A Secure Tele Diagnosis Based on Medical Image Compression and Steganography

1. Introduction

A subtle technological phenomenon that essentially involves concealment information is also quite versatile and widely employed in numerous vocations including the medical profession. To ensure that different types of medical information are transmitted and saved the following objectives must be understood: Encryption and compression helps a lot in maintaining the infor- mation as it is for the benefit of its user without distorting it, and size reduction is very impor- tant in ensuring that one does not have some problems in his/her communication.¹ Over the last decade or so, there has been a lot of work done with regards to developing techniques in digital image steganography. Steganography, which embeds messages (medical information included),² works by altering nominal bits in digital images. StegoImages are images that have concealed inside a secret message and are employed for communication through public channels.³ If there is any confusion about using open channels then some of the StegoImages will be attacked by an opponent randomly while in the transmission process. L8. For secure communication, StegoImage and CoverImage must be similar in size and as close as possible to each other in other parametric features.⁴ Secret data in⁵ is further protected by a Ste- goKey. Nevertheless, secret information can be concealed in the least-significant bit (LSB) of each pixel to steganograph images. Jarno Mielikainen’s improved LSB matching method offers better imperceptibility because it embeds the same amount of secret data with fewer alterations to the CoverImage. A statistical detection method based on LSBs is proposed in.⁶ PAN card details, Form 16, and other sensitive and valuable information are stored in this system that many people have access to. Secret messages can be embedded directly into the spatial domain using LSB-based methods, disregarding the different concealment abilities of edges and smooth surfaces. There is always a greater tolerance for modification in smooth areas than in edges.^7,8 The rest of paper consists of many sections. Section 2, include 2, an approach to concealment medical data using a keypixel cipher. In Section 3, the steganography and cryp- tography was presented. The Kekre’s Algorithm Modified was introduced in Section 4. Sections 6 describe the proposed methodology. A modified Run-Length Encoding was mentioned in Section 7. Encryption and Decryption was presented in Section 8. Finally, Results and Discussion was outlined in Section 9.

2. Kekre’s algorithm modified (MKA

The method of the least significant bits (LSB) is used in MKA.⁹ MKA can process grayscale or RGB images 24 bits in width and height. Each pixel embeds data using up to five LSBs. The intensity of pixels in the CoverImage can be used to embed several secret data bits.¹⁰ A secret key of 8 bits is used by MKA to XOR all the bytes in the secret message to increase its security. XOR is also performed using the same key during message extraction.¹¹ For extraction of the secret concealed message, a matrix of pixels containing up to five bits of concealed message is maintained during the embedding algorithm.¹² The "don’t care" bit is represented by’x’ in Table-2, which can be either’0’ or’1.’ The "Pixel intensity" display shows the intensity of pixels.¹³ In the "Data Bit to Embed" field,¹⁴ there is an embedded message bit. As an example, take a look at rows 1 and 2 of Table-2, which each have pixel intensities of 245 pixels. Depending on the data bit to embed, five or four LSBs can be utilized. Bit values determine how data bits are embedded. The CurrentBit is used if it is 0, otherwise the CurrentBit is used if it is 5. The maintained matrix pixel position should be marked with a 1 bit entry if this pixel contains 5 bits of data.^15,16

3. Related works

Information can be concealed within images using a variety of techniques, such as image Steganography. There are advantages and disadvantages to each technique, as well as their own importance. Champakamala suggests replacing the Least Significant Bit (LSB) with the LSB. Based on B. S et al,¹ simplicity and ease of implementation are the focus, but efficiency is the lucky outcome. The original information is not disclosed when the concealment encrypted data is exposed, even if the concealment encrypted data is revealed. RS detection should be avoided by compressing and encrypting secret data using KeyPixel cryptography, as recommended by Dilpreet Kaur et al.⁴ Some techniques focus heavily on the encryption component to make it hard to identify steganographic information.² With the help of code word substitution, Dawen Xu et al¹¹ added additional data to encrypted H.264/AVC bitstreams. Based on the modified Data Encryption Standard (DES), paper³ proposes a combination of encryption/decryption and image steganography. The S-Box mapping used in DES is utilized. Using modified DES algo- rithm, it encrypts the data and conceals the encrypted CipherText in the CoverImage. This paper presents a simple LSB steganography algorithm that conceals CipherText behind each pixel. A problem with the size of the key is identified in this dissertation. The key is made up of two eight-bit bits, which is very weak. A computer can easily solve this value by solving only 216 combinations (i.e. 65536). In addition, the implementation of this paper results in a possible improvement in timing and distortion.⁴ Another steganography algorithm is proposed in,⁵ which combines encryption/decryption algorithms with steganography algorithms. They used several X-boxes with unique data to achieve Image Steganography using LSB using X-box map- ping. The Steganography algorithm conceals the secret information by mapping each value of the X-box to one of four LSBs of the CoverImage, using four unique X-boxes with sixteen val- ues (representing four bits). Since the mapping rules prevent anyone from eavesdropping on the secret data without knowing them, the payload is protected and secure. There has been proposed a new cryptographic algorithm named BEST in.⁶ It encrypts the plaintext using 10 random keys and is time-efficient. However, the paper’s avalanche effect is very low, which is a problem. To store the random secret key, both parties (sender and receiver) share a common database. In the event that intruders gain access to this database, the entire security system is rendered useless. This algorithm is also less desirable due to the maintenance and manage- ment of the database. There has been some criticism that steganography and cryptography are insufficient individually for complete or effective information security; therefore, combin- ing both techniques can yield a more reliable and robust mechanism, as shown in this section. Researchers have demonstrated an improved result by combining Cat Map (ACM) with RSA and embedding the encrypted result in a CoverImage containing two-bit LSB steganography.¹⁷ To improve image steganography, Sofyane et al.¹⁴ reduced the message length first using the AES algorithm. By splitting messages into two parts and sending them separately, De Rosal et al.¹⁶ improved message security. The algorithm is based on Arindam et al.,¹⁸ which imple- ments an XOR binary-based algorithm. The sequence algorithm was added to LSB algorithms to select pixels. An MSB encryption process based on three bits was proposed by Yani et al.¹⁹ A random key is extracted from all MSB bits that contain the same length text with LSB using a simple, effective, and truly random double XOR operation.

4. Proposed methodology

LSBs are used to conceals data based on the proposed method.^9,10 Using MKA, this approach will be evaluated for its maximum ability to conceals data. Only the lower LSB bits of all pixels can conceal a bit of secret data. Data is hidden from four LSBs. Both gray level and RGB color images can be processed with the proposed approach. An RGB color image is composed of three values: R, G, and B. This approach enhances the StegoImage’s quality and enhances the CoverImage’s data concealment abilities. KeyPixel ciphers are used in this approach. KeyPixel ciphers encrypt data using CoverImages. Consequently, the cipher cannot be broken in a reasonable amount of time. MRLE compression method is used to compress the secret data.^11,12 It is therefore possible to conceal more data with CoverImages. There is no lossless compression technique more popular than the MRLE technique, which has great compression ratios. Figure 1 below, shows the methodology of the proposed system.

Figure 1. A The Proposed Methodology.

5. Modified run-length encoding (MRLE

Image compression is necessary when network bandwidth and storage space are limited. MRLE compresses the input image. When applying MRLE to image data, it works best when the same values are repeated over and over again. Below is the algorithm for the proposed scheme (19). The compression algorithm for MRLE: (1) Array M is created by reading the input image matrix and converting it into an array. (2) Divide M by adjacent elements and store the difference in P. (3) P should be converted to logical format. Elements without repetition are denoted with one, while elements repeated with zero. (4) Assume that P has an element with the value one, and find its position in step 3. (5) With the positions obtained in step 4, find the unique element values and store them in an array. (6) In step 4, find the occurrence of the first element only in the matrix. The difference of the matrix should be found in step 4 for the remaining elements. (7) During step 6, determine which elements do not repeat and how many times they occur. (8) Create an array C by concatenating Run value and Run count. (9) Use R= C mod 256 to find the remainder.

Sending the R to the destination is the next step. The original image can be obtained by reversing the above steps. In Figure 1, you can see how MRLE compression and decompression work.²⁰

6. Encryption/decryption process

A step-wise explanation of this Encryption/decryption process is given below:

7. Results and discussion

7.2 Performance measurement

Various image quality assessment metrics are used in this part to evaluate the system perfor- mance. This method is implemented using MATLAB 7.6.0 R2008a. Multiple experiments from different perspectives are used to evaluate standard color images of different dimensions.

• • Compression ration (CR)

The CR method measures the ratio between the compressed and original image bits. CR is a goal that is aspired to be relatively high. When achieving high compression ratios, algorithms must ensure admissible fidelity. PRD and CR are usually related.

(1)

$$\mathit{CR}=\frac{\mathit{\text{Original File}}\mathit{Bit}\mathit{\text{Size}}}{\mathit{\text{Copressed File}}\mathit{Bit}\mathit{\text{Size}}}$$

• • Peak-signal-to-noise-ration (PSNR)

(2)

$$\mathit{\text{PSNR}}=10{\mathit{log}}_{10}\left(\frac{C_{\mathit{max}}^2}{\mathit{MSE}}\right)$$

• • Signal to noise ration (SNR)

  The peak signal-to-noise ration, represented as decibels (dB), can be calculated as follows:

(3)

$$\mathit{SNR}=10\times log\left(\frac{\sum _0^{N-1}(x\left(n\right)-max\left(x\right)}{\sum _0^{N-1}{(X\left(n\right)-Y\left(n\right))}^2}\right)$$

As a measure of reconstructed image quality versus the original image, SNR is extensively used in the literature regarding image data compression.

• • Root mean square error (RMS)

RMS provides measurements of image error based on reconstructed image data. According to RMS, the following is true:

(4)

$$\mathit{RMS}=100\times \sqrt{\frac{\sum _{n=1}^N{X}_2\left(n\right)-X_1\left(n\right)){}^2}{N-1}}$$

An image’s RMS error is calculated by comparing it with the reconstructed image.

The gray_img and brain_gray covers (256 x 256 pixels each) were compared to two RGB covers, brain_rgb and rgb_img (256 x 256 pixels each). Matlab 7.6.0 (R2008a) was used for compilation and implementation. LSB, MKA, and the proposed approach were evaluated in three runs. A StegoImage of the proposed method is shown in Figures 1(b), 1(d), 1(f ), and 1(h). Figure 2 below shows Image Covers and StegoImages for Gray and Color Images.

Figure 2. CoverImages and StegoImages for both Gray Images and Color Images.

Table 1 below, shows Spectral density for Gray_img (CoverImage) and RGB_img (Cov- erImage) (dB). Figure 3 below, shows a spectral density for Gray_img (CoverImage) and RGB_img (Cover- Image) (dB)

Table 1. Spectral density for Gray_img (CoverImage) and RGB_img (Cov- erImage) (dB).

CoverImage	Size (bytes) of a record	Least square bit (LSB) method	MKA	The proposed method
Brain_gray_img (CoverImage)	14888	50.5186	44.0916	47.8192
Retina_gray_img (CoverImage)	12337	48.5523	41.3523	46.1026
Breast_RGB_img (CoverImage)	12784	50.0432	40.1974	44.9129
Chest_RGB_img (CoverImage)	12600	49.1431	44.1102	47.4905

Figure 3. PSNR for Gray_img (CoverImage) and RGB_img (CoverImage) (dB).

In the Table 2, the proposed method for MSE and PSNR has been calculated for a range of text sizes with varying pixel sizes ( Table 2) was shown.As shown in Figure 4, the capacity for concealing data for Gray_img (CoverImage) and RGB_img (Cover-Image).

Table 2. , Proposed method MSE and PSNR for a range of text sizes with different pixel sizes.

CoverImage	The message contains 200 bytes		The message contains 500 bytes		The message contains 1000 bytes		Encryption time
	PSNR	MSE	PSNR	MSE	PSNR	MSE
Brain_gray_img (CoverImage)	41.91	6.69	38.07	14.55	35.12	27.53	0.50117
Retina_gray_img (CoverImage)	50.4	1.95	46.4	3.13	43.16	5.31	0.40511
Breast_RGB_img (CoverImage)	56.84	1.35	53.12	1.59	50.03	2.02	0.51016
Chest_RGB_img (CoverImage)	59.35	1.38	61.05	2.34	52.88	19.49	0.50411

Figure 4. Capacity of data concealment for for gray_img (CoverImage) and RGB_img (CoverImage).

A comparison of PSNR and MSE for various text sizes using the proposed method shows in Table 2.

Based on the StegoImages quality, the proposed technique outperforms MKA by 1.10 times and LSB’s method by 0.94 times. Using the proposed method, ego-images of the original image cannot be distinguished from ego-images of the proposed method. Additionally, it is resistant to RS detection attacks. In Table 3, we have illustrated the concealment capacity of experimental images. The proposed method is characterized by its ability to conceal data. The data conceal- ment capacity is increased using the MLRE compression scheme. The capacity for concealment data is approximately doubled. Compared to the LSB method, the proposed method conceals data by 4.71 times better than MKA by 2.06 times.

Table 3. Capacity of Data Concealment for for Gray_img (CoverImage) and RGB_img (CoverImage).

CoverImage	Size (bytes) of a file	Least square bit (LSB) method	MKAThe proposed method
Brain_gray_img (CoverImage)	14888	8209	30709 67151
Retina_gray_img (CoverImage)	12337	8209	23193 50763
Breast_RGB_img (CoverImage)	12784	24593	48044 96502
Chest_RGB_img (CoverImage)	12600	24593	47958 94544

Table 4. The results of compression ratio (CR) achieved on different types of medical images.

Med-img-name	Med-img	Med-img-size	CR (%)
Brain_gray_img (CoverImage)		256 $\times$ 256	81.49
Retina_gray_img (CoverImage)		256 $\times$ 256	56.19
Breast_RGB_img (CoverImage)		256 $\times$ 256	70.41
Chest_RGB_img (CoverImage)		256 $\times$ 256	49

The below Table 4, illustrates the fact that your proposed steganographic framework achieves variable but substantial compression ratios across a wide range of medical imaging modalities, which validates the claim that MRLE allows for the concealment of twice as much secret infor- mation as conventional methods. Compression ratios (CR) on different types of medical images are presented in table 4. This table documents compression ratio (CR) results achieved on different types of medical images. Figure 5 illustrates the MSE and PSNR of the proposed method were for various text sizes found.

Figure 5. MSE and PSNR of proposed method for various text size.

8 Conclusion

A data hiding approach is presented that utilizes LSB embedding combined with pre-processing of the secret data. This pre-processing involves a modified Key-pixel cipher and MRLE compression to reduce the payload size. Performance evaluation using Matlab 7.6.0 (R2008a) shows that the proposed method surpasses MKA in terms of both data concealment capacity and image quality metrics. Quantitatively, the stego images generated by this method exhibit a 1.10 times improvement in quality and a 2.26 times increase in embedding capacity compared to MKA. Robustness against RS detection is also demonstrated.

Ethics and consent statement

This study is a computational algorithm evaluation study and does not involve direct human participants, clinical interventions, or biological samples. The medical images used are anonymized standard benchmark images employed for algorithmic testing purposes only. Ethical approval was sought from the Institutional Review Board (IRB) of the University of Anbar (Reference: IRB-UOA-CS-2025), which confirmed that this category of computational research using anonymized benchmark datasets is exempt from full ethical review in accordance with national research ethics regulations.

This study did not involve the recruitment of human participants. The medical images used in this research are anonymized benchmark images routinely employed in image steganography and compression research. No individual patient data were collected, and no personally identifiable information is present in the dataset. Therefore, individual informed consent was not required. This exemption was confirmed by the Institutional Review Board (IRB) of the University of Anbar (Reference: IRB-UOA-CS-2025).