The doctor will see you now
Machine vision is being used for many varied applications within the wide borders of the medical field – from assisting lab researchers to search for novel protein targets for drug discovery, right through to treating individuals with spinal problems. Machine vision is also being used for more traditional applications within the field, namely quality assurance and quality control of drugs labels.
Back pain, whether chronic or acute, is a silent epidemic of modern life. Indeed, recent surveys have suggested that half the population of Germany over the age of 30 experiences some kind of back problem. In turn, the loss of mobility associated with back pain leads to many thousands of working days being lost every year, with significant economic impact. Back problems can be caused by – or lead to – postural deformity, tilted pelvis, scoliosis or kyphosis, while diagnosis of the exact condition is notoriously difficult, making good treatment elusive.
Diers International has developed a new method of examining a patient’s back, using a camera to take a stereographic image of the spine in three dimensions. A grid is projected on to the patient’s back with a light beam, and scanned with a uEye camera from IDS Imaging Development Systems. During the examination, the patient stands two metres away from the scanner. The image is acquired in 40ms, which ensures that a good image is taken despite the fact that people, and children in particular, never stand absolutely still. From the image the system calculates the location of anatomical landmarks, such as the seventh cervical vertebra or the ‘pelvic dimples’, and derives the spine posture from this data, to an accuracy of a tenth of a millimetre. Information from this ‘formetric’ scan can then be used to decide on the best course of therapy to give the patient a swift and comfortable recovery.
A complete 3D model of the spine is reproduced from the stereographic image of the back - taken with a uEye camera from IDS.
The system can also be used to examine movement of the spine in motion. The patient walks on a treadmill while the system captures a dynamic image of the spine, giving specialists an extra dimension of diagnosis.
The system was originally conceived nine years ago, as a project to develop a radiation-free method for biomechanical measurement, which was intended to spare children with scoliosis from frequent x-ray exposure — clinical trials had shown that the children had a much higher cancer risk due to the regular x-ray examinations, which were needed every three to six months.
Digital image capturing and recording has also opened up new possibilities in areas where visual inspection forms a key part of diagnosis, such as dentistry. Dentists have image capturing equipment available to them that enables them to keep a visual record of the oral health of all patients, rather than simple, imperfect, written notes. One such device is the PalmScope from Moritex, which has recently been adapted specifically for dentistry. Dentists using this system can probe the patient’s oral cavity and view the images in real time on an LCD screen built into the unit. The same images or video can be stored for assessment and review at a later date.
One of the biggest fields in medical research spending, and indeed of all scientific research spending, is drug discovery. One of the most promising areas for novel drug discovery is in proteomics – the study of proteins. For example, if a drug can be made to affect a certain protein, then that drug may provide a treatment for a disease linked to the actions of that protein. Such research could eventually lead to effective treatments for conditions such as diabetes, arthritis and heart disease. However, for a scientist to identify a potential drug target, he or she must have an accurate way of identifying the protein.
Syngene, based in Cambridge in the UK, has developed a system that uses machine vision to examine 2D protein gels automatically. The system can process up to 150 different gels in a week, each measuring 26cm square, and can differentiate between different dye colours, whether visible or fluorescent, such as Sypro Orange, Sypro Ruby, Sybr Green, Coomassie Blue and Pro-Q Diamond. The system can accurately identify tiny protein samples, where only a few nanograms of the protein might be present, in a gel sheet containing more that 2,500 spot-samples. The system can also be used to analyse Western blots.
One of the most important concerns for managers of medical laboratories is the traceability of samples passing through the lab. An individual lab may be processing many thousands of individual samples at any given time, each of which must be tracked at every stage of the testing process.
Central Labo is a French company that manufactures plastic medical devices. A particular sample tray manufactured by Central Labo contains 48 individual wells for biological and medical samples. Until recently, each tray was identified at each stage in the testing process via a barcode. Unfortunately this system could not differentiate between individual samples on each tray – meaning that, if one sample was defective, all 48 had to be rejected.
Central Labo now uses a system from Technifor for both imprinting and reading data labels on the individual sample wells. The system uses a laser to mark a two-dimensional barcode, or matrix, on to the sample – as well as letters and numbers that can be read by sight. The DataMatrix can then be read by a machine vision system supplied by Cognex. By reading codes at each step the progress of each and every sample can be recorded, giving lab managers better control over – and knowledge of – all stages in the testing process.
Machine vision is not only being used in drug discovery labs, but also for quality control before pharmaceutical products reach the consumer. Simple typographic errors in drugs labelling has very serious ramifications. For example, a missing hyphen turns a label from ‘take 1-2 tablets daily’ to ‘take 1 2 tablets daily’. Responsibility for such grave errors lies ultimately with proofreaders working in a pharmaceutical company’s quality control department, but inspecting many thousands of labels can place a huge strain on proofreaders, eventually leading to such errors slipping through the net.
Parish Automation supplies the pharmaceutical industry with automated proofreading systems, which use a method of image comparison to spot errors in drug labels. The system captures an image of the label, using a Genie camera from Dalsa. This is converted into a negative image by a processor. The system then ‘subtracts’ the negative image from the ideal ‘master’ image stored in its memory. It then highlights any differences and alerts the operator to the fault.
The real challenge, according to Gary Parish, president of Parish Automation, arises from proofreaders’ suspicion of the accuracy of the system: ‘Many proofreaders are under the impression that they are as accurate as – and faster than – electronic inspection. The best way I found to demonstrate the problems that human inspectors face was to ask the proofreaders to take two drawings from a 'spot the difference' puzzle, and compare their performance with that of the computer. The proofreaders usually take several minutes, while the machine vision-based computer can highlight the differences in less than one second.’
Machine vision techniques are helping advance medical research, diagnosis, and treatment. As such systems become more widespread treatments will become better, giving patients better prospects, and better quality of life for everyone.