Abstract
Osteoporosis means porous bone. If a healthy bone is looked under a microscope, it will look like a honeycomb. In case of osteoporosis, the holes and spaces in the honey comb are much larger. This means bone has lost its density. As bones become less dense they tend to break easily, the bones of the spine are at great risk. This paper is the part of the project which designs and develops a bone mineral density (BMD) detecting ultrasound machine. The unit is fully compact and portable and enables the BMD to be evaluated in real time by determining a parameter known as net time delay (NTD). Computer simulations are used to compare dependencies of NTD and BMD. It shows that the dependencies are very high. Hence concluding that NTD is highly proportional to bone mass and machine learning approach can be used to estimate the BMD of the subject. Looking to the severity of the disease and availability of expertise with ultrasonic system, it is proposed to develop low cost Ultrasonic based Bone Densitometer system to find the NTD thus the density of the bone and then find the T-Score in medical terms using Linear regression model.
Access provided by Autonomous University of Puebla. Download conference paper PDF
Similar content being viewed by others
1 Introduction
Bone density, or bone mineral density (BMD), is the quantity of bone mineral in bone tissue. The concept is of mass of mineral per volume of bone. Improper BMD can cause Osteoporosis i.e. “porous bone”. Artificial intelligence has been on the Radar of technology leaders for decades and these autonomous machines can help us making decisions or carry out complex tasks. Machines can do complex and stressful work faster than Human multiple task at a time. So in order to get appropriate data and accurate result, AI has been used. Current existing system for bone densitometer scanning uses X-rays, hence it is costly. Also, the size and weight of such a system is massive. Using X-rays might have some side effects on a person’s body. So, to avoid all these problems a new idea of using Ultrasonic sound for measuring the bone density is used. Bone density can be easily calculated using the difference between sound intensity at two different points. This system can be made handy, as size of the equipment needed isn’t large. Also, it’s setup is easy as compared to X-ray based system. So, in order to provide neat and elegant solution, we decided to use Ultrasonic sound based scanning system.
1.1 Purpose and Problem Definition
Osteoporosis is a major health related issue, endangering human lives, in turn causing collateral damage to various entities. Bone density importantly influences our lives. Apriori accurate detection of low bone density is a key that could prevent any further damage due to Osteoporosis. Powerful computational tools are needed for predicting the accurate result of low bone density. There is a need of compact and reliable system for predicting the proper bone density related issues. A branch of AI i.e. Machine Learning has many advantages like experiential prediction to produce optimal results in reasonable amount of time, suitable for predicting output by taking past results into account, so proposed system is based on Machine Learning to predict and categorize the bone density.
1.2 Scope
Bone density index prediction is a crucial issue when fighting with Osteoporosis. A direct prediction of person to have Osteoporosis or not cannot be determined by seeing results, as there are some cases in which even if bone density is low, chances of Osteoporosis are less. So, to neatly predict close to accurate results, machine Learning algorithms like Linear Regression are used. This reduces the uncertainty in output prediction and reduces false predictions. An intelligent system based on Machine Learning algorithms has been implemented for prediction of Osteoporosis.
2 Literature Survey
In [1], Normative information on BMD and body structure measured with dual energy x ray absorptiometry (DXA) were acquired from early age to adolescence. The most easily available and most frequently uses method for measuring BMC and BMD in kids is DXA. The major benefits of this method are its brief time, very small dose of radiation, high precision, and the capacity to evaluate BMC and BMD at both axial and appendicular skeletal locations. This article proposed that before the end of the second span of 10 years, there is a slight rise of the skeletal mass in the lumbar spine and complete body. This research offers reference value for kids and young adults for bone density and body composition measured with DXA.
In [2], The article writer has introduced a device that is completely mobile, light and allows the BMD to be assessed in live time by determining a variable known as NTD. The NTD is described as the difference between the travel period of an ultrasound signal via the heel and the travelling time through an equally thick (to the heel) hypothetical item but comprising only soft tissue. Author’s information shows that the fresh instrument and its related ultrasound parameter, the NTD, are highly susceptible to bone mass as measured by DXA scanners. On the other hand, the information demonstrate that the velocity of the ultrasound is much less correlated with BMD, this is due to differences in the quantity of overlying soft tissue density and calcaneus density between individuals. In short, a fresh instrument has been defined which, in specific, has the ability to widen the range of ultrasound application and bone screening in general.
In [3], The writer outlined the bone mineral density test in detail. X rays are utilized to evaluate the mineral quantity in bone. This test is essential for individuals having high potential of osteoporosis, particularly females and older adults. Also called as DXA. It is an significant test for osteoporosis.
In [4], Many efforts have been taken towards enhancing the quality. However, the presence of reliable BMD prediction techniques is still scarce, and they hardly include a real-time analysis simulation module. The main reason for this lack of confidence in prediction is uncertainty in prediction. Past attempts have made a direct prediction of person to have Osteoporosis or not by seeing indexes, which cannot be determined like that, as there are some cases in which even if bone density is low, chances of Osteoporosis are less. Also, as X-ray based systems are tedious to setup and are massive due to X-ray emitter systems, it is troublesome to take them far remote and rural areas. Similarly many researchers have worked in this area [5,6,7,8].
2.1 Benefits of the Proposed System
The solution to the problem defined above requires a system based around a model that can predict and categorize the bone density with improved accuracy despite the uncertainty in the input parameters involved. ML does exactly that. Using only data related to bone densities under analysis and also past data, algorithm can produce results in a reasonable amount of time, and is also suitable for parallel processing. The main advantages of the Proposed System are:
-
1.
Improved accuracy of Results: Algorithm has the ability to produce “closer to optimal” results. This leads to a better intuition of the possible category of bone density.
-
2.
Quicker Predictions: The system can make predictions in a “reasonable amount of time” while also “supporting parallel processing”, which results in quicker predictions and timely actions to respond to the situation with a proper strategy.
3 Proposed Methodology
3.1 Module Description
As Net Time Delay (NTD) of ultrasound and Bone Density (BD) are correlated, a Linear Regression model can be used for predicting BD according to NTD. Linear Regression is a method to predict the dependent variable (Y) based on the values of independent variables (X). It can be used for the cases where we want to predict some continuous quantity (Fig. 1).
Here the X axis is called independent axis, which will be plotted with NTD data. Y axis or dependent axis will be plotted with BD data. Regression line (as shown in blue) will be plotted optimally for all points. As it is assumed that the two variables are linearly related. Hence, we try to find a linear function that predicts the response value (y) as accurately as possible as a function of the feature or independent variable (x). The equation of the regression line is represented as: y = b0 + b1 * x.
Linear Regression to convert the Net Time Delay to Bone Density:
Output:
Thus now, we have got the prediction about bone density for given NTD. The data used is of a male’s lumbar spine bone density, age between 18–23 [1] (Fig. 2).
We have simulated a population mean here, after analyzing the real data we will get the real population mean.
Where BD is predicted value of BMD and popu_mean is the mean of population or data set taken for analysis.
3.2 Hardware and Software Requirements
-
1.
Software Requirements
-
Operating System: Any OS with Python support
-
Platform: Python 3.5 or above pre-installed with data analysis modules
-
Python libraries used: numpy, scipy, matplotlib
-
-
2.
Hardware Requirements
The hardware required for data collection is out of the scope of the project Computational hardware required
-
Minimum 4 GB RAM on system
-
Processor with high clock rate
-
USB input peripheral should be available
-
We will collect the data via a digitizer using PyUsb, do the computing on the NTD and calculate the bone density to get the T-Score.
4 Project Results
The system is trained with data obtained from a male’s lumbar spine bone density, age between 18–23. [1]
For value 0.04 as test value, we get
Predict(0.04)
Output:
We’ve got the prediction about bone density for given NTD. We will use this BD to obtain the statistical T-score for given density. A healthcare provider looks at the lowest T-score to diagnosis osteoporosis.
According to the World Health Organization (WHO):
Normal bone density range has T-score of −1.0 or above.
Low bone density or osteopenia has T-score between −1.0 and −2.5.
Diagnosis of osteoporosis has T-score of −2.5 or below.
Based on T-score, we can predict the category of bone density found for given bone sample, hence can predict chance of Osteoporosis.
Where BD is predicted value of BMD and popu_mean is the mean of population or data set taken for analysis.
5 Benefits of Proposed System
The system in its current form can accurately categorize whether the person is having Osteoporosis or not, depending upon the NTD value obtained from Ultrasonic Transceiver system.
-
1.
Accuracy: As the relation is found to be linearly regressive, it’s easy to get required BMD index and hence t-score from it, to predict.
-
2.
Speed: Once trained with given dataset, system can map any NTD value to BMD value and can predict Osteoporosis very easily.
-
3.
Cost Effectiveness: As this system uses Ultrasonic transceiver instead of classical X-Ray based heavy systems, not only the maneuvering cost, but the cost of operation also gets reduced.
6 Conclusion
The aim of the proposed system has been achieved. A well trained system can now predict Osteoporosis accurately in low cost operations. The ease of calculation is possible due to Linear Regression algorithm used for prediction. Trained well enough for accurate dataset, it provides neat and accurate basis for calculation of BMD value. Apriori detection of Osteoporosis can now be done easily to improvise doctor’s diagnostics.
References
van der Sluis, I.M., de Ridder, M.A., Boot, A.M., Krenning, E.P., de Muinck Keizer-Schrama, S.M.: Reference data for bone density and body composition measured with dual energy x ray absorptiometry in white children and young adults. Arch. Dis. Child. 87(4), 341–347 (2002). https://doi.org/10.1136/adc.87.4.341
Kaufman, J.J., Luo, G., Siffert, R.S.: A portable real-time ultrasonic bone densitometer. Ultrasound Med. Biol. 33(9), 1445–1452 (2007). https://doi.org/10.1016/j.ultrasmedbio.2007.04.007
Krans, B.: Bone Mineral Density Test. https://www.healthline.com/health/bone-mineral-density-test. Accessed 25 May 2018
Bochud, N., et al.: Predicting bone strength with ultrasonic guided waves. Sci. Rep. 7, 43628 (2017). https://doi.org/10.1038/srep43628
Jethe, J.V., Ananthakrishnan, T.S., Lakhe, A.S., Patkar, D.P., Parlikar, R.S., Jindal, G.D.: Development of ultrasonic pulser-receiver for bone density assessment. Int. J. Sci. Res. 8(9) (2019)
Chen, Y., Xu, Y., Ma, Z., Sun, Y.: Detection of bone density with ultrasound. Procedia Eng. 7, 371–376 (2010)
Stein, E.M., et al.: Clinical assessment of the 1/3 radius using a new desktop ultrasonic bone densitometer. Ultrasound Med. Biol. 39(3), 388–395 (2013)
Suzuki, T., et al.: Factors affecting bone mineral density among snowy region residents in Japan: analysis using multiple linear regression and Bayesian network model. Interact. J. Med. Res. 7(1), e10 (2018)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Joshi, B., Agarwal, S., Ragha, L., Yadav, N. (2020). Estimation of Bone Mineral Density Using Machine Learning Approach. In: Thampi, S., et al. Advances in Signal Processing and Intelligent Recognition Systems. SIRS 2019. Communications in Computer and Information Science, vol 1209. Springer, Singapore. https://doi.org/10.1007/978-981-15-4828-4_8
Download citation
DOI: https://doi.org/10.1007/978-981-15-4828-4_8
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-15-4827-7
Online ISBN: 978-981-15-4828-4
eBook Packages: Computer ScienceComputer Science (R0)