Jean-Charles Bazin
CC500-GroupA – “Seeing inside the body from the outside”: this is exactly the main goal of medical imaging. X-ray pictures, MRI, endoscopy and so on are some equipment examples that permit to perform such an unbelievable task. This paper deals with the development of medical imaging systems in diverse points of view. We will first introduce some future imaging sensors and then show some image processing examples. Finally some vision-based medical robotic systems will be presented.
Imaging sensors
Nowadays, there exist three main medical imaging systems: X-ray, MRI and endoscopy. This section presents their capabilities, limitations and future versions.
X-ray
The most common imaging tool is based on x-ray technology and permits to obtain some images similar to figure 1. For example, a chest x-ray makes images of the heart, lungs, blood vessels and the bones of the chest. Therefore it is very useful to diagnose pneumonia, heart problems or other related medical conditions. However, it suffers from some limitations such as radiation exposure and image resolution. First, radiography involves exposing a part of the body to a small dose of radiation to produce pictures of the inside of the body. Thus there exists a small probability that a patient might develop a cancer from the radiation received during the procedure. Obviously, future X-ray systems should be effective with a minimum radiation dose, for example equivalent to the average dose received in daily life. The second limitation is due to low resolution: the radiologist might not detect some important deformations or even very small cancers. Current X-ray technology does not permit to increase image quality a lot. Thus two approaches are possible: either enhance the X-ray picture with image processing algorithms (cf chapter 2) or switch to another imaging system such as MRI.
Figure 1 – typical X-ray images of a skull (left) and a hand (right)
Endoscopy
Endoscopy does not correspond to a see-through system like X-ray but rather a minimally invasive equipment. Concretely, an endoscope is composed of a flexible tube in which a light and a camera are inserted. Recent researches have led to the development of both miniature and high quality cameras so that images obtained by endoscopy provide a very good resolution. Moreover, it permits to obtain color images, contrary to other imaging sensors such as X-ray or MRI. The drawback is that the procedure is far from being comfortable for patients. That is why researchers are trying to manufacture micro cameras whose size does not exceed 2 centimeters and equipped with wireless communication transmitter so that the patient can simply eat the camera.
Figure 2 – Left: an endoscope. Right: image acquired during an endoscopy
MRI
The acronym MRI stands for Magnetic Resonance Imaging and refers to a very powerful medical imaging system. Its key technology is based on magnetic field and thanks to advanced signal analysis, can retrieve the object consistency that the magnetic field goes through. The system is depicted by figure 3 and some MRI examples are sown in figure 4. As you can notice, MRI images consist of a series of slices. Then it is possible to “stick” these images so that the body can be reconstructed in three dimensions. MRI provides a much better accuracy than X-ray pictures but also suffer from some limitations. First, such a system is very expensive, about US$ 1 million. The second limitation does not deal with MRI system itself but the image processing algorithms that stick the images. Finally, the procedure requires that the patient remains immobile during the whole acquisition, typically 10-20 minutes. Future MRI systems will have to face these three important difficulties.
Figure 3 – Procedure of the MRI sensor
Figure 4 – Sequence of images obtained by MRI
Image processing
Analyzing medical images is not an easy task, even for a specialist. That is why some research works have focused on automatically analyzing these images. For examples, some methods permit to detect bone fractures or even breast cancers. Some other researchers have concentrated efforts to obtain accurate 3D reconstruction of some body parts such as chest or brain. The more accurate the 3D reconstruction is, the better diagnosis can be made. Figures 5 and 6 present some results of such algorithms.
Figure 5 – Automatic 3D reconstruction of a brain (right) from a set of X-ray pictures (left) and MRI images (middle)
Figure 6 – Advanced signal analysis methods permit to precisely retrieve the consistency of some body parts. Thus it is possible to precisely display the tissue, bones and blood vessels for example.
Vision-based robotic systems
Recently, the field of machine vision has reached such a high robustness level that it can be applied to medical applications. One of the most impressive results is based on the so-called “augmented reality” technology which adds virtual objects and information in real videos. Figure 7 shows an example where an instrument can be precisely pointed towards a particular area of the brain, without even having to cut the patient’s hair. Figure 8 depicts a critical step during a surgery for Parkinson disease: an electrode has to be accurately positioned in the brain and emit a high current to irradiate some neurons. For such a difficult case, a complex system combining augmented reality, laser and robotic has been developed: the augmented reality part permits to visualize the brain areas and blood vessels, the laser to precisely map the surface of the brain and finally the robot to carefully insert the electrode in the brain without any shaking usually inherent to any human surgeons.
Figure 7 – An augmented reality system to visualize the brain “seeing through” hair and head skin.
Figure 8 – A complex system combining augmented reality, laser and robotic for accurately inserting an electrode into the patient’s brain
Conclusion
This paper has dealt with some aspects of future health systems with respect to medical imaging. First we have presented the three most common imaging systems and explain their capabilities, limitations and future versions. Then some examples of image processing algorithms have been introduced. Finally, some vision-based robotic systems have been shown. It is worthwhile to note that research is very active in medical field nowadays, so we can expect several impressive equipments in very near future.
References
- J.-F. Mangin, O. Coulon, and V. Frouin, “Robust brain segmentation using histogram scale-space analysis and mathematical morphology”, in 1st MICCAI, MIT, Boston, USA pages 1230--1241, LNCS-1496, Springer Verlag, 1998
- Andreas H. König and Eduard Gröller, “3D Medical Visualization: Breaking the Limits of Diagnostics and Treatment” , ERCIM News No.44 - January 2001
- http://en.wikipedia.org/wiki/MRI
