Digital cameras and custom software are helping ophthalmologists screen patients for retinal damage.
By Joe Hallett,Contributing Editor
High-resolution digital cameras and image-analysis software are enabling ophthalmologists to identify patients with eye problems by means of automated diabetic retinopathy. Automated methods screen patients for signs of diabetes and eye diseases with reduced time, effort, and cost of patient evaluation. And lightweight, portable, digital camera instruments are making it possible to obtain rapid evaluations of children and patients at home or in areas with limited access to health-care facilities.
For example, equipment developed by Digital Heathcare Ltd. (Cambridge, England) combines digital imaging and software analysis to spot signs of retinal damage that indicate possible diabetes. According to company managing director Gerry Skews, automating the process will help doctors to address prevalent and costly eye diseases. "We've seen a lot of interest in the analysis process," says Skews. "Interest in diabetes is phenomenal because it is a large and growing problem."
Citing US Center for Disease Control data, Skews reports, "It costs $62 billion per year to combat diabetes in the United States, which is the leading cause of blindness among people of ages 20 to 74. In fact, there was a 30% growth in diabetes between 1990 and 1998," he adds.
Fundus (retinal) cameras are proven devices that permit eye doctors and clinicians to examine the retina of a patient's eye. Typically, photographs provide images that can be displayed for "off-line" analysis, as well as a physical record of an examination.
Symptoms of diabetes as well as various other diseases can be identified by a skilled imaging analyst. Video output enables easier viewing of images, but lacks the resolution and precision for certain applications, including the automatic classification and grading of small defects, such as microaneurysms.
Digital technologies are enabling more efficient use of fundus cameras, since digital images are more easily stored and transported with other patient information than are film images. Meanwhile, advanced research at sites such as the University of Aberdeen in Scotland has led to image-processing algorithms that can be used to screen digital images for signs of a specific disease.
Digital Healthcare, in partnership with Canon Medical Systems (Irvine, CA), has developed an automatic screening system, the Canon Eye-Q line, which is currently going though the US government approval process. "The software can scan thousands of images for pathologies in the retina to determine whether disease is present," says Skews. "We are just completing clinical trials and launch."
He adds, "The system doesn't replace an expert analyst. It just presents images that indicate the need for patient attention, based upon the number of microaneurysms present."
System operation is straightforward, according to Skews. Images are taken in individual "capture stations." The patient sits before a fundus camera, which is attached to a digital camera back (see Fig. 1). After the camera is focused on the retina, the operator takes a picture (see Fig. 2). Automatically, an image file is sent to a Windows-NT-based 800-MHz Pentium III desktop computer with a large hard drive.
Says Canon's Mark Sheckel, "An individual capture station is a stand-alone unit that provides capture and review. But in larger systems, the microaneurysm software resides on a server. We are shipping imaging systems now, but the microaneurysm software module is awaiting US Food and Drug Administration (FDA) approval."
There are two versions of the imaging system; one system uses a Canon EOS D2000 single-chip digital camera that outputs a 1728 x 1152-pixel TIFF image without compression. When the camera button is pressed, the image is captured in the buffer space of the camera back. As soon as a file is generated, image data are pulled down a IEEE-1394 (FireWire) connection to the computer. Signal interfacing is automated to ensure that correct information is downloaded and continuous images are taken.
The other imaging system uses a JVC Company of America (Wayne, NJ) KYF-70 three-chip color camera with 1024 x 1360-pixel resolution. It is synchronized to an 8-bit-per-color Matrox Electronics Systems Ltd. (Dorval, Que., Canada) Meteor-II RGB digital frame store.
The system server with a 36-Gbyte/disk redundant array of independent disks can collect images from up to 250 capture stations via a BaseT-100 Ethernet (a Gigabyte Ethernet connection is under consideration). Analysis software runs on the server to identify aneurysms in selected, or all, images as specified by the clinician. The software reports the number and location of microaneurysms and general measurements of the retina. These data allow comparisons with previous test results to be made.
Results are entered into the database so that images can be recalled from the server based on a number of different parameters, such as patient and doctor names or disease (see Fig. 3). The software uses detection algorithms that were developed at the University of Aberdeen (Scotland) under British government funding. Originally running on Sun Microsystems (Mountain View, CA) workstations, the software was ported to the PC level by Digital Healthcare.
Says Skews, "The computer provides the clinician with more information. It saves laboratory time by choosing not to screen some images; it can screen out healthy patients." Security and integrity of the images and related patient data are important health-care requirements. Digital Healthcare participates in the HL7 working group for hospital information systems, and the system server is DICOM (Digital Imaging and Communications in Medicine) compliant.
"Images are watermarked and encoded at the moment they are taken," says Skews. "There has to be an audit trail, so images are never thrown away. We are waiting for FDA approval of the DICOM module and also the automated screening software. We probably will need clinical trials in the United States. Everything else is approved, and all is commercially available in Europe," he adds.
A portable imaging system makes eye care more accessible to children and handicapped and bedridden patients. Doctors and clinical workers prefer a portable system to obtain retinal images from infants and from patients who do not have easy access to medical facilities. Images are stored for later evaluation.
Digital Healthcare designed the OptiTot digital camera system for pediatric use. It can acquire and store pictures of the retina that might show shaken-baby syndrome damage, as well as diabetic disease. "It was not previously possible to take the system technology to babies," says Skews. "OptiTot is a system for acquiring and storing thousands of pictures of infants, wherever they are located. It consists of a fundus camera, which is supplied by Nidek Co. Ltd. (Aichi, Japan), an off-the-shelf notebook computer, and Digital Healthcare's proprietary software" (see Fig. 4).
"A laptop computer runs a subset of our software," says Skews. "It can be connected to a desktop computer or network and sends the images to a database for review and analysis. Portability is a big operational advantage," says Skews. "The camera instrument is not too heavy to carry, and it takes adequate images. In this approach, we're just looking for images that show major damage to the eye, not microaneurysms."
The medical-imaging movement toward digital technology doesn't mean an end to using film images. But those imaging functions that can be digitally converted will benefit from increased speed and efficiency and may broaden the reach of certain kinds of health care as a result of lower costs.
Says Skews, "There's no substitute for a magnified 35-mm print, but we've raised the resolution for digital retinopathy to a threshold where images are accepted." As for future medical imaging trends, he comments, "We can't afford to put unproven technologies in remote locations. Web-based systems will require further development, and there will be new modules for other eye diseases."
Canon Medical Systems
Irvine, CA 92618
Digital Healthcare Ltd.
Cambridge, England CB4 OGZ
JVC Company of America
Wayne, NJ 07470
Matrox Electronic Systems Ltd.
Dorval, Quebec H9P 2T4, Canada
Nidek Co. Ltd.
Aichi 443 Japan
Palo Alto, CA 94303
University of Aberdeen
Aberdeen AB24 3FX, ScotlandM