Under the background in big data mobile health, how to solve the data transmission among hospitals, doctors, patients, their relationships is an important problem. Patients and their relationships use mobile devices receiving and sending messages. Those devices limit storage space when they send and receive messages. Electronic medical record of each patient contains above 3G and 4G storage space. If all messages are received or sent by device, that would consume a number of energy and overhead. Meanwhile, big data in transmutation may reduce deliver ratio because of data losing. Next work will be discussed this problem.
Precondition
In the real world, patient personal data are confidentiality. At the same time, diagnosis suggests from doctors are protected. Thus, effective data in mobile health are only predicted disease and therapeutic schedule, especially in underdeveloped areas, it is good for patients and doctors to judge disease.
In this paper, effective data in mobile health contain diagnose pictures, diagnosis collection in clinic, data in history diagnosis, diagnostic reports, electronic patient record and so on.
Model Design
In big data mobile health environment, model design is show as Fig. 1.
In Fig. 1, P shows patients; R shows patient’s relationships. When patients and hospital establish relationship, hospital or doctors send messages to the patients by mobile devices, as long as the patients in the hospital’s communication area. Patients receive the messages from the hospital, and then they can put forward data to their own relationships. In this way, patients do not go to a hospital looking over their inspection results, the stress will be reduced in hospital from undertaking transfer of personnel. Patient relationship may get results with the patient’s devices according to the data transmission model.
This communication model likes opportunistic network. All the messages are transmitted by non-link. The scheme of model design can choose opportunistic network as communication method according to the characteristic of opportunistic network. In opportunistic network, all the roles would become nodes. They can transmit messages with no link, and messages can be stored by devices. This process is in Fig. 2.
Figure 2 shows data transmitting in big data mobile health environment. Effective data packets are recorded by sensor devices and they can transmit to mobile devices. According to data collection, all the data can be sent to hospital and doctors. Data will be analyzed. Hospital and doctors receive patients’ request by mobile devices. This request contains patients’ check indicator, illness describing, and patient requirement and so on. Doctors diagnose the condition of patients’ index from mobile device, and then give treatment list for patients. Patients can download list and look over diagnostic results. At the same time, patients may send these results to their relationships. All the processes can be established by 3G or 4G network environment, users may only have a mobile device completed diagnoses. Patients and doctors are nodes in big data opportunistic network, they can transmit information by moving. In this communication method, patients and their relationships need not go to hospitals; they may save money and time. Hospitals can also reduce the stress of populations and guarantee work efficiency. In large population counties, this model may reduce finance with hospital beds, because patients can get medical service at home when they connect their mobile devices.
Effective Data Analysis and Decision
Effective Data in Clinic
In big data communication, storage space of mobile devices are limit, we can not receive all the data when a great number of data packets are created in clinic. How to select effective data become important work.
Figure 3 is diagnosing pictures in PET-CT. It is formed of 12 pictures. There are eight pictures contain yellow-bright areas (SUV Concentration), four pictures display scanning location. For doctors, they judge pathological changes only four pictures in Fig. 3, and then they can point effective images (picture 4, 8, 9 and 12 are marked circle. Picture 8 is maximum concentration area). Accurate marked can help doctors decision-making.
But in the real condition, each patient may have over more 600 pictures. Only 40 pictures may adopt by doctors. Effective data is 6.67 %. Moreover, 600 pictures approximate 1.5G storage space in device. In wireless communication, if doctors or hospitals send all pictures to patients, many network resource must be wasted, especially in big data environment, mass of people join in transmission, there are no enough space storing effective information.
Figure 4 is collection data in clinic. For a patient in hospital, there are 80 M per day stored in database. However, only 1 % data are abnormal. That is to say, 1 % in transmission can help patient and doctor judge disease, 99 % data are useless. In order to keep 1 % effective abnormal data, we must waste a great number of space.
Not only that, patient in clinic may display normal in a day. It seems in Fig. 4 in the morning. But at night, a short moment in clinic, heart rate, blood pressure and blood supply may change very soon. It is dangerous for patient. Those data are important to save life and analyze illness.
From what has been discuss in those figures, it is important to select effective data when we consider to save storage space and transit effective data packets, especially in big data communication environment. So in big data communication, how to make data decision and select effective data is the key.
Effective Data Decision and Transmission in Communication
In big data communication, a great number of electronic medical record (EMR) are commuted in wireless devices. Table 1 shows an EMR for a patient in hospital.
Table 1 Electronic medical record for a patient in hospital
From Table 1, we can see that many type of documents are records in hospital. For a patient, an EMR may occupy 2G storage space and documents conclude different types in database. It is hard to diagnose directly for one or two doctors at moment. Moreover, if all documents which have not been judged by doctors are transmitted by devices in wireless, there are not enough network resource could be used. More seriously, data packets which contain important records for patients would loss. It is insecurity when EMR are transmitted between people. Thus, we must consider how to select effective data in EMR.
Figure 5 is data decision tree in EMR. The root of the decision tree is a patient. The first layer is the basic information assemblies of the patient. The second is sub-assembly, which records detail information. The third layer is diagnosis category. It contains many kinds of parameters. This process in hospital is completed by medical treatment machine.
Traditional method in wireless networks may deliver all diagnosis categories to patients and doctors. In order to reduce energy and overhead in transmission, new method must decrease data packet transmission. We can set normal range of parameters in decision layer. Abnormal data and normal parameter are selected by devices. All the decision data assemble from K
1 to K
n
which can help doctor and patient to judge illness.
Figure 6 shows data assemble exchanging between patients and doctors. Patients send request as data packets. These data packets include diagnostic reports, electronic patient record and image information. When doctors receive these data packets, they can judge patients illness condition and give a treatment recommendation. At the same time, doctors may select a part of effective data which are recorded patient results from medical instruments. From K
i
to K
n
are data packets in EMR data decision. When doctors receive request from patients, they may suggest some important patient results data packets such as K
i
to K
j
and K
m
to K
n
. In this way, patients need not download the whole data packets and then they can know therapeutic regimen clearly.
The process of exchanging data packets are finished by APP device(Accelerated Parallel Processing device). Effective data packets reduce energy consumption and overhead. Patients select effective data packets by mobile device from anywhere and anytime in big data environment. It is not only convenient receive and send messages, but also improve the application for mobile device and mobile health.
This communication likes opportunistic network. All the messages are transmitted by non-link. Next work is how to design effective scheme of big data mobile health in opportunistic network.
Effective Communication Scheme In Big Data Opportunistic Network
In opportunistic networks, the characteristic of the Spray and wait algorithm [10, 15] is that the nodes transmit information to neighbor nodes by spraying data packets. There are serious problems in this algorithm. Firstly, since the storage space of the neighbor nodes is limited, it may not be able to receive all the data packets, which may cause data lost. Secondly, the lost of the data leads to the low deliver ratio. Thirdly, with multiple neighbor nodes, the spraying will consume great routing overhead of the source node. The Spray and wait algorithm is not an optimal selection. The Binary spray and wait algorithm [16] is an optimization of the Spray and wait algorithm. It reduces routing overhead in the way that the source node sends 1/2 data data packets to the neighbor nodes and stores the rest data. However, this algorithm can be adopted only in networks with few nodes. With the increase of nodes and data data packets, the problems in Spray and wait algorithm occur again.
This paper designs ADTRA. The algorithm improves receiving and sending data packets in big data in opportunistic network when data packets are transmitted in mobile health environment. The data transmitting process refer to Fig. 6 in ADTRA. That is to say, the whole data data packets from doctors and patients found an assemble as V. V contains data assemble and is shown as {K
i
, K
i+1……K
n
}. Then in ADTRA, the effective data packets assemble from doctors and patients shows V′. V′ contains {K
i
……K
j
} and {K
m
……K
n
}. There is a relationship between V and V′ is \(V' \subseteq V\). For each element in V and V′, it contains several data packets. If assemble V contains P data packets, V′ contains p data packets. According to \(V' \subseteq V\), it shows the relationship between P and p: \(p \le P\). p is effective data packets from doctors to patients.
The next work will compare with effective data packets and the whole data packets when they are shown in energy consumption, overhead and deliver ratio in big data opportunistic network.
Energy Consumption in Big Data Mobile Health
Mobile devices must cost energy consumption in big data mobile health when these send and receive messages. If it is cost energy more than the node storage, the node would die or the device would not work. Messages can be send and received on the condition that E ≤ E
0, E is send data packets consumption, E
0 is storage energy by node.
Energy consumption in opportunistic network comes from data transmission, signal processing, and hardware operation. Particularly in data transmission, we consider energy consumption in scan, data transmission, and received data. These factors establish some models.
-
1.
Energy consumption in scan shows energy consumption in scan channel with node. Each scan consumes e
s
. Scan period is T. Energy consumption E
s
for node is
$$E_{s} = e_{s} \times \frac{t}{T}$$
(1)
where t shows working time with node.
-
2.
Energy consumption is in transmission. The sent data packet needs to consume energy. Each sent data packet must consume e
t
. p
t
data packets are transmitted. Energy consumption in transmission E
t
is:
$$E_{t} = e_{t} \times p_{t}$$
(2)
-
3.
Energy consumption is in receiving. The condition is the same as receiving data packets by nodes. Each data packet received must consume e
r
. p
r
data packets are transmitted. Energy consumption in transmission E
r
is:
$$E_{r} = e_{r} \times p_{r}$$
(3)
From Eqs. (1)–(3), energy consumption E
c
and surplus energy E
sur
are
$$\left\{ \begin{aligned} &E_{c} = E_{t} + E_{r} + E_{s} \hfill \\ &E_{sur} = E_{0} - E_{c} \hfill \\ \end{aligned} \right. .$$
(4)
Energy can be calculated easily according to the established energy assessment model.
From energy assessment model, we can found a functional relation between node and neighbor when they send and receive message.
In opportunistic network, each node sends data packets to neighbors consuming energy E
s
+ E
t
. If node has n number neighbors and sends p
r
data packets to a neighbor. The energy consumption for a node E is:
$$E = E_{s} + E_{t} = \left( {e_{s} \times \frac{t}{T} + e_{r} \times \sum\limits_{i = 1}^{r} {p_{r} } } \right) \times n$$
(5)
In ADTRA, p
r
is decided by neighbor requests. But in Spray and wait algorithm, each node sends uniform number \(\bar{p}\). It is obvious that \(p_{r} \le \bar{p}\), because in Spray and wait, all data packets are transmitted from node to neighbors when it carries. However in Binary spray and wait, each node sends uniform number \(\frac{{\bar{p}}}{2}\). If effective data packets for ADTRA are half with the total, that is to say \(p_{r} \le \frac{{\bar{p}}}{2}\) ,energy consumption of ADTRA is less than Binary spray and wait; else \(p_{r} \succ \frac{{\bar{p}}}{2}\) when effective data packets for ADTRA are less than half of total.
Overhead in Big Data Mobile Health
There is a pair of close relationship between overhead and data packet transmission. If node sends data packets to neighbors, it may cost overhead. In normal, overhead in opportunistic network shows load level with node. High overhead may bring low deliver ratio and high transmission delay. The more data packets are transmitted, the more overhead may cost with node. Especially in big data mobile health environment, reduce overhead may improve network quality.
Assume that among doctors, patients and hospitals are nodes. Overhead in one transmitting o is:
$$o = \overline{{o_{s} }} \times p_{t}$$
(6)
\(\overline{{o_{s} }}\) is shown average overhead with each node, then \(\overline{{o_{s} }} = \frac{{\sum\nolimits_{i = 1}^{n} {o_{i} } }}{n}\). o
i
is overhead with nodes in opportunistic network.
If node has n number neighbors and sends p
r
data packets to a neighbor. Overhead O in opportunistic network is:
$$O = \overline{{o_{s} }} \times n \times \sum\limits_{r = 1}^{n} {p_{r} }$$
(7)
It is the same condition in energy consumption, in Spray and wait algorithm, each node sends uniform number \(\bar{p}\) and in Binary spray and wait, each node sends uniform number \(\frac{{\bar{p}}}{2}\). Overhead in ADTRA is less than Spray and wait, because \(p_{r} \le \bar{p}\). If effective data packets for ADTRA are half with the total, that is to say \(p_{r} \le \frac{{\bar{p}}}{2}\) ,overhead in ADTRA is less than Binary spray and wait.
Deliver Ratio in Big Data Mobile Health
Data packet deliver ratio is most important parameter on network. In big data mobile health environment, high deliver ratio can be guaranteed messages to send and receive successful among patients and doctors. And then integrality for patient electronic medical record, doctor diagnostic result can be protected.
The deliver ratio has some relationships among energy consumption and overhead when storage and energy limit with nodes. Nodes in opportunistic network can not give service when overhead is flood and energy is empty. It likes mobile device can not as normal. So deliver ratio is founded by triad function:
$$D^{{\left\{ {peer \, to \, peer} \right\}}} = f\left( {E,O,p^{{\left\{ {peer \, to \, peer} \right\}}} } \right)$$
(8)
\(D^{{\left\{ {peer \, to \, peer} \right\}}}\) shows deliver ratio between nodes,E shows energy consumption,O shows overhead. \(p^{{\left\{ {peer \, to \, peer} \right\}}}\) shows transmitted data packets between nodes. And then \(p^{{\left\{ {peer \, to \, peer} \right\}}}\) is:
$$p^{{\left\{ {peer \, to \, peer} \right\}}} = \frac{{p_{r} }}{{p_{s} }}$$
(9)
p
r
is received data packet, p
s
is sent data packet.
The average of deliver ratio among nodes is:
$$\overline{{D^{{\left\{ {peer \, to \, peer} \right\}}} }} = \frac{{\sum\nolimits_{k = 1}^{n} {D_{k}^{{\left\{ {peer \, to \, peer} \right\}}} } }}{n}$$
(10)
\(D_{k}^{{\left\{ {peer \, to \, peer} \right\}}}\) shows deliver ratio between two nodes, n shows nodes quantity.
The average deliver ratio between nodes according to Eqs. (9), (10) is:
$$\begin{aligned} \overline{{D^{{\left\{ {peer \, to \, peer} \right\}}} }} & = \frac{{\sum\nolimits_{k = 1}^{n} {D_{k}^{{\left\{ {peer \, to \, peer} \right\}}} } }}{n} \\ & = \frac{{\sum\nolimits_{k = 1}^{n} {f\left( {E_{k} ,O_{k} ,p_{k}^{{\left\{ {peer \, to \, peer} \right\}}} } \right)} }}{n} \\ \end{aligned}$$
(11)
Energy consumption E
k
and overhead O
k
are functions about p
k
according to Eqs. (5), (7).
So Eqs. (11) can be shown as:
$$\overline{{D^{{\left\{ {peer \, to \, peer} \right\}}} }} = \frac{{\sum\nolimits_{k = 1}^{n} {f(E_{k} (p),O_{k} (p),p_{k} )} }}{n}$$
(12)
Deliver ratio is a function about data packet p according to Eqs. (5), (7), (12).
If messages are delivered by nodes have enough energy and overhead. The deliver ratio is decided by rate of neighbor receiving data packets and node sending data data packets. That is \(D = \frac{{D_{r} }}{{D_{s} }}\). To the whole network, average of deliver ratio \(\overline{D}\) is:
$$\overline{D} = \frac{{\sum\nolimits_{i = 1}^{n} D }}{n}$$
(13)
It tests and verifies average of deliver ratio by simulation.
Delay in Big Data Mobile Health
In network communication, transmission delay is a problem. Especially in mobile health environment, if we may reduce delay in EMR transmission between devices, patients can gain more rescue time.
If a packet is transmitted from node to its neighbor spend time \(t_{0}\). Actual transmission time from two nodes is \(t_{n}\). Delay \(T_{del}\) in transmission is
$$T_{del} = t_{n} - t_{0}$$
(14)
According to Eq. (14), we can calculate delay time between nodes when they transmit and receive data packets.