Unique identification code for medical fundus images using blood vessel pattern for tele-ophthalmology applications
Introduction
Sharing of medical images has wide variety of applications like tele-diagnosis, tele-surgery, tele-medicine, tele-opthalmology, etc. It has been found that security and identification of medical images have been a critical issue for the patients [1], [2] in these applications. Patient identification and distinct identification are important requirements in clinical practices to prevent medical errors. Accurate identification of medical image is a serious issue as any error in such cases can cause threat to proper diagnosis. In addition to transmission of medical images, the issue of correct identification is required for medical images when stored in databases. Hence there is a need to develop efficient and secure algorithms for correct identification of medical images during transmission and storage for such applications.
The existing work to solve the issue of medical image identification is based on steganography and medical image watermarking methods. In the existing literature, different watermarking methods have been proposed to verify medical image integrity. The methods of medical image watermarking can be broadly classified in three different categories [3]. The first category is based on the segmentation of medical image into medically informative (ROI) and noninformative regions (RONI), and then medically noninformative regions are considered for hiding the patent identity. Qershi et al. [4] developed a hybrid ROI-based watermarking method in which the medical image is divided into three regions: ROI, RONI, and border pixels, and then the patient's ID and hash value of ROI are embedded inside ROI using modified DE technique. Wu et al. [5] developed region based watermarking methods for medical image identification by embedding patient information into medically noninformative region. Kim et al. [6] proposed a region-based tampering detection and restoring scheme based on image homogeneity analysis for medical image identification and integrity verification.
The second category in medical image watermarking is the reversible watermarking method that ensures that original image is completely recovered without losing any information of the image. Gouenou Coatrieux et al. [7], [8] developed algorithms for reversible Watermarking to solve the issue of medical image identification. Hirak Kumar Maitya et al. [9], Feng Bao et al. [10], Muhammad Arsalan et al. [11] reported reversible image watermarking methods for medical image identification and integrity verification.
In the third category of watermarking methods, the patient ID is embedded in the host images without degrading the quality of image to an extent where the medical information in the image is not lost. Vleeschouwer et al. [12], Malay Kishore Dutta et al. [13], [14] developed image watermarking algorithms for correct identification of medical images during transmission and storage in such applications.
Robust and imperceptible image watermarking method can be an option to solve the issue of image identification but it may tamper the image and there may be loss of medical information in the image. In the case of medical image identification where any compromise with information loss is not acceptable, watermarking method is not a right choice as any error in such cases can cause threat to proper diagnosis. Hence, there is a need to develop algorithms for accurate identification of medical images without causing any loss of medical information. This paper presents an algorithm for unique identification code generation for medical image identification for tele-medicine applications. This work proposes an algorithm for generation of unique identification code for accurate identification of medical images in such applications and also may be useful for identification of medical images in databases.
The main contribution of this paper is an algorithm to generate unique identification code for fundus images to provide a possible solution in the direction of medical image identification for tele-ophthalmology applications. Local features from the blood vessels of the image are used to generate a unique pattern which ensures that it will comply for accurate identification. This unique pattern is strategically combined with a patient ID resulting into a unique identification code (UIC) making the method a full proof framework for distinct identification of medical images. At the diagnosis center the patient ID is accurately recovered from the unique identification code for that particular image to solve the issue of integrity verification.
Another significant contribution of the proposed method is unlike existing methods of steganography and watermarking, here there is no tampering or embedding in the medical image and hence there is no degradation in the quality of the image. The encryption method used is lossless ensuring that the unique identification code retains all the information. This may be considered as a significant development to ensure distinct identification of medical images accurately without loss of information.
The proposed algorithm seems to be an appropriate solution to solve the issue of threat of loss of identity of medical images during transmission and data storage. The proposed method may be of great help in tele-ophthalmology applications where perfect and uncompromised identification is required for medical images as this has direct relevance to proper diagnosis.
The rest of the paper is organized as follows. Section 2 tells the design challenges associated with generation of identification code for medical images. Section 3 describes the unique features of the fundus image which are used for generation of identification code. The next section discusses the proposed methodology, including the generation of unique identification code for fundus image and recovery of patient ID for integrity verification of image. Section 5 presents the experimental results, and the next section includes the final discussion and remarks on experimental results. Section 7 draws a conclusion to the paper and discusses some future works which may be possible in this area of work.
Section snippets
The design challenges in identification code generation for medical images
There are design challenges associated in identification code generation for medical images to address the issue of medical image identification during transmission and storage of medical images in tele-ophthalmology applications. Some of the design challenges are mentioned below:
- 1.
The identification code used for medical image security should be unique so that the particular image is uniquely identified in medical image database. The generated identification code should be visibly noninformative
Fundus image—unique features
Fundus image is the photograph of the interior surface of the eye, which includes the retinal blood vessels, optic disk and macula. Optic disc is a bright spot where retinal blood vessels converge. Fig. 1 represents digital fundus image of left and right eye. Optic disc, macula, fovea and bold vessels are marked in the fundus image shown below:
In 1935 two ophthalmologists, Drs. Carleton Simon and Isodore Goldstein, while studying eye disease, realized that every eye has its own totally unique
The proposed system
The proposed work provides a solution to the issue of security, identification and integrity verification of fundus images during transmission/storage in tele-ophthalmology applications. To secure the fundus images, unique identification code is generated for fundus images using strategic combination of fundus image unique features and its patient ID. For the identification and integrity verification of fundus image, patient ID is recovered from the unique identification code. Fig. 8 represents
Experimental results
Experiments have been carried out in publicly available DRIVE database of fundus images [33] to generate unique identification code for image identification. DRIVE database contains 40 fundus images of 565 × 584 pixels in size and are provided in TIFF format. Unique identification code generated for these images are of size 160 × 160. The proposed image identification method is tested on all 40 images of DRIVE database and results are reported in this section. Fig. 11(a,b,c) shows five samples
Discussion and final remarks on experimental results
- •
To solve the issue of identification of medical image in tele-ophthalmology applications, a unique identification code is generated for fundus image by strategic combination of unique blood vessel feature and patient ID. The generated unique code is encrypted to maintain the security.
- •
The proposed method does not tamper the original image resulting into 100% accuracy of identification and no loss of information.
- •
The encryption technique used in the proposed method to secure the medical
Conclusion
Identification of medical images during transmission and storage in large databases is an important issue, as any false identification may lead to adverse affect in proper diagnosis. This issue has been addressed in this paper using generation of unique identification code for fundus image. Unlike the existing method of stenography and watermarking, the proposed method does not tamper the medical image for integrity verification. Unique retinal blood vessels pattern near the optic nerve and
Acknowledgement
This work was supported in part by the Grants from Department of Science and Technology, No. “DST/TSG/ICT/2013/37”. Also the authors express their thankfulness to Dr. S.C. Gupta, Medical Director of Venu Eye Institute and Research Center for his kind support.
References (37)
Privacy and security in teleradiology
Eur. J. Radiol
(2010)- et al.
Region-based tampering detection and recovery using homogeneity analysis in quality-sensitive imaging
Comput. Vis. Image Underst
(2011) - et al.
Intelligent reversible watermarking in integer wavelet domain for medical images
J. Syst. Softw
(2012) - et al.
An adaptive threshold based image processing technique for improved glaucoma detection & classification
Comput. Methods Programs Biomed
(2015) - et al.
Automatic model-based tracing algorithm for vessel segmentation and diameter estimation
Comput. Methods Programs Biomed
(2010) - et al.
A pixel-based scrambling scheme for digital medical images protection
J. Netw. Comput. Appl
(2009) - et al.
Potential impact of HITECH security regulations on medical imaging
(2009) - et al.
Strict integrity control of biomedical images, Electronic Imaging 2001
(2001) - et al.
Identification and data hiding using a hybrid ROI-based watermarking scheme for DICOM images
J. Digit. Imaging
(2011) - et al.
Tamper detection and recovery for medical images using near-lossless information hiding technique
J. Digit. Imaging
(2008)
Reversible watermarking for knowledge digest embedding and reliability control in medical images
IEEE Trans. Info. Tech. Biomed
A low distortion and reversible watermark application to angiographic images of the retina,
Joint robust and reversible watermarking for medical images
Tailored reversible watermarking schemes for identification of electronic clinical atlas
Circular interpretation of bijective transformations in lossless watermarking for media asset management
IEEE Trans. Multimedia
Imperceptible digital watermarking in medical retinal images for tele-medicine applications
Digital identification tags for medical fundus images for tele-ophthalmology applications
A new scientific method of identification
N. Y. State J. Med
Cited by (14)
RetIS: Unique Identification System of Goats through Retinal Analysis
2021, Computers and Electronics in AgricultureCitation Excerpt :Although, the process of acquiring images for retinal analysis is not time-efficient, but the uniqueness in the vascular patterns of the retinal images makes it nearly impossible to forge contrary to that of iris. The uniqueness of the retinal vascular pattern for every individual based (used for authentication) (Deljavan Amiri et al., 2009; Fukuta et al., 2008; Singh et al., 2016) had been established and later published in a research paper by Drs. Carleton and Goldstein during the study of eye diseases (Bevilacqua et al., 2008). Later, it had been supported by Dr. Tower in the course of studying two identical twins wherein amidst of most likely physiological factors, the retinal vascular pattern had been found completely different in one another (TOWER, 1955).
IOT based sustainable diabetic retinopathy diagnosis system
2020, Sustainable Computing: Informatics and SystemsCitation Excerpt :It deposits exudates, lesions, microaneurysms, hemorrhages and cotton wool spots. The manual assessment of retinal blood vessel is impossible, since measurement of vascular width is most crucial [8]. This automated sustainable disease detection system significantly reduces the load of ophthalmologists.
Imperceptible watermarking for security of fundus images in tele-ophthalmology applications and computer-aided diagnosis of retina diseases
2017, International Journal of Medical InformaticsCitation Excerpt :Transmission of medical images over the network and storage of images in databases is an essential requirement for providing e-health services. [1]. Identification and authentication of medical images is a critical and growing issue for patients in e-health services like Tele-medicine, Tele-diagnosis, Tele-radiology applications [2,3]. Accurate identification of medical images and correct patient verification is an essential requirement to prevent error in medical diagnosis and tele-medicine applications.