Compression of Medical Images Using Wavelet Transform Compression of Medical Images Using Wavelet Transform and Metaheuristic Algorithm for Telemedicine Applications

- Medical image compression becomes necessary to efficiently handle huge number of medical images for storage and transmission purposes. Wavelet transform is one of the popular techniques widely used for medical image compression. However, these methods have some limitations like discontinuity which occurs when reducing image size employing thresholding method. To overcome this, optimization method is considered with the available compression methods. In this paper, a method is proposed for efficient compression of medical images based on integer wavelet transform and modified grasshopper optimization algorithm. Medical images are pre-processed using hybrid median filter to discard noise and then decomposed using integer wavelet transform. The proposed method employed modified grasshopper optimization algorithm to select the optimal coefficients for efficient compression and decompression. Four different imaging techniques, particularly magnetic resonance imaging, computed tomography, ultrasound, and X-ray, were used in a series of tests. The suggested method's compressing performance is proven by comparing it to well-known approaches in terms of mean square error, peak signal to noise ratio, and mean structural similarity index at various compression ratios. The findings showed that the proposed approach provided effective compression with high decompression image quality.


░1. INTRODUCTION
Medical images play a significant role in identification, diagnosis and surgical planning of disorders. These images have large volume of relevant information that assist physicians or doctors in analysing more accurately and planning the diagnosis for patients. For the purposes of medical history and future reference of patients, medical images should be stored. Due to the advancements in imaging techniques, the amount of generation of medical images for both diagnosis and telemedicine applications has increased in recent years. However, because to a lack of memory and limited transmission bandwidth, hospitals find it difficult to store and transmit medical images. To rectify the aforementioned concerns, image compression is introduced. Image compression aims to minimize the size of medical images for efficient transmission and storage while keeping image quality for diagnostic purposes [3].
Several image compression techniques have been developed so far for compressing medical images [12]. However, most of the techniques able to yield good outputs at low Compression Ratio (CR) and memory and computation cost are yet to be a critical issue. Therefore, compressing medical images at high CR without losing important data is an extremely challenging task. Recently, metaheuristic algorithms are integrated with image compression methods to enhance the CR and reduce storage space demand. Researches have used the metaheuristic algorithm as an optimization method for image compression [5] [6] [11]. Medical images are compressed, and certain critical details are lost, posing a risk during analysis and diagnosis. Thus, this paper presents a novel hybrid compression algorithm by integrating Integer Wavelet Transform (IWT) and metaheuristic algorithm. The major contribution of this paper are as follows: • A new optimized image compression method is proposed by using Integer Wavelet Transform (IWT) and Modified Grasshopper Optimization Algorithm (MGOA), named as MGOA-IWT. • Hybrid Median Filter (HMF) is proposed to suppress noisy data.
• The proposed algorithm, MGOA select the optimal coefficients for compression and decompression

Compression of Medical Images Using Wavelet Transform and Metaheuristic Algorithm for Telemedicine Applications ░ 2. REVIEW OF PREVIOUS METHODS
Alkinami et al. [1] combined WT and Particle Swarm Optimization (PSO) for developing an efficient method for image compression scheme. Hosny et al. [2] presented a compression method by using Legendre moments and whale optimization algorithm. Sreenivasulu et al. [3] presented a method for image compression which is based on wavelet based modified region grouping algorithm. Ammah et al. [5] used Daubechies WT and Huffman encoding for compressing biomedical images. However, WT based compression has some drawbacks like poor reconstruction due to floating points. Medical image compression by applying harmony search algorithm was introduced by Haridoss et al. [4]. Region of Interest (ROI) based medical image compression by applying modified rider optimization algorithm was implemented by Sreenivasulu et al. [6]. Honsy et al. [7] proposed a Tchebichef moments and artificial bee colony algorithm-based picture compression approach. Wu et al. [9] introduced image compression method using genetic algorithm and discrete WT. A hybrid algorithm based on IWT and Particle Swarm Optimization (PSO) algorithm was invented by Vijayvargia et al. [11]. Bouetta et al. [12] used discrete WT and genetic algorithm for compressing images without loss of data. The prime objective of this work is to integrate IWT and MGOA to reduce the size of medical images of different modalities for efficient storage and transmission. The complete processes involved in the proposed method is evinced in figure -1.

Noise Suppression
The Hybrid Median Filter (HMF) is used to eliminate noise from medical images in this study. Let a kernel with size of 3 X 3 shown in Figure. 2, HMF replaces the pixel 'P' from the median values of three values, as given in Equation (1), In this investigation, salt and pepper noise for MRI, CT and Xray images and speckle noise for US images is considered to validate the de-noising potential of the HMF. Kernel size of HMF is 3 X 3. Sample images before and after filtering is given in figure-3.

Decomposition
In this research work, IWT is adopted for decomposition. The IWT has three phases split, predict and update. Split: Let the input signal x(n) is divided into even samples Xe(n) and odd samples Xo(n) Predict: Both Xe and Xo are obtained by dividing X. So, there exists a correlation between them. A Predictor P is applied on the Xe(n) and then difference between P[Xe(n)] and Xo(n) turns out as the detail signal. Detail signal can be expressed as, Update: An update operator U is applied on D(n) and then Xe(n) is modified by U[D(n)] to get low frequency component, A.

Coefficient Optimization
GOA is a recent bio inspired algorithm developed by Saremi et al. [8] for solving optimization problems. According to Saremi et al. [8], the swarming behavior of grasshopper can be expressed as,

International Journal of Electrical and Electronics Research (IJEER) Open Access | Rapid and quality publishing Research Article | Volume 10, Issue 2 | Pages 161-166 | e-ISSN: 2347-470X
Where, X is the Position of the i th grasshopper, S, G and A are social interaction, gravity force and wind advection respectively. Social interaction, S can be defined as, Image type Noisy image De-noised image US Figure 3: Sample medical image before and after noise suppression X i Where, dij is distance between i th and j th grasshopper and s is the Social force and s can be defined as, Gravitational force, G can be computed using Equation (11), Wind advection, A can be calculated using Equation (12), Where, f denotes intensity of attraction, l is attractive length scale, g is gravitational constant, ̂,u and ̂ are centre of earth, constant drift and wind direction . Substituting Equation (7), Equation (11) and Equation (12) in Equation (6), Equation (13) cannot be used to solve real time or complex problems, because grasshoppers achieve their comfort zone and do not cluster on a single location. Therefore, Equation (13) is modified by adding some parameters to solve complex problems. The modified version is given in Equation (14).
Where, is upper bound in the d th dimension, is lower bound in the d th dimension, ̂ is value of the dimension in the target, w is decreasing coefficient , ,l is current iteration, L is maximum number of iterations, wmax and wmin are maximum value and minimum value respectively. (14), the value of w decreases linearly from 1 to minimum value to achieve best solution. But, this characteristic causes local optima problem. To address such an issue, w update equation proposed in this work and delineated in Equation (16). The proposed update equation is adaptive with iteration and non-linear.

As in Equation
In this work, MGOA is adopted to select the optimal coefficients in order to enhance the quality of recovered image. Based on the CR, Optimal Number of Coefficients (ONC) are computed as represented in equation (17).
During first iteration, decomposed image is divided into m X m (8 X 8) sub block. Equation (17) is utilized for each submatrix to select optimal coefficients where maximization of PSNR as fitness function of the MGOA. The selected coefficients are encoded, compressed and then decompressed. The PSNR between the input and reconstructed image is computed. This process is repeated to a predefined iteration. The selected coefficients which gives higher PSNR is taken as optimized coefficients. Optimization workflow is given in the form of algorithmic steps in Table. 1.

░ 4. RESULTS AND DISCUSSION
For validation, a collection of medical images from various modalities such as MRT, CT, US, and X-ray are acquired from a publically available source.

Image Quality Evaluation Metrics
To gauge the compression ability of the proposed method, four performance measuring metrics namely Mean Square Error (MSE), PSNR and Mean Structural Similarity Index (MSSIM) between the original (O) and recovered (R) images were computed. The higher values of these metrics, the better the compression obtained.

Performance Analysis and Comparison
The proposed medical image compression method has been implemented in MATLAB 2019a and validated with a large set of medical images of different imaging modalities. The medical images of four different imaging modalities, MRI, CT, US and X-ray were compressed using both standard GOA and MGOA for CRs varies from 40% to 90%. The compression efficiency of both GOA and MGOA was assessed in a quantitative form where the obtained values were reported in Table 2 for MRI and CT images and Table 3 for US and X-ray images. From observations, the proposed MGOA-IWT algorithm significantly improves compression when compared to the GOA-IWT compression algorithm in almost all the CRs.

International Journal of Electrical and Electronics Research (IJEER) Open Access | Rapid and quality publishing Research Article | Volume 10, Issue 2 | Pages 161-166 | e-ISSN: 2347-470X
For MRI image and CR is 40%, GOA-IWT method achieved mean MSE, PSNR and MSSIM are 0.0730, 59.68dB and 0.8824 respectively. The proposed MGOA-IWT compression method attained mean MSE is 0.008, mean PSNR is 79.12 dB and mean MSSIM is 0.9988. The obtained values clearly showed that the proposed compression method produce better results than that of the conventional compression algorithm. Figure 5 provides the graphical delineation of Table 2.
To further prove the performance, a comparison with the existing methods was done such as Harmony Search based Compression (HSA) method [9], PSO based method [10] and Whale Optimization Algorithm (WOA) based method [11]. Figure 6 shows the comparison in terms of PSNR for MRI images at various CRs. It is inferred from the figure 9 that the MGOA-IWT compression method achieved higher PSNR for almost all CRs which indicates that the decompressed image is closer to the original images in comparison with the existing methods taken for comparison. The overall analysis confirmed the superiority of the MGOA-IWT compression method.  Table 3. Performance comparison of the GOA-IWT and MGOA-IWT for US and X-ray image   IMAGE TYPE  METRICS GOA-IWT  MGOA-IWT  CR (%)  40  50  60  70  80  85  40  50  60  70  80