COMMERCIAL PROFILE - DUBLIN INSTITUTE OF TECHNOLOGY:WITH SKIN cancer on the rise and Irish people in their 60s and 70s now five times more likely to develop it than their parents the need for quick, safe and reliable diagnostic tools for this fatal condition has never been greater. Now, thanks to the work of the TeaPOT (engineering advancing people oriented technology) research group in the School of Electrical Engineering Systems in Dublin Institute of Technology (DIT), people with concerns in relation to a mole can have the condition checked out quickly and cheaply using a new online service known as Moletest (moletestuk.com).
The Science Foundation Ireland-funded TeaPOT group was established in May 2007 to consolidate ongoing research into technology that interacts with humans or with the human body. The group’s activities include research and teaching in biomedical signal engineering, human-machine interfaces, assistive technology, rehabilitation engineering and health informatics.
The Moletest service is based on a new approach to analysing medical images developed by Stokes Professor at DIT Jonathan Blackledge and Dr Dmitri Dubovitski from the Bauman Moscow State Technical Institute. The system allows users to upload digital images of their suspect moles to the website. It can then identify any anomalies in these photographs and advise users to visit a medical practitioner if there is a chance of a mole becoming cancerous.
The service uses an easy-to-understand traffic light approach to screening for non-melanoma and melanoma skin cancer. Green denotes a “normal” lesion, amber “borderline”, and red a cancerous one.
All customers need to do is upload a 5MP image or better of the suspect mole to Moletest’s website – an image of this quality can even be taken on some mobile phones – pay a fee of roughly £40 (€48) and wait for their results, which they will normally get within 24 hours.
The system – which is supervised and audited by a panel of advisory dermatologists – evaluates the customer’s image against a database of known results to see if there are any characteristics consistent with previous cases of cancer.
The system continues to “learn” by comparing its findings with later clinical diagnoses of dermatologists following biopsies and other examinations, using these comparisons to inform future analyses.
The Moletest product development team is led by Prof Blackledge and Prof Rino Cerio, the UK’s only professor of dermatopathology.
“The incidence of malignant melanoma has quadrupled over the last 30 years due to the advent of cheap air travel and patients failing to get moles checked until it is far too late,” says Prof Cerio. “Although a rare form of cancer, melanoma accounts for over 75 per cent of skin cancer deaths – most of which could have been avoided with early detection. This innovative technology has the potential to vastly increase early detection, saving thousands of lives that would have otherwise been lost to this very treatable skin cancer.
“Many people, especially men, don’t visit their GP to have a suspect lesion examined until it is too late. Using Moletest consumers will be able to see if there is a need to seek medical advice in under 24 hours. However, consumers must understand that moles develop throughout their lives and monitoring is important – especially for those in high-risk groups.”
The technology used in the system is based on Prof Blackledge’s long-time interest in an area of mathematics known as fractal geometry. “In traditional Euclidian geometry we have the three X, Y and Z dimensions – and Einstein gave us one more in time,” he explains. “Fractal geometry is all about the dimensions in between these. This has applications in areas like computer vision, cryptography and financial analysis.”
In this case it is being applied in the area of computer vision. Euclidian geometry is perfectly adequate for analysing images of objects in the normal physical world, but has limitations in the area of medical imaging.
“If you look at a medical image it is usually a mess and doesn’t have definable lines to it. What fractal geometry allows us to do is analyse its textural properties as well as its shape.”
This ability to measure texture lies at the heart of the Moletest system. When users upload good quality colour digital images the specific area of interest is identified automatically using a unique object location algorithm. Various features are then identified and measures obtained using both conventional Euclidean and fractal geometric parameters. A combination of these parameters is used to generate what is known as a “feature vector”, which is then compared with historically equivalent cases where the medical outcome is known.
This is where the system gets really smart. It doesn’t simply rely on a database of images to decide if a mole may be cancerous or not. It measures more than 40 different parameters in respect of each image, correlates these to its database of dermatologists’ diagnoses and assesses which parameters are common. It then uses what Blackledge describes as a “fuzzy inference engine” to come up with the final result.
The fuzzy inference engine operates in the same way as the human mind, Blackledge explains. “If it looks bad it probably is bad. At the end of the day the only sure test is a biopsy where you take material out and examine it under a microscope and even then errors can be made. Using Moletest to analyse the image of the mole you will get accuracy consistent with a dermatologist looking at it.”
The technology has now been commercialised. An exclusive license has been awarded by DIT to Moletest Limited which was launched in the last fortnight with capital investment from the UK-based World Wide Leisure Group.
This may be just the beginning for Professor Blackledge and the TeaPOT group. While a licence has been awarded for this particular application of the technology, the intellectual property rights remain with DIT. It has many other potential applications, including the analysis of medical images created by systems such as MRI, CT and ultrasound scanners. Indeed, it was with such images in mind that the technology was initially developed – its first commercial application was driven in part by practicality.
“I have worked in these areas and what we are trying to achieve is to get a computer to do what a radiologist can do when reading the images,” says Blackledge. “But if we wanted to use this software with an MRI or CT scanner this would mean it was restricted to places with millions of euro worth of equipment in place. Moletest, on the other hand, works with almost any type of image capture device making it very accessible. It’s the only area that is so user friendly and that’s what led us to the development of this system.”
He is already considering further applications. “One area we are looking at is to use the same approach for the next stage in the diagnostic process,” he explains. “The same technology can be used to assist radiologists in the analysis of biopsy material, for example. It can also be used to assist in the examination of cervical smear tests. It has great potential to assist radiologists. In the future we may see it used in areas such as MRI and CT scans as well.”
