MIPHAS: Military Performances and Health Analysis System

Gennaro Laudato

, Giovanni Rosa

, Simone Scalabrino

1,2

, Jonathan Simeone

, Francesco Picariello

Ioan Tudosa

, Luca De Vito

, Franco Boldi

, Paolo Torchitti

, Riccardo Ceccarelli

Fabrizio Picariello

, Luca Torricelli

, Aldo Lazich

and Rocco Oliveto

1,2

STAKE Lab, University of Molise, Pesche (IS), Italy

Datasound SRL, Pesche (IS), Italy

LESIM lab, University of Sannio, Italy

XEOS, Roncadelle (BS), Italy

Formula Medicine, Viareggio (LU), Italy

TexTech Technologies, Reggio Emilia (RE), Italy

Ministero della Difesa, Roma (RM), Italy

{fpicariello, ioan.tudosa, devito}@unisannio.it, {f.picariello, l.torricelli}@textechtechnologies.com,

{franco.boldi, paolo.torchtti}@xeos.it, riccardo.ceccarelli@formulamedicine.com, aldo.lazich@marina.difesa.it

Keywords:

Wearable Devices, Machine Learning, Healthcare, Decision Support System.

Abstract:

In the last few years wearable devices are becoming always more important. Their usefulness mainly lies in the

continuous monitoring of vital parameters and signals, such as electrocardiogram. However, such a monitoring

results in an enormous amount of data which cannot be precisely analyzed manually. This recalls the need of

approaches and tools for the automatic analysis of acquired data. In this paper we present MIPHAS, a software

system devised to meet this need in a well-deﬁned context: the monitoring of athletes during sport activities.

MIPHAS is a system composed of several components: a smart t-shirt, an electronic component, a web

application, a mobile APP and an advanced decision support system based on machine learning techniques.

This latter is the core component of MIPHAS dedicated to the automatic detection of potential anomalies

during the monitoring of vital parameters.

1 INTRODUCTION

Wearables that are able to monitor health parame-

ters, such as electrocardiogram (ECG), are among the

technologies that are going to change the world we

live in. That was stated by Microsoft co-founder and

billionaire Bill Gates at the 18th edition of the MIT

Technology Review’s annual roundup

. This feel-

ing is supported also from the scientiﬁc literature that

clearly shows an exponential trend in the number of

publications per year with titles including wearable

electronics (Khan et al., 2019). In the recent years

the research community has devoted a lot of effort

to the ﬁeld of wearable technologies for the health-

care. Park et al. (2002) presented the “Georgia Tech

Wearable Motherboard (GTWM)”, which provides a

versatile framework for the incorporation of sensing,

monitoring and information processing devices. It in-

https://cnb.cx/2XxF9XZ

volves the use of optical ﬁbers and special sensors.

The data bus integrated into the structure transmits

the vital signs information to the monitoring devices

such as an ECG Machine, a temperature recorder, a

voice recorder, etc. Paradiso et al. (2005) designed

WEALTHY, a system where breathing pattern, elec-

trocardiogram, electromyogram, activity pattern or

behavior, temperature can be listed as physiological

variables to be monitored. The system is based on a

wearable interface implemented by integrating fabric

sensors, signal processing techniques and telecommu-

nication systems, on a textile platform. Curone et al.

(2010) devised a prototype system, based on a smart

garment, which enables the detection of health-state

signs of the users (heart rate, breathing rate, body

temperature, blood oxygen saturation, position, ac-

tivity, and posture) and environmental variables (ex-

ternal temperature, presence of toxic gases, and heat

ﬂux passing through the garments),to process data

198

Laudato, G., Rosa, G., Scalabrino, S., Simeone, J., Picariello, F., Tudosa, I., De Vito, L., Boldi, F., Torchitti, P., Ceccarelli, R., Picariello, F., Torricelli, L., Lazich, A. and Oliveto, R.

MIPHAS: Military Performances and Health Analysis System.

DOI: 10.5220/0008989401980207

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 5: HEALTHINF, pages 198-207

ISBN: 978-989-758-398-8; ISSN: 2184-4305

and remotely transmit useful information to the op-

eration manager. Villar et al. (2015) introduced Hex-

oskin

, a line of cutting-edge smart clothing that in-

clude body sensors into garments for health monitor-

ing. Hexoskin monitors (i) ECG, and derived fea-

tures such as: Heart Rate Variability (HRV), QRS

events, and Heart Rate Recovery, (ii) Breathing Rate

and Minute Ventilation (L/min), (iii) activity inten-

sity, peak acceleration, steps, cadence, positions and

(iv) sleep activity. Balestrieri et al. (2019) recently

introduced ATTICUS, an Internet of Medical Things

(IoMT) system for implementing personalized health

services through the monitoring of ECG, respiration

rate measurement, galvanic skin response estimation,

skin temperature measurements, and activity classiﬁ-

cation and monitoring. Matias et al. (2018) presents

a bracelet able to perform health abnormalities detec-

tion based on both vital signs, and accelerometer data

collection from the user.

All such systems are designed to work in the

health monitoring scenario, where physicians can re-

motely assist patients. However, monitoring vital

signs would be greatly beneﬁcial also when practicing

sports. Indeed, many health problems can arise when

a person is under speciﬁc stressful situations, such as

running. Moreover, personal trainers could use the in-

formation acquired through a continuous monitoring

of vital signs to keep track of the improvements of the

athlete.

In this paper, we introduce MIPHAS (Military

Performances and Health Analysis System), a hard-

ware/software system able to detect and analyze the

vital signs of an individual through the use of wear-

able systems when practicing sports. The novelty of

the MiPHAS system lies in the ability to (i) adapt to

the real conditions of the person, modifying the moni-

toring activity; (ii) analyze the data collected through

the use of automatic learning techniques in order to

suggest anomalous and/or critical situations both in

the vital signs or in the performance of the athlete.

The remainder of the paper is organized as fol-

lows. In Section 2 we provide an overview of

MIPHAS, while in Section 3 we describe in details

the core component of MIPHAS, i.e., the decision

support system. Finally, in Section 4 we conclude the

paper and outline directions for future works.

2 MIPHAS OVERVIEW

The MiPHAS project provides the monitoring of the

electrical activity of the heart, respiratory activity,

https://www.hexoskin.com/

temperature and user “dynamics”, aiming at offering

the following services of detection:

• Cardiac Arrhythmia, through an activity analy-

sis of the electrical system of the heart, with par-

ticular focus on bradyrhythmia, tachyarrhythmia

and on Atrial Fibrillation (AF). The focus on these

speciﬁc kinds of disease is due to the fact that car-

diovascular diseases are the leading cause of death

in the world, and, in particular, atrial ﬁbrillation

is one of the most frequent types of cardiac ar-

rhythmia (Fuster et al., 2001; Mathew et al., 2009;

Petrutiu et al., 2006).

• Heat Stress and Heat Stroke, through an ac-

curate estimate of body temperature obtained by

a heat map. Exertional Heat Stroke (EHS) – a

medical emergency deﬁned as life-threatening hy-

perthermia with core body temperature ≥ 40.5

◦

and Central Nervous System (CNS) dysfunction

(Navarro et al., 2017) – mainly affects individu-

als performing rigorous physical activities, such

as athletes, soldiers, or laborers (Gaudio and Gris-

som, 2016).

• Falls, through the analysis of the user’s posture

and movements.

MIPHAS is a project based on an innovative tech-

nology designed to monitor the vital signs of individ-

uals during training sessions. Indeed, MIPHAS has

been devised for athletes and specialized trainers, as

ﬁnal users of the system. The objectives of the sys-

tem are therefore based on an individualistic approach

(e.g., personalized training strategies and techniques),

thanks to the continuous and automatic analysis of

the physical conditions of an athlete. The architecture

turns out to be the composition of a multi-component

technology, which includes:

• A Wearable Sensor Part, in the form of a t-shirt

equipped with innovative sensors able to detect

and acquire vital signs;

• A Hardware Component, capable of transmitting

the signals detected by the wearable sensors. This

component is equipped with a smart hardware,

which is also capable of modifying its behavior

according to the needs identiﬁed by the software

component;

• A Firmware and Software Platform, able to an-

alyze the vital signs detected in order to repre-

sent the health status of a person, the performance

level and any information deemed useful in order

to offer signiﬁcant data for monitoring the physi-

cal activity of the individual;

• An Artiﬁcial Intelligence (AI) Module, able to an-

alyze the data obtained from continuous moni-

MIPHAS: Military Performances and Health Analysis System

199

Figure 1: The MIPHAS wearable prototype with the place-

ment of electrodes and sensors to detect vital signs. The

prototype includes eight temperature sensors (from T1 to

T8), four electrodes for the analysis of respiratory activ-

ity (from R1 to R4) and four electrodes for cardiac activity

(from V1 to V5 and N). The IMU/central unit is inside the

ECU (Electronic Control Unit) located at the height of the

belt.

toring and to identify any anomalies, classifying

them according to their criticality level.

An overview on each component of MIPHAS will

be described in the following subsections.

2.1 The Wearable Component

The wearable component in MIPHAS has been built

to acquire (i) real-time three-leads ECG, (ii) thoracic

bio-impedance, for the breath wave measurements,

(iii) eight-points body temperature, and (iv) inertial

data, basically used for activity recognition and fall

detection. Figure 1 shows the preliminary prototype

of the wearable.

The sensors positioning conﬁguration follows

these rules:

• Cardiac Activity: The wearable allows the detec-

tion of the electrical activity of the heart through a

three-lead ECG. The electrodes are arranged ac-

cording to a typical standard Holter conﬁgura-

tion (Dower et al., 1988), obtaining a continuous

ECG relative to the precordial leads V1, V3, V5.

The fourth electrode, embedded into the wearable,

represents the neutral (N).

• Temperature Acquisition: To obtain a complete

thermal map of the upper body and maintain the

non-invasive constraint, the MiPHAS system has

8 temperature sensors located in 8 distinct areas

of the upper body, according to the model pro-

posed by Wissler (1964) deﬁned for the simula-

tion of the human thermal system. To guaran-

tee the non-invasive constraint, the MIPHAS sys-

tem acquires the skin temperature of the individ-

ual through sensors integrated in the wearable.

• Respiratory Activity: The MIPHAS system ana-

lyzes the patient’s respiratory activity by measur-

ing thoracic bio-impedance, a technique widely

used for long-term monitoring. In fact, impedance

pneumography (IP) signals can be used to assess

respiratory variables such as volume and respi-

ratory rate (Folke et al., 2003; Houtveen et al.,

2003). Wang et al. (2014) carried out a study com-

paring the accuracy of the IP according to four

different conﬁgurations (positioning) of the elec-

trodes. From the comparison of the obtained IP

signals it emerged that the most accurate and sta-

ble conﬁguration appears to be the conﬁguration

shown in Figure 1, since being the electrodes far

from the central area of the chest are less sensitive

to the respiratory movement and are able to guar-

antee a better contact with the leather. Moreover,

such a conﬁguration of the electrodes is also less

sensitive, compared to the others, to the assisted

posture.

• Movement Dynamics: The wearable is equipped

with an Inertial Measurement Unit (IMU), aimed

at detecting user activities and potential falls. The

IMU is positioned close to the waist. Such a

spot is largely considered preferable in applica-

tion of automatic fall detection (Mao et al., 2017;

Ranakoti et al., 2018).

The textile part of the wearable prototype shall

meet key requirements. It has to be: (i) comfort-

able in order to ﬁt prolonged activity intervals with-

out affecting negatively the performances, (ii) stable

when worn in order to favourite the stability of sig-

nals acquisition, (iii) customizable for different sizes

as a common ’fashion’ garment. To met these techni-

cal speciﬁcations we decided to adopt linear knitting

techniques with three innovative yarns:

• Bioceramic Elastic Yarn: Used as a base for the

wearable due to its technical characteristics, it is a

decisive element for the perceived comfort of the

garment. This material provides a positive side ef-

fect on peripheral circulation due to FIR (Vatan-

sever and Hamblin, 2012) that is supposed to be

part of the subjectively perceived comfort. Fur-

thermore, the material also provides the structure

on which the sensors are applied and the other

functional yarns incorporated;

• High Grip Silicone Yarn: The highly deformable

wire is used to obtain a ”second skin” effect, nec-

essary to keep the motionless sensors in their po-

sition and to minimize movement artifacts on the

HEALTHINF 2020 - 13th International Conference on Health Informatics

200

Figure 2: The MIPHAS wearable ﬁnal prototype.

removal of the biomedical signal. This particular

yarn considerably increases friction without elas-

tic compression (elastic force of the thread itself).

The high friction is due to the intrinsic properties

implemented in the material);

• High Conductivity Shielded Stainless Steel Yarn:

For the creation of the electrical circuit, a high-

performance stainless steel yarn made by capil-

laries coated with an insulated polymeric cover-

ing was adopted. This yarn has a high mechanical

strength, good inertia to oxidizing agents and is

suitable for being integrated into the wearable de-

vice as it maintains a softness that is not perceived

as a foreign body. inside the garment;

• Carbon-black based Electrode: a conductive cloth

that acts as dry electrode. Considering the re-

sults obtained in the literature regarding the use of

dry electrodes in electrocardiographic examina-

tions (Chlaihawi et al., 2018; Guo et al., 2016), the

electrodes must be carbon-based and each elec-

trode must cover a surface of at least 2 cm

• IC Thermal Sensors: speciﬁcally designed ac-

tive sensors for superﬁcial skin temperature esti-

mation.

In Figure 2 is presented the ﬁrst prototype of

the wearable, built according to the above mentioned

speciﬁcations.

2.2 The Hardware Component

This section contains the speciﬁcations related to the

BIOX, the data acquisition board used to acquire the

signals coming from the several sensors embedded in

the MIPHAS’s wearable. The board is connected to

the internet via the NB-IoT interface and it can also

connect to a local device through a Bluetooth (Low

Energy) interface. A Generic Attribute Proﬁle Server

(GATT

) has been implemented with several services

https://bit.ly/2hIE3EB

concerning the following measures: eight tempera-

ture measurements, heart rate, respiratory frequency

and thoracic bio-impedance signal, pitch and roll an-

gles and fall detection, SD memory status and battery

level, data streams related to the ECG channels.

The following are the main types of measurements

provided by the various sensors integrated on the data

acquisition board and the minimum/maximum sam-

pling frequencies that are used for the acquisition of:

• Body Temperature: mapping the body tempera-

ture to eight measurement points. The sampling

frequency is approximately 1 Hz, while the ADC

resolution is 12 bits.

Electrocardiogram: acquisition of ECG signals

from three synchronous channels. The sampling

frequency is 320 Hz and the ADC resolution is

equal to 16 bits;

• Respiratory Wave: thoracic bio-impedance mea-

surement at four terminals. Two currents and two

voltages measured with a sampling frequency of

20 Hz. The ADC resolution is equal to 16 bits;

• Dynamics: (1) orientation measurements, ex-

pressed in terms of pitch and roll angles, acquired

with a frequency of 1 Hz. The ADC resolution

is 12 bits; (2) Fall Detection, the event is gen-

erated asynchronously. The accelerometer has a

dedicated event-detection interrupt (fully conﬁg-

urable) which allows the identiﬁcation of a free-

fall episode.

The general architecture of the MIPHAS elec-

tronic component, with all the devices connected to

the microcontroller and the interfaces that have been

used for the data transport and transmission, is listed

below:

• ADS1294

, four-channel ADC with integrated

front-end for the ECG signals conditioning; this

device communicated with the microcontroller

via SPI interface;

• MAX30002

, chip for the thoracic bio-impedance

measurement with four terminals, two for current

driving and two for voltage reading; this device

communicates with the microcontroller via SPI

interface;

• Micro SD, which communicates with the micro-

controller via SPI interface;

• Three LEDs, one red, one green and one blue,

which are used to deﬁne the battery status. The

LEDs are connected to three digital microcon-

troller output pins;

https://bit.ly/33J9sgt

https://bit.ly/2nY15hX

MIPHAS: Military Performances and Health Analysis System

201

• Eight LMT70

temperature sensors, which pro-

vide an output voltage proportional to the mea-

sured temperature. (The signals supplied by these

sensors is acquired using the 12-bit ADC inte-

grated on the microcontroller. Moreover, they are

in order to reduce the energy consumption of the

device);

• Micro-USB, for interfacing the board to a PC via

USB;

• SARA N211

, NB-IoT module; this module com-

municates via UART interface with the microcon-

troller;

• BlueNRG-MS

, Bluetooth module this module

communicates via SPI with the microcontroller;

• LSM6DS3

, sensor containing triaxial accelerom-

eter and gyroscope, which communicates via I2C

with the microcontroller;

• LIS3MDL

, triaxial magnetometer, which com-

municates via I2C with the LSM6DS3 module;

• JTAG interface, for microcontroller programming

and ﬁrmware debugging;

• The PDN (Power Distribution Network), which

supplies the power to the microcontroller and to

all the devices connected to it, based on the volt-

age supplied by the battery. The voltage supplied

by the battery is measured by the microcontroller

via ADC in order to estimate its state of charge.

The schematic is depicted in Figure 3.

In Figure 4 is depicted the preliminary hardware

prototype of the BIOX board.

The schematic of the BIOX is depicted in Figure 3,

while Figure 4 shows the preliminary hardware pro-

totype. Figure 5 shows the 3D rendering of the case

designed to house the BIOX board. In addition to the

case, it is also shown the design of the sled housing

to anchor the BIOX to the wearable of the MiPHAS

system and to allow a reliable connection of the elec-

trodes and sensors to the electronics.

The MIPHAS ﬁrmware implements a GATT pro-

ﬁle server Bluetooth Low Energy (BLE) with services

and features according to the working principle de-

picted in Table 1.

We show in Figure 6 the system which integrates

the MIPHAS’s wearable, hardware and ﬁrmware.

Moreover, it is shown how, with this conﬁguration, it

https://bit.ly/2oSRsl3

https://bit.ly/2iBsZsK

https://bit.ly/2J1Nspi

https://bit.ly/2Mq2Huf

https://bit.ly/2J0rcMm

Figure 3: General architecture of the MIPHAS hardware

component.

Figure 4: The preliminary prototype of MIPHAS hardware

component.

is possible to acquire a real-time multi-channel elec-

trocardiogram.

2.3 The Software Component

Figure 7 shows the architecture of the MIPHAS’ sys-

tem. The server is mainly composed by three compo-

nents:

• Application Server: in the the MIPHAS system, it

is the component that provides the infrastructure,

the support and execution functions;

• Decision Support System (DSS): it is the compo-

nent that implements the decision support system

of the MIPHAS system. Based on machine learn-

ing techniques, it has the task of making a ﬁrst

assessment of the alarms received;

• Alert Manager: it is the component that is in-

volved in the process of notifying the medical spe-

cialists on the alarms received and conﬁrmed by

the DSS. The notiﬁcation of the alarm is made

only to the sport medical specialists who are in

charge of the athlete for whom an anomalous sit-

uation has been found.

HEALTHINF 2020 - 13th International Conference on Health Informatics

202

Table 1: List of services implemented by the MIPHAS’s ﬁrmware.

Feature Transmission Frequency (via BLE) Source

Measurement of body temperature Sent every 1 s Eight points located on the shirt

Measurement of the user’s Heart Rate Sent every 1 s Evaluated on one of the three ECG signals acquired

Electrocardiogram Sent every 31 ms, 10 samples per channel Three ECG channels acquired through the electrodes in V1, V3, V5

Measurement of the user’s Respiratory Rate Sent every 1 s Evaluated on the thoracic bio-impedance signals

Respiratory Wave (thoracic bio-impedance signal) Sent every 500 ms, 10 samples Acquired through the electrodes in R1, R2, R3, R4

Pitch and Roll angles Sent every 1 s Inertial Measurement Unit

Identiﬁcation of falls Sent in asynchronous mode Inertial Measurement Unit

Estimation of the state of the battery [%] Sent every 1 s BIOX

SD memory occupied in percentage [%] Sent every 1 s BIOX

Figure 5: MIPHAS hardware case 3D Rendering.

Figure 6: The preliminary test of MIPHAS and an example

of the acquired ECG signals.

The users can interact with the system in two dif-

ferent ways:

• Active: Regards the interaction between the

MIPHAS system and the all the actors. Espe-

cially, the “Personal trainer” and the “Manager”

access to a web-based platform while the “Ath-

lete” to a application for mobile devices;

• Passive concerns the interaction between the

MiPHAS system and the “Athlete” during a train-

ing session. These do not have access to any fea-

tures of the MIPHAS system; they only supply

data to the system thanks to the combined use of

the wearable component and the BIOX. The wear-

able deals with the detection of vital signs, while

the BIOX deals with its acquisition and sending,

via Bluetooth Low Energy (BLE) protocol to a

personal gateway. It is then the task of the per-

sonal gateway to send data, via the HTTPS proto-

col, to the MiPHAS server.

The sending of personal gateway data to the

MiPHAS server is done online, i.e., the vital signs

of the individual are streamed every second during a

training session.

Speciﬁcally, the structured data, i.e., personal and

clinical data of the athlete are stored in a relational

database. The vital signs automatically monitored, in-

stead, are stored in a document DBMS. The choice

to store this data in a no-SQL database is motivated

by two factors: (i) scalability, because given the large

amount of data to be managed, a document database

offers performances that are certainly superior to a re-

port database as regards information retrieval and (ii)

ﬂexibility, because although the structure of a report

has been deﬁned to send the monitored vital signs, the

latter could be modiﬁed, for example by changing the

sampling frequency of a data (e.g., the ECG) or by

adding new sensors to the wearable. In this scenario,

a document database offers greater ﬂexibility than a

relational database.

2.3.1 The Mobile Application

The MIPHAS app is designed for the proﬁle “Ath-

lete”. In the home there are two buttons (Figure 8 A)

that allow, respectively, to start a new training session

and to view the history of the workouts already per-

formed. In case the athlete wishes to perform a new

training session, she has to be submitted to a relax-

ation session, lasting one minute, in order to record

MIPHAS: Military Performances and Health Analysis System

203

Figure 7: The architecture of MIPHAS.

Figure 8: MIPHAS Mobile App: (A) Home section (B) re-

cap of the assigned training session (C) record of the com-

pleted training session.

and store her baseline heart rate and respiratory rate

values. After the minute, the athlete displays a re-

port of her baseline heart rate and respiratory rate val-

ues and the assigned training program (Figure 8 B).

At this point the athlete can start the actual training

session. When completed the training, the system re-

quires the athlete to perform a new relaxation session,

always lasting 1 minute, so as to be able to monitor

the basal heart rate and respiration rate values again.

After the post-workout relaxation time, the app shows

the athlete a screen with the report of the training ses-

sion just ended (Figure 8 C).

2.3.2 The Web Platform

The web platform has been devised for the other two

proﬁles of the MIPHAS system: (i) the “Gym man-

ager”, responsible for managing personal and clinical

data of athletes, (ii) the “Personal trainer”, who takes

care of the athlete’s training and is interested in mon-

itoring vital signs during a training session in order

to assess how the athlete is responding to the training

protocol.

When accessing the homepage (Figure 9, the sys-

tem shows a mask divided into 3 sections, the ﬁrst two

sections are displayed both to the “Personal trainer”

and to the “Gym manager”, the third section is shown

Figure 9: The Home section in MIPHAS web platform.

Figure 10: The Heart Rate comparison graph.

only to the “Personal trainer”. In the ﬁrst section, the

system shows four information boxes containing: (i)

the number of total athletes registered, (ii) the number

of red alarms in the last 30 days, (iii) the number of

yellow alarms in the last 30 days and (iv) the number

of white alarms in the last 30 days. In the second sec-

tion a histogram is shown on the left with the alarms

recorded in the last 30 days and to the right a pie chart

containing the distribution of the alarms in the last 30

days. In the histogram for each day the system reports

3 different colored rectangles: red, yellow and white,

representing respectively the number of red, yellow

and white alarms occurred in the last 30 days.

The third section “Athletes” is visible only to the

personal trainer and shows the list of all the athletes

assigned to the personal trainer. The system gives also

the possibility to view the progress and the details of

the current training session (if any).

A very useful feature is represented by the Com-

pare Workouts section. Here, the personal trainer can

compare the training sessions obtained by an athlete

on different days or compare the sessions carried out

by two different athletes. Through this feature, the

personal trainer can evaluate the progress and results

of the athletes and choose whether to modify certain

training sessions for particular athletes. The compari-

son is based on the data monitored during the train-

ing sessions. The Data Summary section displays

some descriptors of the trainings, such as: (i) the

HEALTHINF 2020 - 13th International Conference on Health Informatics

204

effective duration, in minutes (ii) the average Heart

Rate and Respiratory Rate during the relaxation be-

fore the training (iii) the average Heart Rate and Res-

piratory Rate obtained during relaxation after com-

pleting the training session (iv) the Cardiac, Respi-

ratory and Thermal activity during the entire training

session. Furthermore, in the Details section, the Per-

sonal Trainer has the possibility to check the graphs

with the Heart Rate (an example in Figure 10, Respi-

ratory Rate and Temperature waveforms based on the

data acquired during the training sessions.

3 ARTIFICIAL INTELLIGENCE

IN MIPHAS

The Decision Support System (DSS) is the core soft-

ware for the automatic detection of potential critical

situations. Intelligent decisions are made at different

levels: part of the DSS is embedded in the personal

gateway, while another part is deployed as a stan-

dalone service. The mobile gateway integrates - what

we have deﬁned - a MIPHAS Local DSS (MLDSS),

considering that the data, automatically analyzed by

this component, are related only to one athlete. On the

other hand, we have deﬁned as MIPHAS Global DSS

(MGDSS) the one installed on the server because it is

enriched with the data from all the MIPHAS system

users.

When the electronic component detects an

anomaly, the DSS is warned. Depending on the

source of the warning (e.g. Heart, Respiration, etc.), a

speciﬁc component of the - Global or Local - DSS

is triggered. The DSS now can conﬁrm or reject

the warning. In case it is conﬁrmed, an alarm (with

a severity information) is forwarded to the personal

trainer. The personal trainer can analyze in near real-

time the anomaly and decide whether to conﬁrm or re-

ject the anomaly. The decision of the personal trainer

is provided to the DSS that can use the (positive or

negative) feedback to enrich its knowledge base.

3.1 Local Decision Support System

The local DSS has the duty to perform the automatic

analysis of:

• Fall Detection. The process of fall detection in

MIPHAS is managed in such a way that the elec-

tronic component, according to an algorithm in-

ternal to the sensor, detects (optimistically) a free

fall event. Once a potential fall has been identi-

ﬁed, the DSS of MIPHAS system has the task of

rejecting a potential False Positive or conﬁrming

the episode, through a more speciﬁc elaboration

than the one implemented by the electronic com-

ponent. Speciﬁcally, for an accurate identiﬁcation

of falls, in MIPHAS it is integrated the method

proposed by Mao et al. (2017). Basically, the au-

thors propose a method based on the evaluation

of a generalized instantaneous Root Mean Square

(RMS) of the triaxial acceleration. To discrim-

inate better a fall event from other rapid move-

ments, in addition to considering the RMS of the

acceleration of the human body, the authors also

analyze the individual posture after the increase

in acceleration. Posture is analyzed through an al-

gorithm that uses the Euler angle, a quantity that

represents the spatial orientation of a body;

• Heat Stress & Heat Stroke. Thanks to a continu-

ous temperature analysis, the DSS needs to iden-

tify a particular situation of heat stress that could

lead to a fatal heat stroke. This is done through

the procedure described by Palma et al. (2017):

an alarm is generated if the patient’s heart rate is

higher than a threshold f (in MIPHAS, the default

is 100 bpm) and there is a change in body tem-

perature at least equal to 0.1° C per minute for a

deﬁned period of time t (in MIPHAS, the default

is 20 minutes).

3.2 Global Decision Support System

We have designed the global DSS in order to perform

these automatic analyses:

• Atrial Fibrillation. When alerted, the MIPHAS

DSS acquires a consistent portion of 3-leads ECG

data. On these signals a well-known QRS detec-

tion method is applied, the Pan-Tompkins algo-

rithm, according to the work of Pan and Tomp-

kins (1985). The outcome of this last step, is used

to trigger the detection algorithm. In MIPHAS,

the algorithm embedded to automatically detect

Atrial Fibrillation is the one proposed by Zhou

et al. (2015). This method is one of the best ap-

proaches in the state of the art, showing highly

precise results in the classiﬁcation metrics (Sensi-

tivity = 97.37%, Speciﬁcity = 98.44%, Accuracy

= 97.89%).

• Respiratory Disorders. When alerted, the

MIPHAS DSS buffers a consistent portion of the

bioimpedance signal. The offset is removed from

the signal, by removing the mean. After this op-

eration, a zero-crossing algorithm has been ap-

plied in order to derive the respiration frequency

from the bioimpedance waveform. Once obtained

a pattern of respiration frequencies, an algorithm

MIPHAS: Military Performances and Health Analysis System

205

Figure 11: The MIPHAS MGDSS modular architecture.

inspects the presence of potential anomalous val-

ues in the pattern;

• Human Activity Recognition (HAR). The HAR

component in MIPHAS is crucial because it al-

lows the system to reject possible False Positive.

If the respiration rate increases, it does not cer-

tainly indicate a respiratory disease if the athlete is

running. Thus, the HAR is an information which

helps the DSS to do a bivariate analysis, before

conﬁrming an anomaly. The algorithm works on

a consistent section of the generalized RMS accel-

eration data: ﬁrst, the method operates to remove

the DC component from the accelerometer data.

Once deleted, the FFT is computed. The funda-

mental frequency, resulting from FFT, provides

the frequency range where the Human Activity

belongs. At the moment, we have experimented a

3-classes identiﬁcation (Standing, Walking, Run-

ning) method.

The MGDSS is designed to be reliable, safe and

easily extensible. The MGDSS system deﬁnes a se-

ries of detectors, that are, components capable of de-

tecting speciﬁc problems based on the measurements

available. MGDSS is structured in different modules

(see Figure 11). Some are basic modules, while others

are plug-ins. There are two basic modules:

• Core Module: contains all the entities of the sys-

tem and deﬁnes the abstract class Detector, which

represents a generic pathology detector;

• Main Module: the entry point of MGDSS. It con-

tains all the classes that allow you to manage the

detectors, starts the system and provides services

outside.

The plug-in modules currently deﬁned are the fol-

lowing (see Figure 11): (i) Atrial Fibrillation mod-

ule which contains the detector of atrial ﬁbrillation

and all the classes necessary for the identiﬁcation of

the pathology, (ii) Breath Problems module as con-

tainer of the detector of respiratory diseases and all

the classes necessary for the identiﬁcation of such dis-

orders and (iii) Arrhythmia module which contains the

generic cardiac arrhythmia detector and the classes

necessary for the identiﬁcation of this pathology; (iv)

Activity Recognition module: contains all the classes

that allow to recognize the activity that the athlete is

doing (e.g., running or resting).

When the intervention of MGDSS is requested for

a given athlete, all concrete detectors related to the

warning are instantiated. For example, if the warning

concerns the fall, the heat stroke detector will not be

started in order to optimally allocate the available re-

sources. An instance of a detector refers to a speciﬁc

athlete: in other words, if several athletes are mon-

itored simultaneously, it is possible that several in-

stances of the same detectors are running at the same

time. MGDSS periodically checks for new data for

the patient in question and provides this data to all the

detectors installed for it.

The detectors access a patient’s data and, based on

these, they decide if there is any anomaly. If at least

one of the detectors reports an anomaly, it is registered

and sent back to the MiPHAS backend, which, in turn,

alerts the physician. Some of the detectors can be just

a support for other detectors: they are not intended to

report anomalies, but to provide additional informa-

tion to help the other detectors to take a decision. At

the moment, the only support detector implemented is

the Human Activity Recognition (HAR) detector. Pe-

riodically, such a detector is called in order to update

patient activity. Like the other detectors, the HAR de-

tector also has an instance for each patient.

Finally, a Detector Manager (DM) orchestrates all

the detectors. This component provides operations

that allow data to be forwarded to all the detectors

of an athlete and to perform a global detection in a

transparent way (without necessarily know which de-

tectors have been instantiated and in which moment).

4 CONCLUSIONS

In this paper we have presented MIPHAS, a tele-

monitoring system aimed at continuously monitor-

ing athletes during their training sessions and offer-

ing support to the trainers in their decision-making

processes. MIPHAS is basically composed of three

components: a smart t-shirt, an electronic device,

called BIOX, and several software components.

The wearable - made of speciﬁc fabrics for the

tele-healthcare applications - embeds several sensors.

The electronic device is the component dedicated to

the data acquisition of vital parameters. This unit is

also in charge of transmitting data in real-time to the

software components. These are mainly composed of

a web application, a mobile app and a DSS.

HEALTHINF 2020 - 13th International Conference on Health Informatics

206

We have an interesting agenda for future works.

First of all, we plan to exhaustively experimenting

MIPHAS. The experimentation will involve profes-

sional athletes. We also plan to enrich the DSS with

more reﬁned algorithms, in order to reach a fully

functioning phase of Continuous Learning. Finally,

we also plan to reduce the invasiveness of the elec-

tronic component by dividing it in distinct physical

modules.

REFERENCES

Balestrieri, E., Boldi, F., Colavita, A. R., De Vito, L.,

Laudato, G., Oliveto, R., Picariello, F., Rivaldi, S.,

Scalabrino, S., Torchitti, P., and Tudosa, I. (2019). The

architecture of an innovative smart t-shirt based on the

internet of medical things paradigm. In Proc. of 2019

IEEE Int. Symp. on Medical Measurements and Appli-

cations.

Chlaihawi, A., Narakathu, B., Emamian, S., Bazuin, B., and

Atashbar, M. (2018). Development of printed and ﬂex-

ible dry ECG electrodes. Sensing and Bio-Sensing

Research, 20, 9–15.

Curone, D., Secco, E. L., Tognetti, A., Loriga, G., Dud-

nik, G., Risatti, M., Whyte, R., Bonﬁglio, A., and Ma-

genes, G. (2010). Smart garments for emergency op-

erators: the ProeTEX project. IEEE Transactions on

Information Technology in Biomedicine.

Dower, G., Yakush, A., Nazzal, S., Jutzy, R., and Ruiz, C.

(1988). Deriving the 12-lead electrocardiogram from

four (EASI) electrodes. Journal of Electrocardiology.

Folke, M., Cernerud, L., Ekstrom, M., and Hok, B. (2003).

Critical review of non-invasive respiratory monitoring

in medical care. Medical & Biological Engineering &

Computing, 41(4), 377–383.

Fuster, V., Ryden, L., Asinger, R., Cannom, D., Crijns,

H., and et al., R. F. (2001). ACC/AHA/ESC Guide-

lines for the Management of Patients With Atrial Fib-

rillation: Executive Summary A Report of the Amer-

ican College of Cardiology/American Heart Associa-

tion Task Force on Practice Guidelines and the Eu-

ropean Society of Cardiology Committee for Practice

Guidelines and Policy Conferences (Committee to De-

velop Guidelines for the Management of Patients With

Atrial Fibrillation) Developed in Collaboration With

the North American Society of Pacing and Electro-

physiology. Journal of the American College of Car-

diology, 38(4):1231–1265.

Gaudio, F. and Grissom, C. (2016). Cooling Methods in

Heat Stroke. The Journal of Emergency Medicine,

50(4), 607–616.

Guo, S., Han, L., Liu, H., Si, Q., Kong, D., and Guo, F.

(2016). The future of remote ECG monitoring systems.

Journal of Geriatric Cardiology, 13(6):528-30.

Houtveen, J., Groot, P., and de Geus, E. (2003). Valida-

tion of the thoracic impedance derived respiratory sig-

nalsusing multilevel analysis. International Journal of

Psychophysiology, 59(2):97–106, 2006.

Khan, S., Ali, S., and Bermak, A. (2019). Recent de-

velopments in printing ﬂexible and wearable sens-

ing electronics for healthcare applications. Sensors,

19(5):1230.

Mao, A., Ma, X., He, Y., and Luo, J. (2017). Highly

Portable, Sensor-Based System for Human Fall Moni-

toring. Sensors, 17(9), 2096.

Mathew, S., Patel, J., and Joseph, S. (2009). Atrial ﬁbrilla-

tion: mechanistic insights and treatment options. Eu-

ropean Journal of Internal Medicine, 20(7):672–681.

Matias, I., Pombo, N., and Garcia, N. (2018). Towards a

fully automated bracelet for health emergency solu-

tion. In Proceedings of the 3rd International Con-

ference on Internet of Things, Big Data and Security,

pages 307–314. INSTICC, SciTePress.

Navarro, C., Casa, D., Belval, L., and Nye, N. (2017). Exer-

tional Heat Stroke. Current Sports Medicine Reports,

16(5), 304–305.

Palma, O. A., Ceballos, M., Rocio, R. V., Basto, C., and

Barzallo, B. (2017). Heat stroke detection system

based in iot. In Proc. of 2017 IEEE Second Ecuador

Technical Chapters Meeting.

Pan, J. and Tompkins, W. J. (March 1985). A real-time qrs

detection algorithm. IEEE Transaction on Biomedical

Engineering, bme-32(3).

Paradiso, R., Loriga, G., Taccini, N., Gemignani, A., and

Ghelarducci, B. (2005). Wealthy-a wearable health-

care system: new frontier on e-textile. Journal of

Telecommunications and Information Technology.

Park, S., Mackenzie, K., and Jayaraman, S. (2002). The

wearable motherboard:a framework for personalized

mobile information processing (pmip). In 39th Annual

Design Automation Conference (ACM).

Petrutiu, S., Jason, N., Nijm, G., Al-Angari, H., Swiryn, S.,

and Sahakian, A. (2006). Atrial ﬁbrillation and wave-

form characterization. IEEE Engineering in Medicine

and Biology, 25(6): 24–30.

Ranakoti, S., Arora, S., Chaudhary, S., Beetan, S., Sandhu,

A., Khandnor, P., and Saini, P. (2018). Human Fall

Detection System over IMU Sensors Using Triaxial

Accelerometer. Advances in Intelligent Systems and

Computing, 495–507.

Vatansever, F. and Hamblin, M. R. (2012). Far infrared radi-

ation (ﬁr): its biological effects and medical applica-

tions. Photonics & lasers in medicine, 1(4):255–266.

Villar, R., Beltrame, T., and Hughson, R. L. (2015). Vali-

dation of the hexoskin wearable vest during lying, sit-

ting, standing, and walking activities. Applied Physi-

ology, Nutrition, and Metabolism, 40(10):1019–1024.

Wang, H., Yen, C., Liang, J., Wang, Q., Liu, G., and

Song, R. (2014). A Robust Electrode Conﬁguration

for Bioimpedance Measurement of Respiration. Jour-

nal of Healthcare Engineering, 5(3), 313–328.

Wissler, E. (1964). A mathematical model of the human

thermal system. The Bulletin of Mathematical Bio-

physics, 26(2), 147–166.

Zhou, X., Ding, H., Wu, W., and Zhang, Y. (2015). A real-

time atrial ﬁbrillation detection algorithm based on the

instantaneous state of heart rate. In Plos One Journal.

MIPHAS: Military Performances and Health Analysis System

207