Prediction of Drug Users Addiction Level with Methadone

Treatment based on Brainwave Maximum Amplitude using ANFIS

Method

Arjon Turnip

, Erwin Sitompul

, George Michael T.

, Shelly Iskandar

, Dessy Novita

, Dwi Esti

Kusumandari

Department of Electrical Engineering, Universitas Padjadjaran, Indonesia

Study Program of Electrical Engineering, Faculty of Engineering, President University, Indonesia.

Electrical Engineering, Institut Teknologi Nasional, Indonesia

Faculty of Medical, Universitas Padjadjaran, Indonesia

Technial Unit for Instrumentation Development, Indonesian Institute of Science, Indonesia

Keywords: Methadone, EEG, drugs, and ANFIS.

Abstract: The use of drugs outside the doctor's instructions tends to damage nerve function in users. The ability to detect

drug users early is a major obstacle in overcoming drug abuse. If the brain system in humans is damaged, it

usually causes permanent disability and is difficult to repair. In this study, a classification method to identify

the level of brain damage in drug users was proposed. In the experiment, drug images were randomly

displayed to stimulate the subjects' memory of using certain drugs. Recorded brain signals from eight subjects

(addiction, methadone treatment (rehabilitation), and control) were performed. Brain waves in the form of

alpha, beta, theta, and delta are used as features for the classification process using the ANFIS method. The

classification results related to drug use with an accuracy rate of 96.97% were achieved.

1 INTRODUCTION

This study aims to determine the level of addiction of

a drug user after using methadone on the brain. The

brain is the central regulatory structure that regulates

most of the movement, behavior, and body functions

such as heart rate, blood pressure, body fluid balance,

and body temperature for every living thing,

especially humans. The large part of the brain can be

divided into four lobes, namely the front (frontal),

back (occipital), middle (parietal), and side

(temporal) brain.

The cerebral cortex is the outermost layer of the

brain, which extends in two hemispheres and is

connected by the corpus callosum. Overall, each

hemisphere is divided into four lobes namely frontal,

parietal, temporal and occipital. This division shows

that each lobe works based on their respective

functions. The frontal lobes are separate from the

parietal, and temporal lobes, where they are

connected by central and lateral sulci, respectively.

The frontal lobe generally functions to regulate

emotional regulation, planning, reasoning, and

problem solving. The parietal lobe functions to

connect all sensory information such as touch,

temperature, pressure, and pain. The temporal lobe

functions to process sensory information from the

parietal lobe such as hearing, recognizing language,

and forming memories. The occipital lobe is the main

center for processing information in the form of

visuals such as interpreting the depth, distance,

location, and identity of the object seen.

The information content in each brain activity

according to its function can be recorded and

processed for various needs such as disease detection,

robot applications in the form of wheelchairs (Turnip

et al, 2015), games, entertainment, and others

(Turnip, Hidayat & Kusumandari, 2017; Simbolon et

al, 2019). Of the many medical instrumentations,

electroensephalogram (EEG) is a tool that can be used

to study information from recorded electrical activity

in the brain, including recording and interpretation

techniques. EEG signals contain information on

electrical activity in the brain, including the state of

electrical and mental disturbances in nerves. EEG

signals have a complex shape, are easily buried by

Turnip, A., Sitompul, E., Michael T., G., Iskandar, S., Novita, D. and Kusumandari, D.

Prediction of Drug Users Addiction Level with Methadone Treatment based on Brainwave Maximum Amplitude using ANFIS Method.

DOI: 10.5220/0010370603010307

In Proceedings of the International Conference on Health Informatics, Medical, Biological Engineering, and Pharmaceutical (HIMBEP 2020), pages 301-307

ISBN: 978-989-758-500-5

301

noise, small amplitudes and do not have a standard

pattern, so visual analysis is not easy.

There are many types of EEG signals, one of

which is the P300 signal type (Karamacoska & Barry,

2019). EEG signals can also be interpreted in the form

of delta (δ), theta (θ), alpha (α), beta 1 (β1), beta 2

(β2), and gamma (γ) waves. Delta waves (δ) are

conditions that arise when a person is sleeping well.

Theta wave (θ) is a condition that occurs when a

person is lightly sleeping, and in a happy state. Alpha

waves (α) are conditions that appear when a person is

relaxed and their eyes are closed. Beta waves (β) are

conditions that appear when a person is doing

activities in terms of remembering such as a state of

thinking. Gamma waves (γ) are conditions in which

brain activity integrates various stimuli (Neto &

Rosa, 2017).

Several previous studies related to the use of EEG

for interpretation of brain signal information have

been done (Simbolon et al, 2015; Turnip et al, 2019).

Research on methadone was conducted using the

fuzzy method as a classifier in the journal entitled

Drug Abuse Identification based on EEG-P300

Amplitude and Latency with Fuzzy Logic Calssifier

compiled by (Turnip, Kusumandari, & Pamungkas,

2018). This study aims to determine the level of a

person's addiction to drugs using fuzzy logic. EEG

Signal Classification Using AAR and SVM with

Eeggyroscope Sensor of Emotiv Epoc was conducted

to determine the results of the EEG signal

classification using the AAR and SVM methods. This

study also links ICA as one of its methods. The

accuracy of this study shows that the number is quite

high, namely 92%.

Furthermore, in the literature entitled An

Adaptive Method for Feature Selection and

Extraction for Classification of Epileptic EEG Signal

in Significant States, feature selection and extraction

to get good classification results are carried out

(Harpale & Bairagi, 2018). This study also uses the

ANFIS method as a classifier with approximate

accuracy 96,48%

EEG signal classification using the K - Means

algorithm and Fuzzy C Means Clustering in a study

entitled EEG Signal Classification using K-Means

and Fuzzy C Means Clustering Methods is proposed

(Hegde, Nagananda, & Harsha, 2015). Drug-related

research entitled The Effects of Methadone

Maintenance Treatment on Heroin Addicts with

Response Inhibition Function Impairments: Evidence

From Event-Related Potentials discusses the effects

of methadone on the brain response of heroin users

(Yang, et al., 2015). The results of this study indicate

that there are differences in brain response after using

methadone. A similar study using methadone, namely

Aging Opioid Users' Increased Risk of Methadone-

Specific Death in the UK explains the level of risk of

death caused by excessive methadone use (Pierce,

Millar, Robertson, & Bird, 2018). Subsequent

research studies (Characteristics of Adherence to

Methadone Maintenance Treatment Over A 15-Year

Period among Homeless Adults Experiencing Mental

Illness) proved that adherence to methadone

maintenance treatment over a 15 year period (among

homeless adults with mental illness) provided

significant improvement.

Risks of Methadone use as Substitute Therapy for

Opioid Addiction during Pregnancy and use of

Clonidine as a Plausible Alternative by (Munin, Iqbal,

& Stowe, 2016) discusses the risks of using

methadone in therapy during pregnancy. Another

study in the form of EEG Signal Classification Using

PSO Trained RBF Neural Network for Epilepsy

Identification discusses the classification of EEG

signals using the RBF Neural Network method with

an accuracy of 98% (Satapathy, Dehuri, & Jagadev,

2017). The application of the ANFIS method to detect

brain tumors was also successfully carried out in a

study entitled "Brain Tumory Detection Using

Artificial Neural Network Fuzzy Inference System

(ANFIS) (J.Deshmukh & Khule, 2014). However, the

accuracy of this study is still relatively low, namely

50-60%.

In this study, the use of the ANFIS method to

predict drug users addiction levels in rehabilitation

patients using methadone was proposed. The

predictors that are developed are expected to be able

to assist the medical team in providing methadone

doses to patients, so that the suspicion of over or

under-dosage can be overcome. Either over or under

dosage will slow down the process of eliminating

dependence on drug use. Even over dose is thought to

cause death or at least will increase dependence on

drugs.

2 METHOD

The process of this research is divided into 4 stages,

namely: Signal Recording, Filtering Process, Data

Extraction, and Data Classification. Recording of the

EEG signal was carried out at Hasan Sadikin Hospital

Bandung. The tools used are Electro-cap, Electro-gel,

electro-gel special syringe, Mitsar EEG-202,

computer, MatLab 2020 software.

Brain recording data of the subjects used were as

many as 8 subjects after fasting to consume

methadone for 8 hours before the experiment. Each

HIMBEP 2020 - International Conference on Health Informatics, Medical, Biological Engineering, and Pharmaceutical

302

subject was subjected to 2 tests, namely different pre

and post stimuli. The different pre and post were the

subjects' EEG recording data before and after using

the methadone, respectively. Stimulus different is a

collection of pictures, one of which contains pictures

of drugs. The image set consists of 4 images of non-

target stimuli (different images with drugs) and 1

target stimulus (drug-like) in 10 different sequences

(20 seconds total). Each subject was recorded with the

condition before and one hour after consuming

methadone. The recorded data is then extracted based

on the brain wave. Figure 1 is EEG Mitsar 201 and

experimental scenario.

Each subject will be assigned an Electro-cap that

has been given Electro-gel for each channel.

Recording is done using the Mitsar EEG-201 device

which is connected to a computer and has WinEEG

software. Prior to recording, the electrode impedance

calibration for each channel was performed with an

impedance less than 5k. Raw data that is recorded

is an EEG signal which is very sensitive to noise.

Therefore, the signal will be preprocessed to sort out

the data that will be considered noise.

The signal that has been extracted from the

filtered signal will study the pattern of changes from

before and after consuming methadone. The pattern

obtained is then used as input for the classifier to

produce a decision on the outcome of each subject.

The classification used is ANFIS logic as a method

for machine learning decision making.

Figure 1: Recording process using Mitsar - EEG 201

Artificial Neuro Fuzzy Inference System (ANFIS)

is an architecture that is functionally the same as the

fuzzy rule base Sugeno model. ANFIS architecture is

also the same as a neural network with radial

functions with certain limitations. It can be said that

ANFIS is a method in which setting rules use a

learning algorithm for a set of data. ANFIS also

allows the rules to adapt. The first order ANFIS

structure is shown in Figure 2. In the Figure there are

5 layers with different functions for each layer. The

box symbol represents an adaptive node, meaning

that its parameter values can change with learning.

Meanwhile, the circle symbol represents a non-

adaptive node whose value is fixed.

Figure 2: Structure of ANFIS prediction model

3 RESULTS AND DISCUSSIONS

3.1 EEG Signal Pre-processing

The signal is filtered using a Band Pass Filter in the

frequency range 0.5 Hz - 70 Hz (signal frequencies

outside this range will be considered noise). Figure 3

is the EEG signal raw data with background noise

content. The portions that are marked with red colour

are known as artefacts.

The filtered EEG signal is then extracted.

Extraction is carried out to obtain brain waves in the

form of alpha, beta, theta, delta, beta, and gamma

waves. Signal extraction was carried out using EEG

Spectra method from WinEEG application. EEG

Spectra produces extraction data in the form of

amplitude, power spectrum, and percent of the EEG

signal. The extraction results are then known as

features to be used as input to the classifier. The

feature chosen in this study is the average percentage

of the maximum amplitude of each wave. Table 1 is

the percentage of the maximum amplitude of each

channel for one subject in one experiment, namely

before consuming methadone. The same was done for

each subject both before and one hour after taking

methadone. The results from each subject were then

averaged to obtain a single amplitude value for each

subject in each experiment.

3.2 Clasification Process

After going through the recording process using an

EEG signal recording device, data will be obtained in

the form of numbers from the translation of the

recorded brain signal. The selected feature is the

difference in the average percentage of the maximum

amplitude, namely in Tables 2 and 3 which consists

of six variables, namely Delta, Theta, Alpha, Beta1,

Beta2, and Gamma (S is the subject). The value of

each wave was obtained from the difference in the

average maximum amplitude before (Table 2) and

Prediction of Drug Users Addiction Level with Methadone Treatment based on Brainwave Maximum Amplitude using ANFIS Method

303

after one hour (Table 3) consuming Methadone for

each subject.

Figure 3: Shape of the brain signal

Table 1: Maksimum amplitudo of one subject for each

channels

Table 2: Average maximum amplitude before Methadone

consumption

δ θ α β1 β2 γ

1,6 5,4 1,8 0,8 1 13,7

1,7 4,3 0,5 0,8 6,7 7,1

1,7 3,9 8,2 0,7 0,9 7,2

2 4,6 5,7 0,7 1 21,8

1,5 6,9 5,3 0,8 1 20,4

1,8 4,2 5,5 0,6 11,2 21,4

2,1 4,1 6,4 0,8 0,9 14,4

1,6 5,3 9 0,7 6,3 1,5

Table 3: Average maximum amplitude after one hour

Methadone consumption

S δ θ α β1 β2 γ

S1 2,2 5,1 6,8 0,8 1 1,3

S2 1,6 5,1 0,5 0,7 1 15,6

S3 1,9 5 6,9 0,7 1,1 1,4

S4 1,7 6,5 2,3 0,7 0,9 1,4

S5 1,9 6,1 3,7 0,7 1 8,5

S6 1,9 4,9 1,9 0,7 11,4 15

S7 1,7 4,1 3,6 0,8 0,9 28,3

S8 1,7 5,2 4,1 0,7 1 8,5

The classification results are used to detect the

level of a person's addiction to drugs in relation to the

administration of methadone doses in rehabilitation

patients. The results from the difference in maximum

amplitude from Table 3 to Table 2 for each subject

are used as input in the ANFIS method as in Table 4.

Table 4. Difference of average maximum amplitude

S Delta

(δ)

Theta

(θ)

Alpha

(α)

Beta1

(β1)

Beta2

(β2)

Gamma

(γ)

Output

S1 1 0.5 9.9 0.1 5.7 8 25.20

S2 0.3 1.6 4.9 0 0 28.9 35.70

S3 0.6 1.9 3.6 0.1 5.9 14.6 26.70

S4 0.1 2.7 1.5 0.1 5.6 0 10

S5 0.8 0 3.3 0 5.7 8.5 18.30

S6 0.5 1.5 1.3 0.2 5.9 14 23.40

S7 0 0.8 2.1 0.1 5.7 34.3 43

S8 0.5 0.7 0 0.1 0.4 27.4 29.10

The column output data in Table 4 is used as a

reference for the classifier. The value ranges are

grouped into 4 categories, where for each category the

value are:

• HA (Heavily Addicted): >40

• MA (Moderate Addicted): <40 - ≥ 30

• SA (Slightly Addicted): <30 - ≥20

• NA (Not Addicted): <20 - ≥ 0

The grouping of values for each feature (Delta,

Theta, Alpha, Beta1, Beta2, and Gamma) is a

predictive output related to the addiction level of

rehabilitation patients. This data is the difference

from the signal before consuming methadone and

after consuming methadone. The six waves will later

become input in this classification process.

Classification process Based on the image above, the

following is an explanation for each description in

Figure 4: (a) how to input training data in the ANFIS

algorithm, (b) showing the ANFIS display, (c)

showing the input and output design on the

classification for each category predefined, (d)

HIMBEP 2020 - International Conference on Health Informatics, Medical, Biological Engineering, and Pharmaceutical

304

iterates 100 times on the data. (e) shows the

comparison of training data and tested data.

Figure 4: Clasification process using ANFIS method

Each input requires the maximum and minimum

value of the data for the smallest and largest input

variables. Based on Figure 5, the following is an

explanation for each description of the input settings:

(a) Delta variable, the value range starts from 0 to 1.

Each category gets a value of 0.25, (b) the Theta

variable, the range of values starts from 0 up to 3.

Each category gets a value of 0.75, (c) the Alpha

variable, the range of values starts from 0 to 10 with

the category of getting a value of 2, 5, (d) the Beta1

variable, the range of values starts from 0 to with 0.02

with each category getting a value of 0.05, (e) the

Beta2 variable, the value range starts from 0 to 6 with

each category getting a value of 1.5, (f) the Gamma

variable, the range of values starts from 0 to 35, with

each category getting a value of 8.75.

Figure 6 shows the Rule Viewer of the input and

output model design in the ANFIS classifier. A hybrid

training algorithm is used where a combination of the

gradient descent algorithm and a least squares

algorithm is used for an effective search for the

optimal parameters. The main benefit of ANFIS is

that it converges much faster, since it reduces the

search space dimensions of the backpropagation

method used in neural networks. 4096 If-then rules

and threemf type of membership function are used.

The developved predictor model designed by ANFIS

method is shown in Figure 7. This study started from

recording brain signals before and after 1 hour of

consuming methadone from 8 subjects who were drug

addicts. There will be a comparison of the results of

the two recordings starting from the increase and

decrease in the subject's brain signal. This

comparison becomes the data that we will process for

classification. To get a comparison of the two data,

the difference between the two experiments was

calculated. After getting the difference between the

two data, the classification process in the ANFIS

prediction model application that has been built is

carried out. The iteration process of 40 epochs was

carried out to increase the accuracy of the prediction

model. Furthermore, the input form is redesigned to

improve the accuracy results before testing the

prediction model.

Figure 5: Design input on ANFIS

The comparison of clasification results using

ANFIS method and calculation process is shown in

Figure 8. It can be seen that both result almost the

same except subjects 4 and 5. The slightly different

of the classification results on those subject was

suspected that both subjects consumed methadone

before the experiment. However, generally the

classification results of 96.97% is shows that this

prediction model can be used to predict the addiction

Prediction of Drug Users Addiction Level with Methadone Treatment based on Brainwave Maximum Amplitude using ANFIS Method

305

level of a drug user. Table 4 shows the predicted

addicted level for all subjects. It were obtained that

only one subject was highly addicted. Six subject

were moderate addicted and one subject was not

addicted.

Figure 6: Viewer IF-Then rules

Figure 7. Developved ANFIS Structure

Figure 8. Comparison of Addiction level of calculated data

with predicted by ANFIS method.

Methadone-induced increased brain signal

activity resulted in several different patterns of

enhancement for each subject. Each pattern was

evaluated in each subject to determine the level of

dependence of a patient on drugs in this study

represented by methadone. The increase in the

amplitude value of several brain waves before and

after consuming Methadone indicates an increase in

brain signal activity in certain parts. Some of the

unusual enhancements to the subject are indicated by

numbers soaring higher than other subjects. This is

thought to be due to interference in recording brain

signals.

Table 4: Classification results

S ANFIS Calculation Addiction level

S1 25,20 27,0 MA

S2 35,70 43,0 HA

S3 26,70 27,0 MA

S4 10,00 27,0 MA

S5 18,30 0,0 NA

S6 23,40 27,0 MA

S7 43,00 27,0 MA

S8 29,10 27,0 MA

4 CONCLUSIONS

The classification of brain signal activity before and

after 1 hour of consuming Methadone using the

ANFIS process resulted in a fairly good prediction.

Using 4 categories to determine the addiction level of

each input resulted in an accuracy rate of 96.97%.

This high degree of accuracy makes ANFIS modeling

feasible for medical needs. However, the drawback is

in the spike in brain signal values that fluctuate based

on disturbance or stimulation when the ANFIS

determines the category associated with the desired

value. ANFIS predictor represent a useful tools for

solving the non linearity problem of drug user

prediction level prediction. Training data for the

present study of ANFIS prediction was randomly

collected from several simulations in MATLAB. The

simulation results proved that ANFIS predictor can be

applied successfully to predict the drug user addiction

level because of its effectiveness and fast processting

time.

ACKNOWLEDGMENTS

This research was supported by Technical

Implementation Unit for Instrumentation

Development, Indonesian Institute of Sciences, and

12345678

96.97%

Output Matlab

HIMBEP 2020 - International Conference on Health Informatics, Medical, Biological Engineering, and Pharmaceutical

306

Department of Electrical Engineering, Universitas

Padjadjaran, and Toba Research Center, Indonesia.

REFERENCES

Athia, S. (2009). Logika Fuzzy. Bandung: Bandung.

Grace Y.WangaRobKydd, T. A. (2015). Changes in resting

EEG following methadone treatment in opiate addicts.

Volume 126, 943-950.

Harpale, V., & Bairagi, V. (2018). An adaptive method for

feature selection and extraction for classification.

Journal of King Saud University – Computer and

Information Sciences.

Hegde, N. N., Nagananda, P. M., & Harsha, M. (2015).

EEG Signal Classification using K-Means and Fuzzy C

Means Clustering Methods. International Journal of

Science Technology & Engineering.

J.Deshmukh, R., & Khule, R. (2014). Brain Tumor

Detection Using Artificial Neural Network Fuzzy

Inference System (ANFIS). International Journal of

Computer Applications Technology and Research.

Karamacoska, D., & Barry, R. (2019). EEG-ERP dynamics

in a visual Continuous Performance Test. International

Journal of Psychophysiology, 249-260.

Karamacoska, D., & Barry, R. (2019). EEG-ERP dynamics

in a visual Continuous Performance Test. Journal of

Psychophysiologi, 249-260.

Kusumadewi, S., & Hartati, S. (2010). NeuroFuzzy

Integrasi Sistem Fuzzy dan jaringan Syaraf Edisi 2.

Yogyakarta: Graha Ilmu.

Munin, M., Iqbal, Z., & Stowe, D. F. (2016). Risks of

Methadone use as Substitute Terapy for Opioid

Addiction during Pregnancy and use of Clonidine as a

Plausible Alternative. Journal of Addiction Medicine

and Therapeutic Science.

Musthafa, A. R., & Tjandrasa, H. (2018). EEG SIGNAL

CLASSIFICATION USING AAR AND SVM WITH

EEGGYROSCOPE. 1-2.

Neto, F. S., & Rosa, J. (2017). Depression biomarkers using

non-invasive EEG: A review. Journal Neuroscience and

Biobehavioral Reviews 105, 83 - 93.

Parpouchi, M., Moniruzzaman, A., Rezansoff, S. N.,

Russolillo, A., & Somers, J. M. (2017). Characteristics

Of Adherence To Methadone Maintenance Treatment

Over A 15-Year Period Among Homeless Adults

Experiencing Mental Illness. Addictive Behaviors

Reports 6.

Philippu, A. (2019). Neurotransmitters are released in brain

areas according to ultradian rhythms: Coincidence with

ultradian oscillations of EEG waves. Journal of

Chemical Neuroanatomy, 66-72.

Pierce, M., Millar, T., Robertson, J. R., & Bird, S. M.

(2018). Ageing Opioid Users’ Increased Risk Of

Methadone-Specific Death In The UK. International

Journal of Drug Policy 55.

Salusi, H. (2020). Aktivitas Gelombang Otak Dengan

Sinyal EEG P300 Pada Empat Daerah Berbeda.

Pekanbaru.

Satapathy, S. K., Dehuri, S., & Jagadev, A. K. (2017). EEG

signal classification using PSO trained RBF neural

network for epilepsy identification. Informatics in

Medicine Unlocked 6.

Shepshelovich, D., & Yahav, D. (2019). Assessment of

frequency and reporting of design changes among

clinical drug trials published in influential medical

journals. European Journal of Internal Medicine.

Shiffman, S., & Basak, S. (2018). An automated mapping

method for Nissl-stained mouse brain histologic

sections. Journal of Neuroscience Methods, 219-227.

Simbolon,A. I., Turnip, A., Hutahaean .J., Siagian, Y. and

Irawati, N., An experiment of lie detection based EEG-

P300 classified by SVM algorithm. International

Conference on Automation, Cognitive Science, Optics,

Micro Electro-Mechanical System, and Information

Technology (ICACOMIT), Bandung, 2015, pp. 68-71,

doi: 10.1109/ICACOMIT.2015.7440177.

Simbolon, A., Nadiya, U., Suhendra, M., et al., Alteration

in Resting EEG for Different Dosage of Methadone

Treatment, International Conference of Computer

Science and Information Technology (ICoSNIKO), 28-

29 Nov. 2019. DOI:

10.1109/ICoSNIKOM48755.2019.9111607.

Turnip, A., Kusumandari, D. E., & Pamungkas, D. S.

(2018). Drug Abuse Identification based EEG-P300

Amplitude and Latency with Fuzzy Logic Calssifier.

Conference: 2018 International Conference on Applied

Engineering (ICAE).

Turnip, A., M. A. Suhendra and Mada Sanjaya W. S. 2015.

Brain-controlled wheelchair based EEG-SSVEP

signals classified by nonlinear adaptive filter. 2015

IEEE International Conference on Rehabilitation

Robotics (ICORR), Singapore, 2015, pp. 905-908, doi:

10.1109/ICORR.2015.7281318.

Turnip A., Hidayat, T., Kusumandari, D. E., 2017.

Development of brain-controlled wheelchair supported

by raspicam image processing based Raspberry pi. 2nd

International Conference on Automation, Cognitive

Science, Optics, Micro Electro--Mechanical System,

and Information Technology (ICACOMIT), Jakarta,

2017, pp. 7-11, doi:

10.1109/ICACOMIT.2017.8253377.

Turnip, A., Kusumandari, D., Novita, D., et al., An Increase

of the Central Frequency of Alpha Brain Wave Effected

by Soy Peptide in Open Eyes Condition based EEG

Signals, IEEE International Conference of Computer

Science and Information Technology (ICoSNIKOM),

28-29 Nov. 2019.

Yang, L., Xu, Q., Li, S., Zhao, X., Ma, L., Zheng, Y. . . . Li,

Y. (2015). The effects of methadone maintenance

treatment on heroin addicts with response inhibition

function impairments: Evidence from event-related

potentials. Journal of food and drug analysis 23.

Prediction of Drug Users Addiction Level with Methadone Treatment based on Brainwave Maximum Amplitude using ANFIS Method

307