Biomedical Engineering
Cancer Detection Using Atypical Cell Detection
Early detection of cancer is extremely important for successful treatment.
In
microscopic examinations, the pathologist has to spend hours examining the slides.
This is tiring for the pathologist, and time wasting. Partially automating the task
of atypical cell detection, while relying on the pathologist to make the final decision,
in this context, is a real need. Our atypical cell detector identifies cancerous
cells by examining the images of a pap smear (on a slide under an automated microscope).
This will automate the tedious manual examination process which is mandatory in
some countries. The most valuable time of the pathologist, who usually is a very
senior doctor, is saved tremendously by this system. Diagnosis assistance using
an expert system in conjunction with our atypical cell detector is the future direction.
Painless Glucometry
The percentage of diabetes patients is alarmingly high all around the world.
Most
of these patients have to prick their own fingers and monitor the glucose level.
Some patients need the reading daily or even more frequently. Painful pricking and
obtaining a blood sample, therefore, is not the solution at all. The device that
we plan to build makes the glucose monitoring non-invasive! We do this by using
an antenna kept just touching the skin which gives an output related to the biological
parameters. This does not need a blood sample and the monitoring process is painless.
Electronic Systems
Traffic Light Attachment to for the Visually Impaired
There is a renewed interest in Sri Lanka to make the government offices and facilities
accessible for the disabled. This is very welcome as we need to treat our disabled
war heroes with the best possible way. For example, a large number of offices are
not wheel-chair accessible. Public transport is not at all ready to serve the disabled.
Let alone public transport. If a visually impaired person chooses to walk, even
crossing the road is a big challenge. Most of our traffic lights are not visually-impaired
friendly. Ironically, the traffic light system enabling patients to cross the road
to reach the Eye Hospital, Kandy, a major eye hospital in our country, is not visually-impaired
friendly!
Unfortunately, modifying the existing traffic lights to emit audio signals
to help the visually impaired is prohibitively expensive. Given the fact that most
of the major junctions in our country are unprotected, we can perceive the cost
of replacing the existing traffic lights with visual-impaired-friendly traffic lights.
What is feasible, as a result, is to devise an attachment to be mounted on the existing
traffic lights to give audio signals in response to changes in traffic signals.
However, such a device should not trample the existing traffic-light circuitry.
Such an attempt will involve reconstructing and reprogramming the traffic-light
circuit. What is appropriate, therefore, is a minimally intrusive attachment.
Computer Vision
Room Recognition for Indoor Surveillance
Unmanned outdoor driving, automatic location recognition,
and real-time vehicle and pedestrian detection are few amazing
achievements recently showcased in computer vision conferences. These techniques
work well in outdoors. Their counterpart, indoor recognition tools, can immensely
assist in security, surveillance, and blind-assistance applications. With such applications
in mind, we have developed a computer vision based system that is able to do indoor
surveillance using room recognition. system can be mounted on a robotic plat form
like a vacuum-cleaning robot and sent on its own to investigate the indoors. There
is no detection capability such as detecting intruders, anomalies yet. However,
this ability could also be incorporated.
Scene Recognition for Assisting the Visually Impaired Persons
Visually impaired people rely on their cane, the guide dog,
and
Braille to navigate and identify places such as elevators. Computer vision can help
immensely in this regard, by building devices for surrogate sight. We propose a
wearable camera-based system which would observe the scene as the person walks,
and produce audio cues such as "stop sign", "bend", and "play ground". We have been
successful in adapting and experimenting with systems for read-sing detections,
and scene recognition. We plan to build the wearable system with these capabilities.