Determination of Thorax Exposure Factors in Conventional X-rays

Imaging using the Artificial Neural Network Method

Donni Maulana Sipa and Jamrud Aminuddin

Department of Physics, Faculty of Mathematics and Natural Science, Universitas Jenderal Soedirman, Purwokerto, Indonesia

Keywords: X-ray, Exposure Factor, Neural Network.

Abstract: The application of artificial intelligence in the medical field is indispensable for providing optimum results.

Conventional X-ray imaging is the fastest, most common and least expensive diagnostic imaging system

available. However, an effective X-ray examination depends on the range of radiation given to the subject.

The radiation from an X-ray primarily depends upon X-ray tube current (mA), tube voltage (kVp) and

exposure time(s); these parameters define the dosage. X-ray radiation has a negative impact on the human

body; this danger is not visible, but X-ray radiation can damage human cell tissue. This work aims to explore

and analyze X-ray exposure parameter levels to the thorax with an artificial neural network, which helps to

diagnose exposure of the tissue that is being irradiated. By entering distance, weight and height into the

software, radiographers will get the optimum exposure factor settings for the patients’ thorax. The subjectivity

of exposure factor settings from radiographers can be objective, and optimum exposure settings for patients

can result in lower radiation with a good, detailed image, thereby reducing the impact of X-ray radiation.

1 BACKGROUND

A radiograph from X-ray imaging is produced based

on the suitability of the exposure factor chosen by a

radiographer. The exposure factor is chosen based on

the region of examination, the patient's body weight,

projection position, distance from X-ray to the patient

and the patient's physical condition (Carlton et al.,

2019). If the value of the exposure factor is too high

or too low, the radiograph will yield a shadow, which

has no diagnostic information, and the inappropriate

exposure factor also triggers a high radiation value. In

other words, the image on X-ray film will be over-

bright or over-dark, making it difficult to read by a

doctor and giving the effects of excessive radiation to

the patient (Gois et al., 2019).

The value of the exposure factor can be seen by

the state of the patient; underweight patients and

overweight patients have different exposure factor

settings. Because of the differing patient surface

areas, radiographers need to ensure that the value of

the exposure factor is a match to the patient. Every X-

ray device and every radiographer needs to have clear

parameters for overweight or underweight patients;

this makes the settings of exposure factor subjective

by the radiographer itself (Elster, 2010).

Exposure factor calculation using fuzzy logic has

been developed by Santoso et al., (2016). Fuzzy logic

succeeded in calculating the optimum exposure

factor. Instead of using fuzzy logic, however, we

employed the use of an artificial neural network.

Therefore, the aim of this work is to explore and

analyse X-ray exposure parameter levels to the thorax

with an artificial neural network, which helps to

diagnose exposure of the tissue being irradiated.

2 METHOD

2.1 Exposure Factor

A good image depends on the exposure factor. Image

quality also represents the amount of radiation

received by the patient during the imaging technique.

Three main factors that determine the image quality

are the kilovolt potential (kVp), which controls the

penetrating power of the X-ray; the milliampere

(mA), which controls the number of X-rays; and the

exposure time (S), which controls the duration of

exposure (Hiswara, 2002). These combinations will

decide the contrast sensitivity, detail and noise of the

radiograph. Variables from imaging techniques must

Sipa, D. and Aminuddin, J.

Determination of Thorax Exposure Factors in Conventional X-rays Imaging using the Artiﬁcial Neural Network Method.

DOI: 10.5220/0009388100330037

In Proceedings of the 4th Annual International Conference and Exhibition on Indonesian Medical Education and Research Institute (The 4th ICE on IMERI 2019), pages 33-37

ISBN: 978-989-758-433-6

be precise so the radiograph from X-ray imaging

yields a good image with minimum radiation; this

also decreases the risk of X-rays radiation

(Lampignano and Bontrager, 2014).

2.1.1 Kilo Volt Potential

Energy from X-rays is controlled by a voltage

regulator. The potential difference setting usually has

the keV (kilo electron volt) or kVp (kilo Volt

potential) label. These are the important parts that

regulate the potential difference between anode and

cathode. The higher value of the potential difference,

the more energy is produced by X-rays (Omura,

2018). The high energy produced is contrary to the

contrast of the radiograph—the higher the energy, the

lower the image contrast.

2.1.2 Milliampere

Besides the potential difference, the current also

influences the imaging technique. This setting has the

label mA (milliampere); this represents how much

filament flows. The higher value of the mA (the tool

will get hotter) that is flowed through the filament, the

more electrons available in the ‘space charge’ to

accelerate through the target; the result is a high flux

of photons when energy flows. The effect of the

current is quite linear. To duplicate X-rays from the

tube, it can be done by doubling the previous tube

current settings. Changing the amount of current will

affect the blackness of the radiograph but it doesn’t

affect the contrast (Plaats, 1965).

2.1.3 Exposure Time

The final setting is the exposure time. With the s

(second) label, the exposure time is often associated

with regulating the tube current. The combination of

current and exposure time is often called mAs, or

milliAmpere second. For example, a 100-mA current

setting and a 0.5 s exposure time is the same as 50

mAs, as is a 50-mA current and 1 s exposure time; the

result is the same: 50 mAs. The combination of these

two factors is directly proportional to the effect on the

film (Sari and Fransiska, 2018). To produce a darker

radiograph, the value of mAs must be increased; and

to produce a brighter image, the value of mAs must

be reduced.

2.2 Body Mass Index

BMI is a comparison between weight and squared

height. The method of measurement is to measure his

weight and height. Then the BMI can be calculated

by:

BMI = (weight (kg)) / (height (m)) ^ 2

To determine the nutritional status of children

under five years of age (0–60 months) and children

aged 5–19 years, the BMI value should be compared

with the standard BMI value according to the

Republic of Indonesia’s Ministry of Health (2010). At

this time, the index is most often stated with Z-scores

or percentiles. Theoretically, the Z-score can be

calculated in the following way:

ZScore

BMI  Mean of BMI from reference

Standard Deviation from reference

Classification is distinguished in the 0–60-month

age group and also in the 5–18-year age group

(Munish, 2015). Classification of BMI for ages 0–60

months is presented in Table 1, while BMI

classification for children aged 5–18 years is

presented in Table 2.

Table 1: BMI for newborn baby 0–60 months.

Category Z-score value

Abnormal Z-score < -3

Underweight -3 ≤ Z-score < -2

Ideal -2 ≤ Z-score < 2

Overweight 2 ≤ Z-score < 3

Obese Z-score ≥ 3

Table 2: BMI for kids/teenagers 5–18 years old.

Category Z-score value

Abnormal Z-score < -3

Underweight -3 ≤ Z-score < -2

Ideal -2 ≤ Z-score < 2

Overweight 2 ≤ Z-score < 3

Obese Z-score ≥ 3

In adults, measurement of nutritional status is

done by using body mass index (BMI). BMI

calculation is the same as above. The results are

compared with the BMI threshold values according to

the Indonesian Ministry of Health, and the boundary

values are presented in Table 3. In adults the age

factor is not considered when calculating BMI,

because the height is usually relatively stable, so

variations only occur in body weight (Baş Mor,

2018).

The 4th ICE on IMERI 2019 - The annual International Conference and Exhibition on Indonesian Medical Education and Research Institute

Table 3: BMI for adults.

Category BMI value

Abnormal BMI < 17.0

Underweight 17.0 ≤ BMI < 18.5

Ideal 18.5 ≤ BMI < 25.0

Overweight 25.0 ≤ BMI < 27.0

Obese BMI ≥ 27.0

2.3 Linear Neural Network

A linear network can only solve linearly separable

problems since its transfer function is linear. This

allows their outputs to take on any value.

Figure 1: Layer in Neural Network.

Like the human brain, nerve tissue also consists of

several neurons, and there are connections between

these neurons. Figure 1 shows the structure of

neurons in which neurons will transform information

received through the output connection to other

neurons.

By using the function newlind in Matlab toolbox

as follows:

Net = newlind (P, T)

This returns a linear layer designed to output T

given input P (MathWorks, 2019).

This reference data was obtained from

Conventional X-rays device kVp and mAs sheet in a

hospital, thus the data will be used as an input (P) for

artificial neural network.

Table 4: Exposure Factor for Thorax in Neural Network.

Age

Proje

ction

Exposure Factor

Abnor

mal

Under

weigh

Ideal

Over

weig

Obese

kV/m

Newborn

baby

AP 55/2

(0–2 y) LAT 55/2.5

Baby AP 55/2.5

(2–5 y) LAT 55/3

Kids/Teen

ager

AP 51/4 53/4 55/4 57/4 59/4

(5–18 y) LAT 61/4 63/4 65/4 67/4 69/4

Adults AP 61/16 63/16 65/16 67/16 69/16

(> 18 y) LAT 71/20 73/20 75/20 77/20 79/20

3 RESULTS

Without using loss function and optimizer in the

training of the linear neural network, the

determination of X-rays that come out for overweight

patients and with underweight patients is different.

Overweight patients require higher doses because of

the larger surface area and density of the body.

Likewise, underweight patients will require a lower

dose because of smaller surface area and body

density.

To obtain a good-image quality, optimal X-ray

output settings are needed with the patient because

the higher the X-ray output settings, the greater the

dose received by patients.

4 CONCLUSIONS

An artificial neural network program can be used to

determine the thorax exposure factor in the

conventional X-ray devices. By entering body weight

and body height, this program calculates the optimum

value of exposure factors that can be used to the

patient.

The output of this program is the value of kVp and

mAs. Compared to the reference, the value of kVp

and mAs is under the predetermined range, which

means the software can determine the optimum

exposure factor. Optimum exposure factors yield a

minimum dose of radiation and good image quality.

Determination of Thorax Exposure Factors in Conventional X-rays Imaging using the Artiﬁcial Neural Network Method

Figure 2: Thorax AP Graph.

Figure 4: Thorax AP.

ACKNOWLEDGEMENT

The authors are thankful to the Dean of Faculty of

Mathematics and Natural Sciences at Jenderal

Soedirman University and his staff for providing the

necessary facilities to complete this paper.

REFERENCES

Baş Mor, H., Altinsoy, N., & Söyler, İ. (2018). Estimation

of Adult Patient Doses for Chest X-Ray Examinations

and Comparison with Diagnostic Reference Levels

(DRLs). Radiation Protection Dosimetry, 182(3), 377–

Figure 3: Thorax Lat Graph.

Figure 5: Thorax Lateral.

385. http://doi.org/10.1093/rpd/ncy076.

Bontrager, K. L., & Lampignano, J. P. (2014). Bontrager's

Handbook of Radiographic Positioning and Techniques

(8th ed.). St. Louis, MO: Elsevier.

Carlton, R. R., Adler, A. M., & Balac, V. (2019). Principles

of Radiographic Imaging: An Art and a Science (6th

ed.). Clifton Park, NY: Delmar, Cengage Learning.

Elster, A. D. (2010). Effective Doses in Radiology and

Diagnostic Nuclear Medicine: A Catalog. Yearbook of

Diagnostic Radiology, 2010, 214–217.

Gois, M. L. C., Schelin, H. R., Denyak, V., Bunick, A. P.,

Ledesma, J. A., & Paschuk, S. A. (2019). Human

factor in exposure from conventional radiographic

examinations in very and extremely low birth weight

patients. Radiation Physics and Chemistry, 155, 31–37.

http://doi.org/10.1016/j.radphyschem.2018.09.016.

0‐2y 2‐5y 5‐18y >18y

Abnormal Underweight

Ideal Overweight

Obsese

0‐2y 2‐5y 5‐18y >18y

Abnormal Underweight

Ideal Overweight

Obsese

The 4th ICE on IMERI 2019 - The annual International Conference and Exhibition on Indonesian Medical Education and Research Institute

Hiswara, E. (2002). Pokok Dosimetri, Lokakarya Proteksi

Radisasi, Pusat Standarisasi dan Penelitian (PSPKR).

Jakarta: Badan Tenaga Nuklir Nasional.

Kementrian Kesehatan RI (2011). Standar Azntropometri

Penilaian Status Gizi Anak.

Linear Neural Networks. (2019). Retrieved October 7,

2019, from https://www.mathworks.com/help/deep

learning/ug/linear-neural-networks.html

Munish, K. S. (2014). Accuracy of Conventional X-Rays in

Diagnosing Airway Foreign Bodies among Children.

Indian Journal of Public Health Research and

Development, 6(2), 4–6. http://doi.org/10.5958/0976-

5506.2015.00061.3.

Omura, Y. (2018). Impact of X-ray Radiation on SOI

MOSFET: Insulator Film Degradation and Hot-Carrier

Reliability. ECS Transactions, 85(8), 59–64.

http://doi.org/10.1149/08508.0059ecst.

Santoso, S., Haddin, M., Nuryanto, E., & Utomo, A. S.

(2016). Penentuan Faktor Ekposi Pada Pembangkit

Sinar-X Konvensional Dengan Menggunakan Logika

Fuzzy. Prosiding SNST Fakultas Teknik, 1(1), 56–61.

Sari, A. W., & Fransiska, E. (2018). Pengaruh Faktor

Eksposi dengan Ketebalan Objek pada Pemeriksaan

Foto Thorax Terhadap Gambaran Radiografi. Journal

of Health, 5(1), 17–21. http://doi.org/10.30590/vol5-

no1-p17-21.

van der Plaats, G. J. (1965). Medical X-Ray Technique:

Principles and Applications (2nd ed.). Springfield, IL:

C. C. Thomas.

Determination of Thorax Exposure Factors in Conventional X-rays Imaging using the Artiﬁcial Neural Network Method