, 10–17. [Google Scholar] [CrossRef] [Green Version]

**Figure 1.**System overview and signal processing workflow of the proposed epileptic seizure prediction system.

**Figure 2.**(

**a**) Block diagram and (

**b**) image of the proposed RRI telemeter. (

**c**) Screenshot of the developed Android app showing memo buttons; current heart rate appears on the heart while monitoring. Instrumentation amplifier (Inst. Amp), analog-to-digital converter (ADC), receiver and transmitter (Rx/Tx).

**Figure 3.**Bland–Altman plot comparing the RRIs obtained from the reference ECG and the proposed system. The measurement bias indicated by the dashed-line and the stochastic error range indicated by the solid lines defined by the limits of agreement are sufficiently low for HRV analysis. No significant proportional bias is observed in the result of the regression analysis of this plot.

**Figure 4.**Results of MSPC analysis for seizure episodes (

**a**) A1 and (

**b**) B1. The horizontal lines and vertical lines indicate the control limits and the seizure onset, respectively. The colored bands denote $C\left[t\right]=\u01a4$, indicating the discriminated preictal change.

**Figure 5.**Result of MSPC analysis for the interictal episodes of (

**a**) patient A, selected to include false positives; and (

**b**) patient B, without the false positives.

**Table 1.**Patient demographic and clinical characteristics. Medications were carbamazepine (CBZ), gabapentin (GBP), levetiracetam (LEV), valproic acid (VPA), rufinamide (RFN), lamotrigine (LTG), clonazepam (CZP), zonisamide (ZNS), and lacosamide (LCM).

Patient | Sex | Age | Seizure Foci | Medication (mg/day) |
---|---|---|---|---|

A | F | 31 | Right temporal lobe | CBZ 200, GBP 1200 |

B | M | 54 | Left mesial temporal lobe | LEV 500, VPA 1000 |

C | M | 20 | Left temporal lobe | LEV 1000 |

D | F | 25 | Undefined | RFN 600, LTG 150, LEV 2500, VPA 400 |

E | F | 42 | Occipital lobe (undefined lateralization) | LEV 2000, GBP 600, CZP 1, ZNS 300 |

F | M | 9 | Right frontal lobe | VPA 400, CBZ 200 |

G | F | 14 | Undefined | LEV 1750, LCM 50 |

**Table 2.**Collected episodes. Seizures were focal-impaired awareness seizures (FIAS), focal to bilateral tonic–clonic seizures (FBTCS), and focal aware seizures (FAS). “FIAS→FBTCS” means that FIAS symptomatically changed to FBTCS, which was treated as a consecutive seizure.

Patient | Seizures | Total Duration (h:min) | Interictal Duration (h:min) | Control (age/gender) | Total Duration (h:min) |
---|---|---|---|---|---|

A | 3 FIAS | 70:14 | 0:53 | 31/F | 5:08 |

B | FIAS→FBTCS | 40:35 | 13:47 | 57/M | 4:48 |

C | 2 FIAS | 32:18 | 9:21 | 20/M | 7:17 |

D | FIAS | 105:52 | 2:57 | 25/F | 7:08 |

E | 2 FAS | 85:03 | 2:43 | 45/F | 11:21 |

F | 2 FIAS | 28:39 | 8:07 | 9/M | 6:07 |

G | 3 FAS | 86:45 | 2:28 | 16/F | 7:14 |

Patient | Total RRIs | RRI Outliers | Failure Rate (%) |
---|---|---|---|

A | 245,920 | 11,564 | 4.7 |

B | 163,520 | 577 | 0.4 |

C | 159,240 | 5805 | 3.6 |

D | 474,630 | 31,665 | 6.7 |

E | 343,770 | 14,668 | 4.3 |

F | 144,530 | 2446 | 1.7 |

G | 419,440 | 11,866 | 2.8 |

**Table 4.**Seizure prediction performance. The duration of prediction is shown with the start and the end of exceedance (e.g., −05:28 to −05:17) means that the statistic exceeded the control limit from 5 min 28 sec to 5 min 17 sec prior to a seizure. NA indicates not available, meaning that the statistic did not exceed the control limit. Sen (sensitivity) summarizes the true positive ratio of seizure prediction.

Seizure | Duration (min:s to min:s) | |
---|---|---|

Q | T^{2} | |

A1 | −05:10 to −04:58 | NA |

A2 | −05:16 to −02:19 | NA |

A3 | NA | NA |

B1 | −07:06 to −05:16 | NA |

C1 | −14:40 to −14:25, −12:41 to −11:40 | NA |

C2 | NA | −09:56 to −08:16 |

D1 | −13:05 to −11:15 | NA |

E1 | −16:09 to −14:44, −09:36 to −06:39 | NA |

E2 | −14:43 to −10:36 | NA |

F1 | −12:43 to −12:06 | −13:18 to −10:28, −06:34 to −02:52 |

F2 | −15:41 to −12:42, −10:13 to −08:44 | NA |

G1 | −14:23 to −13:39, −08:37 to −04:54, −03:38 to −01:05 | NA |

G2 | −03:52 to −02:59 | NA |

G3 | −12:18 to −09:33 | NA |

Sen | 85.7% | 14.3% |

Patient | False Positive Rate (times/h) | False Positive Rate of Healthy Control | ||
---|---|---|---|---|

Q | T^{2} | Q | T^{2} | |

A | 3.34 | 1.11 | 0 | 2.14 |

B | 0.29 | 1.16 | 0 | 1.46 |

C | 1.62 | 1.62 | 0.69 | 1.92 |

D | 0.43 | 1.71 | 0.14 | 0.56 |

E | 0.67 | 0.34 | 1.32 | 0.59 |

F | 0.73 | 4.76 | 1.30 | 0.65 |

G | 0.74 | 0.37 | 0.97 | 1.82 |

Total | 0.62 | 1.34 | 0.93 | 1.02 |

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Yamakawa, T.; Miyajima, M.; Fujiwara, K.; Kano, M.; Suzuki, Y.; Watanabe, Y.; Watanabe, S.; Hoshida, T.; Inaji, M.; Maehara, T.
Wearable Epileptic Seizure Prediction System with Machine-Learning-Based Anomaly Detection of Heart Rate Variability. *Sensors* **2020**, *20*, 3987.
https://doi.org/10.3390/s20143987

**AMA Style**

Yamakawa T, Miyajima M, Fujiwara K, Kano M, Suzuki Y, Watanabe Y, Watanabe S, Hoshida T, Inaji M, Maehara T.
Wearable Epileptic Seizure Prediction System with Machine-Learning-Based Anomaly Detection of Heart Rate Variability. *Sensors*. 2020; 20(14):3987.
https://doi.org/10.3390/s20143987

**Chicago/Turabian Style**

Yamakawa, Toshitaka, Miho Miyajima, Koichi Fujiwara, Manabu Kano, Yoko Suzuki, Yutaka Watanabe, Satsuki Watanabe, Tohru Hoshida, Motoki Inaji, and Taketoshi Maehara.
2020. "Wearable Epileptic Seizure Prediction System with Machine-Learning-Based Anomaly Detection of Heart Rate Variability" *Sensors* 20, no. 14: 3987.
https://doi.org/10.3390/s20143987