Non-Invasive Anemia Detection Tool with Application of Mini

Spectrometry Base Machine Learning

Theresia Laura da Costa, Elsa Putri Alﬁatun, Risa Picelia Dian Kusuma and Sari Ayu Wulandari

Department of Biomedical Engineering University of Dian Nuswantoro Semarang, Indonesia

ﬁ

Keywords:

Anemia, FCM, Mini Spectrometry, Non-Invasive, PCA.

Abstract:

Anemia is a condition in which the level of hemoglobin (Hb) in the body is reduced. Prolonged anemia

can cause heart problems, pregnancy disorders, and even death. According to the 2018 basic health research

data, anemia sufferers in Indonesia have increased to 48.9%. to reduce the level of anemia, early detection

is needed, but the existing tools are usually invasive, namely using blood samples, which certainly reduces

public interest. This study aims to make an efﬁcient non-invasive anemia detection tool as an option in ane-

mia detection. This tool was developed using the working principle of mini spectrometry, which recognizes

light sources in mini spectrometry using Near-infrared because the Hb wavelength is within the near-infrared

wavelength range. The Hb wavelength is 1700-1725 nm and the near-infrared wavelength is 1000-2500 nm.

The Photo-NIR detector is used as a sensor because it can capture signals according to the near-infrared wave-

length. The method used in signal processing is the Principal Component Analysis (PCA) method for feature

extraction and two feature variations are produced. Furthermore, grouping was carried out using the Fuzzy C

Means (FCM) method so as to produce anemic and non-anemic data based on the degree of membership. The

results of this study obtained an accuracy of 88%. In conclusion, the non-invasive detection tool succeeded in

separating anemic and non-anemic samples. Therefore, a non-invasive detection tool is needed as an option

for the detection of anemia.

1 INTRODUCTION

According to the 2018 basic health research data, ane-

mia sufferers in Indonesia increased to 48.9% from

the previous37.1%, with the age group 15-24 years

and 25-34 years (Sholikhah et al., 2021). Based on

data from the 2019 Semarang City Health Ofﬁce, the

prevalence of anemia in Semarang in the group of

young women has increased to 43.75% and in the

group of pregnant women to 15.4%. meanwhile,

data from Kendal District Health Ofﬁce in 2018, 715

itu 721 pregnant women experienced anemia, so in

2019, the maternal mortality rate reached 103.28 out

of 100,00 live births caused by bleeding.

Anemia is a health disorder caused by a lack of

red blood cells in the blood. Red blood cells are also

known as hemoglobin (Hb) (Agustina et al., 2022).

The normal standard for hemoglobin levels in the

blood is 12 g/dL, if the hemoglobin level in the blood

is below the normal standard, it can be said that

the person has entered symptoms of anemia. Cur-

rently, anemia detection is still using invasive meth-

ods. The use of invasive methods in detecting ane-

mia has several drawbacks, namely it is less efﬁcient

and causes discomfort to its users because it requires

taking blood samples to detect anemia by inserting a

needle into the patient’s arm, then conducting a lab-

oratory examination and ﬁnding out the results takes

a long time (Bernecker et al., 2019; Dervieux et al.,

2020). Therefore, one of the efforts that must be made

in overcoming this problem is to make a detector or

tool to detect anemia that does not need to use nee-

dles (non-invasive), as the newest innovation that can

be an option in anemia detection.

Based on previous research related to non-invasive

detection of anemia, namely through the conjunctiva

of the eye based on digital image processing. In this

study, the feature extraction method was carried out

using the SVM classiﬁcation. The results obtained

have an accuracy of 72,916%. However, this study

still has drawbacks, namely the accuracy of the de-

tection results is affected by the intensity of light. If

the light intensity obtained is less, then the resulting

accuracy results are less precise (Hasan and Ismaeel,

2020). In addition, research related to anemia detec-

tion has also been carried out before detecting Hb. In

Laura da Costa, T., Alﬁatun, E., Kusuma, R. and Wulandari, S.

Non-Invasive Anemia Detection Tool with Application of Mini Spectrometry Base Machine Learning.

DOI: 10.5220/0012441400003848

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Advanced Information Scientiﬁc Development (ICAISD 2023), pages 38-45

ISBN: 978-989-758-678-1

this study, Hb detection was carried out based on the

light intensity received by the sensor. However, in this

study there were still some data that could not be read

(Morscher et al., 2014).

This non-invasive technology was developed us-

ing the working principle of mini spectrometry, which

recognizes light dispersion using pastern recognition

algorithms on the ﬁnger of anemic patients (Mum-

tazmi et al., 2022). The light source in mini spectrom-

etry uses near-infrared because the Hb wavelength is

within the near-infrared wavelength range. The Hb

wavelength is 1700-1725 nm and the near-infrared

wavelength is 870-2500 nm (Nasruddin et al., 2021;

Nidianti et al., 2019). The photo-NIR detector is used

as a sensor because it can capture signals according to

the near-infrared wavelength. The catch of the photo-

NIR detector is in the form of analog signal data that

is converted to digital (ADC). The ADC signal data

is then processed using the principal component anal-

ysis (PCA) feature extraction method, then using the

fuzzy c-means (FCM) classiﬁcation method is used to

ﬁnd points in clusters based on their degree of mem-

bership to separate anemic and non-anemic data. Data

resulting from the detection of anemia or non-anemia

can be viewed quickly in realtime, without requiring a

long time, namely by using the Blynk IoT application

which can be accessed using a smartphone/PC.

2 MATERIAL AND METHODS

2.1 Hardware Design

In designing non-invasive anemia detection hardware,

the tools needed are ESP32 to function as a micro-

controller, and near-infrared as a light source with a

wavelength of 1000-2500 nm Near-infrared is used as

a light source because the wavelength of Hb 1700-

1725 nm is included in the near-infrared wavelength

range (Nasruddin et al., 2021). Furthermore, a photo-

NIR detector is used as a sensor because the photo-

NIR detector is a sensor that can capture the wave-

lengths generated by near-infrared itself, a USB cable

as a connection device with a voltage source, a poly-

carbonate chip that functions in decomposing light

from near-infrared. DC to DC stepdown is used as a

liaison near-infrared, sensors and also servo motors to

ESP32, and servo motors that function to rotate poly-

carbonate plates.

The ﬁrst step in making this detection tool is to

connect the servo motor, near-infrared and photo-NIR

detector to the ESP32 using a DC to DC stepdown ca-

ble. After that, the polycarbonate chip is placed on

the device parallel to the near-infrared above the servo

motor, so that the light produced by near-infrared can

be decomposed using the polycarbonate chip. The

last process is to connect the voltage source to the

ESP32 using a USB cable to operate properly.

The workings of this anemia detection tool use the

working principle of mini spectrometry, namely rec-

ognizing light dispersion using pattern recognition al-

gorithms on the ﬁngers of people with anemia (Mum-

tazmi et al., 2022). Detection is carried out by decom-

posing light from near-infrared through polycarbon-

ate chips to obtain a light color with the same wave-

length as the Hb wavelength, namely 1700-1725 nm

(Nasruddin et al., 2021). Light with the same wave-

length as the Hb wavelength is directed to the patient’s

index ﬁnger which is in the ﬁnger slot. Some of the

light is absorbed by the ﬁnger and some of the light is

passed on. The transmitted light will be captured by

the photo-NIR detector and used in the detection of

anemia. The following is a hardware manufacturing

block diagram shown in Figure 1.

Figure 1: Hardware Manufacturing Block Diagram.

2.2 Machine Learning

The steps taken during the data processing began with

taking the ADC data obtained from the measurement

process directly using a non-invasive anemia detec-

tion tool, in the form of the resulting wavelength data.

Then the data is read by the system and characterized

by calculating the maximum value, standard devia-

tion, and average (mean) of each data to produce a

feature vector value. Furthermore, the feature extrac-

tion process was carried out using the PCA method to

reduce the number of variables (which were initially

very large) to fewer to facilitate analysis at a later

stage. The next step after feature extraction using the

PCA method is the clustering process. The clustering

Non-Invasive Anemia Detection Tool with Application of Mini Spectrometry Base Machine Learning

process was carried out using the FCM method, which

aims to determine the results of grouping anemic and

non-anemic data. After the data has been successfully

grouped, a testing process is carried out to determine

the accuracy of the system so that the results obtained

are accurate.

A block diagram of the data processing process is

shown in Figure 2.

Figure 2: Data Processing Process.

1. Data Sampling

Data sampling was carried out at the Muham-

madiyah Kendal Hospital with as many as 20 ane-

mic samples and 20 non-anemic samples. In one

patient ﬁve times data collection was carried out,

whereas in 1 retrieval 50 data will be taken so that

the total data taken in one patient is 250 data. The

sample data forms a matrix with a size of 250 rows

x 40 columns which will be used as input for the

time characterization domain.

2. Characterization

Characterization is the stage used to ﬁnd the char-

acteristics of each signal. The results of the char-

acterization process are in the form of a feature or

ordinary matrix called a feature vector. Time do-

main characterization is done in a way that calcu-

lates the maximum value, standard deviation, and

average (mean) of each ﬁnite data. The feature

vector returns a value of 200 rows x 3 columns,

where data one-100 = anemia and data 101-200 =

non-anemia.

3. Feature Extraction Using Principal Component

Analysis (PCA) Method

Principal Component Analysis (PCA) is an algo-

rithm that is used to reduce or reduce data in-

formation but does not eliminate the information

contained in the data. The reason for using the

PCA method in the ﬁrst data processing is because

the PCA method is a simple and easy method to

implement but produces great accuracy in the data

reduction process [10]. The function of PCA itself

is to reduce the number of variables (which were

initially very large) to become fewer to facilitate

analysis at a later stage. In the initial stages, the

data measured by the photo-NIR detector is char-

acterized by an n x m matrix, where n indicates the

amount of data and m indicates the characteristics

of the data. Next, the calculation of the average

difference value for each data is carried out, using

equation (1).

C =

∑

i=1

− ¯x) (1)

After calculating the average difference value for

each data, then calculating the variance and co-

variance matrices of the sample data using equa-

tion (2).







1 σ

2 ... σ

1 σ

2 ... σ

... ... ... ...

1 σ

2 ... σ







j =

∑

(i −

i)( j − ¯y)

n − 1

(2)

Next, the search for eigenvectors and eigenvalues

from the previously obtained covariance matrix is

carried out. The process of ﬁnding eigenvectors

and also eigenvalues using MATLAB, where if

the matrix A is square, and the eigenvalues are

x1 ̸= 0, then it is known that AX is a scalar

multiple of X. eigenvalues are scalars which when

multiplied by the column vector X is the same as

matrix A multiplied by the same column vector,

or can be deﬁned by equation (3).

AX = λX (3)

The results of the eigenvectors that have been

obtained are then sorted from the largest to the

smallest value. After sorting, the eigenvector ta-

ble is then multiplied by the initial matrix so that

the PCA results are PC1, PC2, and PC3. In pro-

cessing this data, the PCA results taken were PC1

and PC2 had an eigenvalue of more than 1. The

data processing algorithm using PCA can be seen

in Figure 3.

ICAISD 2023 - International Conference on Advanced Information Scientiﬁc Development

Figure 3: Data processing algorithm using PCA.

4. Clustering Using the Fuzzy C Means (FCM)

Method

The FCM algorithm is used for data clustering

after the data has been reduced using the PCA

method. The FCM algorithm process begins by

reading the data to be lustered in the form of an i

x j matrix, where i is the number of rows of data

and j is the number of columns of data. Next is to

determine the reference constant values to be used

such as the number of clusters (k), then the rank

(w), the maximum iteration (maxiter), the small-

est error (e), and the initial iteration (iter). The

value of this reference constant will determine the

number of iterations and the accuracy of the clus-

tering results. After determining the value of the

reference constant, the initial matrix U is formed

by randomly generating u

k numbers. In gener-

ating uik values, the rule is that the number of

numbers in one row must equal one. After that,

the cluster center value is calculates as described

in the V

jmatrix. V

jis known by using equation

(4).

k j

∑

i=1

((U

)

i j

))

∑

i=1

)

(4)

Then the objective function (P(iter)) is calculated

using equation (5).

iter

∑

i=1

∑

k=1

∑

j=1



i j

−V

k j



)

(5)

After P(iter) is determined, whether or not the

iteration continues is determined by two condi-

tions. The ﬁrst requirement is that the iteration

must be more than or equal to the maximum iter-

ation value, otherwise the iteration will continue.

If so, it will proceed with checking for the sec-

ond condition, namely the difference in the value

of the objective function of the i-th iteration with

the (i-1) iteration must be less than or equal to the

smallest error. If not, then the iteration will con-

tinue. To continue the iteration, it is necessary to

update the u

k value. The u

k value is updated us-

ing equation (6).

∑

j=1



i j

−V

k j



−1

w−1

∑

k=1

∑

j=1



−V

k j



−1

w−1

(6)

If these two conditions are met properly, the itera-

tion is complete and the V

j matrix resulting from

the iteration will be used as the cluster center. So

that the data will be clustered based on its distance

to the canter of the V

jcluster. The data clustering

process using the FCM method is shown in Figure

5. Testing

Testing on data processing begins with setting the

cluster center point each data, namely for the ﬁrst

cluster center point is anemia and the second clus-

ter center point is non-anemic. The testing pro-

cess is then carried out by ﬁnding the distance be-

tween the ﬁrst cluster and the second cluster in

each data using equation (7).

ClusterDistance =

measures

−C1

)



measures

−C1



(7)

After calculating the cluster distance, the next step

is to ﬁnd the value minimum distance between

cluster 1 (C1) and cluster 2 (C2). For anemia data

if the value minimum = cluster distance 1 then

the data can be said to be correct, while for non-

anemia data is said to be correct if the minimum

value = cluster distance 2. After knowing the cor-

rect amount of data, accurate calculations are car-

ried out from the results of data processing that

has been done using the formula equation (8).

Non-Invasive Anemia Detection Tool with Application of Mini Spectrometry Base Machine Learning

Figure 4: Data clustering process using the FCM method.

Accuracy =

Correctamount

Totalnumber

× 100 (8)

2.3 Blynk IoT

Blynk is an IoT platform that is used to remotely con-

trol hardware, display sensor data, store data, and vi-

sualize it using iOS and Android applications (Sep-

tiana et al., 2018). Several types of microcontrollers

are compatible with Blynk IoT such as NodeMCU

ESP8266, Arduino, Rasberry Pi, and ESP32 via the

Internet (Utari et al., 2019). Blynk IoT consists of

several main components, this is shown in Figure 5.

1. Blynk App: used to control hardware and display

data on widgets.

2. Blynk Server: this is a storage service that is re-

sponsible for all the relationships between appli-

Figure 5: Blynk IoT components.

cations and hardware.

3. Blynk Libraries: This includes various widgets

such as control buttons, display formats, notiﬁca-

tions and time management that allow hardware

to send data obtained from sensors to be displayed

on applications effectively.

3 RESULT

3.1 Hardware Design

After the process of designing the system, making the

toll, testing and also repairing the tool that was devel-

oped, the following is a display of the implementation

of the tool that has been made. The display of tool im-

plementation is shown in Figure 6.

Figure 6: Display of tool implementation.

ICAISD 2023 - International Conference on Advanced Information Scientiﬁc Development

3.2 Machine Learning

1. Data sampling

Table 1: Anemia sample data.

No. Number of Patient with Hb Levels

1. 4 2 3 2 2 2 . . . 1

2. 7 6 5 5 5 5 . . . 2

3. 16 10 5 6 6 6 . . . 5

4. 18 10 11 13 13 13 . . . 7

5. 21 10 12 14 14 14 . . . 7

. . . . . . . . . .

. . . . . . . . .

250. 42 46 46 43 43 43 . . . 45

Table 2: Non-Anemia sample data.

No. Number of Patient with Hb Levels

1. 5 6 5 6 5 5 . . . 5

2. 5 13 12 7 9 6 . . . 12

3. 6 14 15 9 13 8 . . . 15

4. 10 16 18 12 17 15 . . . 17

5. 10 17 19 15 21 16 . . . 20

. . . . . . . . . .

. . . . . . . . .

250. 55 52 46 55 56 47 . . . 53

Based on Table 1 and Table 2, analogous data

were found on the results of anemia measure-

ments in anemic patients and also in non-anemic

patients who had been carried out.

2. Characterization

Table 3: Vector feature values.

Data Maximum Standard Deviation Mean

1 19,06 27,24 48,20

2 21,33 19,46 48,80

3 19,20 26,22 48,80

4 18,78 27,36 48,00

5 16,02 20,66 42,60

. . . .

100 21,46 24,98 61,20

101 16,47 22,30 45,60

102 20,76 16,04 52,20

103 14,89 24,14 47,20

104 17,05 23,86 44,80

105 17,46 22,80 43,80

. . . .

200 17,46 22,80 43,80

Table 3 shows the result of the characterization

process, namely the value of the feature vec-

tor. Each data will be searched for the maximum

value, standard deviation and average value. Data

1-100 are data taken from anemic patients, while

data 101-200 are data taken from non-anemic pa-

tients.

3. Feature Extraction Using Principal Component

Analysis (PCA) Method

Feature extraction will be carried out using fea-

ture vector data measuring 200 rows x 3 columns

as shown in Table 3. Where the data will be ex-

tracted using PCA features using Matlab to ﬁnd

the eigen values which have been sorted from the

largest number as in Table 4 and the eigen vector

values which are sorted according to the order of

the eigen values as shown in Table 5.

Table 4: Eigen values.

Eigen Values

112,5291

16,9215

1,1329

Table 5: Eigen vector.

Eigen Vector

0,8833 -0,3162 -0,3462

0,3112 -0,1570 0,9373

0,3507 0,9356 0,0403

The matrix values of the eigen vectors are then

multiplied by the transpose of the initial matrix

which measures 200 rows x 3 columns to pro-

duce the Principal Component (PC) as shown in

Table 6, with 1-100 anemic data and 101-200 non-

anemic data.

Table 6: Principal Component.

Data PC1 Ekstraksi

PCA PC2

PC3

1 2,9189920 3,7818193 0,8478797

2 1,4286449 -4,0443551 2,4594854

3 3,1339610 2,6162275 0,7277167

4 2,6967718 4,0015579 0,6579699

5 -5,2803313 -0,1259826 -0,3266830

. . . .

100 14,3562939 -2,8171040 -1,4957786

101 -1,9160600 0,3886640 -0,8788068

102 3,0524977 -8,2283799 0,6027757

103 -0,3493570 1,8525220 -2,8399143

104 -1,8936031 2,0093340 0,0091286

105 -3,0207303 1,2692095 0,6978164

. . . .

200 -3,0207303 1,2692095 0,6978164

4. Clustering Using the Fuzzy C Means (FCM)

Method

The result of the FCM classiﬁcation is in the form

of a cluster center point, where there are 2 cluster

Non-Invasive Anemia Detection Tool with Application of Mini Spectrometry Base Machine Learning

center points, namely cluster 1 center point (non-

Anemia) and cluster 2 center point (Anemia).

Where is the center point of cluster 1 (c1x,c1y)

and the center point of cluster 2 (c2x,c2y) as

shown in Table 7.

Table 7: FCM classiﬁcation results.

Cluster Center Point

# X Y

1 -0,5593 1,95906

2 1,25848 -6,5747

Based on the cluster center point that has been ob-

tained. It is known that the cluster 1 center point

is the non- anemia data center point and the clus-

ter 2 center point is the anemia data center point.

Therefore, the PC value of non-anemia data must

be close to the cluster 1 center point and the PC

value of anemia data must be close to the cluster

2 center point. If not, the data is declared wrong.

5. Testing

Table 8 shows the correct amount of data for each

anemic and non-anemia data based on the results

of calculating the minimum distance values in

cluster 1 and cluster 2 for each data. Of the 100

anemia data, there are 10 incorrect data and 90

correct data. In non-anemic data, there are 14 in-

correct data and 86 correct data.

So the total correct data from all data, both anemia

data and non-anemia data, is 176 out of 200 data,

and the accuracy obtained from this tool is 88%.

The results in table 8 are the result of the previous

process, where anemia data must be close to the

cluster 2 center point and non-anemia data must

be close to the cluster 1 center point. If not, the

data is said to be wrong. The amount of data pro-

cessed is 200, of which 100 are anemia data and

100 are non-anemic data.

Table 8: Accuracy of data processing anemia detection tool.

Patient Data Accuracy

Amount of

Data

Correct

Data

Incorrect

Data

Anemia 100 data 90 data 10 data 3*88%

Non-Anemia 100 data 86 data 14 data

Total 200 data 176 data 24 data

3.3 Blynk IoT

The program code is written using Arduino IDE

1.8.19 environment, this code starts to pre-

pare the necessary library for the ESP32 mod-

ule <ESP32Servo.h> and Blynk application

<BlynkSimpleEsp32.h>. The anemia detection

result signal is read through the ESP32 pin IO26.

The patient’s index ﬁnger should be placed so that

it touches the tip of the available ﬁnger slot. The

ESP32 microcontroller processes data by converting

analog data to digital information using Analog to

digital conversion. The ESP32 module connects to

the internet hotspot using the same hotspot name

(SSID) and (PASSWORD) and then sends data to

the Blynk application platform. The Blynk IoT

application receives data through a virtual channel

(V5) to be displayed so that it can be seen by users

on their smartphones as shown in Figure 7.

Figure 7: Anemia detection results on the Blynk Applica-

tion.

4 DISCUSSION

The tool designed is a noninvasive anemia detection

tool using the working principle of mini spectrom-

etry as an option in the anemia detection process.

This is because the currently circulating anemia de-

tection devices still use invasive methods. This tool

works by reading the wavelength of the transmitted

light. The less light that is transmitted and captured

by the photo-NIR detector after passing through the

patient’s ﬁnger, it is written that the patient is classi-

ﬁed as anemic, conversely if more light is transmit-

ted and captured by the photo-NIR detector, the pa-

tient is included in a non-anemia patient. Because the

ﬁnger of an anemic patient will absorb more emitted

light so that the light that is transmitted is less. In-

versely proportional to the ﬁnger of a non-anemic pa-

tient (Ningsih et al., 2019).

Many systems have been proposed for an anemia

detection system, but until now the anemia detection

process still uses blood samples. Even though the

blood sampling is small, if it is necessary to do it

repeatedly, it still causes discomfort for the patient

(Septiana et al., 2018). In addition, the anemia de-

tection system is carried out by recognizing images

of patient blood samples (Utari et al., 2019). Because

the tool used still uses blood samples, we made a non-

invasive anemia detection tool without the need for

blood samples in the detection process.

To improve this research, in the future the authors

can develop this tool as a Hb monitoring tool that is

equipped with an alarm so that medical personnel can

provide faster treatment if a patient is detected with

a drastic decrease in Hb. This tool will be a new de-

ICAISD 2023 - International Conference on Advanced Information Scientiﬁc Development

velopment and produce a sophisticated tool in dealing

with cases of anemia.

5 CONCLUSIONS

This study aims to implement an anemia detection

system using the working principle of mini spectrom-

etry with the PCA data processing method and data

clustering using the FCM method. This tool was cre-

ated to be the tool of choice in the process of non-

invasive detection of anemia.

The results obtained from the manufacture of this

anemia detection system are that this tool can distin-

guish anemic patients from non-anemic patients with

an accuracy of 88%. This tool is very useful in the

process of detecting anemia, which was previously

done with invasive methods now in this tool detection

of anemia is done with non- invasive methods to re-

duce the prevalence of anemia. To improve research,

this tool can be further developed as an Hb monitoring

tool.

ACKNOWLEDGEMENTS

The author would like to thank the Muhammadiyah

Kendal Hospital for giving the author permission to

take samples at the hospital. Respondents who are

willing to carry out the detection process with our tool

and parties who play a direct role in designing this

anemia detection tool.

REFERENCES

Agustina, A., Anam, K., and Bulan, S. (2022). Fabri-

cation of spectro-animation as a learning media to

improve student understanding on bohr’s concept of

atomic structure. Inovasi-Jurnal Diklat Keagamaan,

16:136–149.

Bernecker, C., K

ofeler, H., Pabst, G., Tr

otzm

uller, M.,

Kolb, D., Strohmayer, K., Trajanoski, S., Holzapfel,

G., Schlenke, P., and Dorn, I. (2019). Cholesterol de-

ﬁciency causes impaired osmotic stability of cultured

red blood cells. Frontiers in Physiology, 10.

Dervieux, E., Bodinier, Q., Uhring, W., and Th

eron, M.

(2020). Measuring hemoglobin spectra: searching for

carbamino-hemoglobin. Journal of Biomedical Op-

tics, 25.

Hasan, D. and Ismaeel, A. (2020). Designing ecg moni-

toring healthcare system based on internet of things

blynk application. Journal of Applied Science and

Technology Trends, 1:106–111.

Morscher, S., Driessen, W., Claussen, J., and Burton, N.

(2014). Semi-quantitative multispectral optoacoustic

tomography (msot) for volumetric pk imaging of gas-

tric emptying. Photoacoustics, 2:103–110.

Mumtazmi, N., Bulan, R., and Munawar, A. (2022). Anali-

sis karakteristik spektrum tanah berbasis near infrared

spectroscopy (analysis of soil spectrum characteristics

based on near infrared spectroscopy. Jurnal Ilmiah

Mahasiswa Pertanian, 7.

Nasruddin, H., Syamsu, R., and Permatasari, D. (2021).

Angka kejadian anemia pada remaja di indonesia. Ju-

rnal Ilmiah Indonesia, 1:357–364.

Nidianti, E., Nugraha, G., Aulia, I., Syadzila, S., Suciati, S.,

and Utami, N. (2019). Pemeriksaan kadar hemoglobin

dengan metode poct (point of care testing) sebagai de-

teksi dini penyakit anemia bagi masyarakat desa sum-

bersono, mojokerto. Jurnal Surya Masyarakat, 2:29.

Ningsih, E., Fajrin, H., and Fitriyah, A. (2019). Pendeteksi

hemoglobin non invasive. Medika Teknika : Jurnal

Teknik Elektromedik Indonesia, 1:10 18196 010102.

Septiana, E., Peningkatan, P., Pengenalan, A., Pane, E.,

Wibawa, A., and Purnomo, M. (2018). Peningkatan

akurasi pengenalan emosi pada sinyal electroen-

cephalograpy menggunakan multiclass ﬁsher discrim-

inant analysis.

Sholikhah, A., Mustar, Y. S., and Hariyanto, A. (2021).

Anemia di kalangan mahasiswi: Prevalensi dan kai-

tannya dengan prestasi akademik. Medical Technol-

ogy and Public Health Journal, 5(1):8–18.

Utari, E., Listyalina, L., and Puspaningtyas, D. (2019).

Aplikasi self-organizing mapping sebagai alat deteksi

anemia pada citra sel darah merah. Jurnal Gizi Klinik

Indonesia, 16:64.

Non-Invasive Anemia Detection Tool with Application of Mini Spectrometry Base Machine Learning