Dr. Lior Weizman, a post-doctoral student in the laboratory of Prof. Yonina Elder from the Faculty of Electrical Engineering at the Technion, is developing fast, quiet and accurate MRI technology
Taking a still photo - a task we casually perform today with the smartphone - was a difficult and oppressive challenge about 150 years ago. First, the client was required to make an appointment at a special photography studio, and when his turn came, he was forced to sit motionless for a long time, with the help of head restraints, in a room surrounded by smoke - the smoke that was spread by clumsy and primitive flashes. At the end of the long and expensive process, a black-and-white image awaited him.
The MRI scans accepted today are reminiscent of the old-fashioned photography experience: we have to make an appointment - sometimes months in advance - and then lie motionless for at least 50 minutes inside a narrow device that is not recommended for those suffering from claustrophobia. The patient, who during all this time suffers from jarring noises that are able to penetrate through headphones and earplugs, is warned that any movement he makes may damage the quality of the image.
Will there come a day when an MRI scan will resemble, from the patient's point of view, a still photo using a smartphone? There is no doubt that this day is still far away, but recently the road to it has been shortened thanks to a joint research conducted with the funding of the Eshkol scholarship (Ministry of Science) and under the auspices of the Centers of Excellence (i-core) program of the OT and the Israeli government. Prof. Yonina Elder and Dr. Lior Weizman from the Technion, Dr. Dafna Ben Bashat from the Ichilov Hospital and Dr. Assaf Tal from the Weizman Institute participated in the study.
The premise of the study is that a large part of the information collected in the existing MRI scans is not necessary for the accurate diagnosis. This information includes cross-sections (XNUMXD images) grouped in sequences. Each sequence is made up of many sections, which make it a complete "volumetric image" of the entire organ. The existing method, in which all sequences are fully scanned, consumes a very long scanning time.
The approach examined in the joint research is based on the utilization of the similarity between different sections and between different sequences. Since there is a lot of similarity between close sections and successive sequences, it is possible to sample only part of the total information and thus shorten the scanning time without losing significant information. Furthermore, this process significantly improves the quality of the image - a change that greatly helps the medical team.
In the study, another way to shorten the scan time was examined: using scans performed on the patient in the past. Using an algorithm developed by the research group, the system checks the similarity between the current scan and previous scans, and based on this test allows for a great saving in the information that must be purchased for the current scan.
![LIOR_MRI[1]](https://www.hayadan.org.il/images/content3/2015/10/LIOR_MRI1-500x209.png)
The use of the new methods led to a dramatic reduction (of up to 85%) in the amount of information that must be collected in an MRI scan. This means shortening the duration of the scan by 85%. Furthermore, in another study, the group was able to reduce the noise generated inside the MRI machine by about 18% (18 decibels).
"Accelerating and quieting MRI scans has occupied many researchers for a long time," explains Dr. Weizman, "however, most solutions require expensive dedicated hardware, and also do not allow shortening the scan time without losing a lot of essential information. Our idea - utilizing similarity between sequence to sequence and between section to section, and using previous scans - makes it possible to dramatically reduce the scanning time while improving its quality. The results of the study were recently published in the prestigious journal Medical Physics, and now we hope that medical centers and manufacturers of MRI devices will adopt its results and thereby improve the MRI scan and the 'user experience'."
3 תגובות
The signal is random or sampled in the sense that the system that samples it does not know what it will look like before sampling.
In addition, algorithms for sharpening the display that have long existed in smartphones and space telescopes sharpen the image much more than would be possible without them and allow objects to be viewed with sharpness that is not possible otherwise. Therefore, even if Yona's XSAMPLING is not used, but in an evaluation based on nearby pixels, there is no fear of information loss, but it is received sharply through a power multiplier. the algorithm.
Professor Yunina Elder developed what is known as XSAMPLING. The theory allows sampling at frequency X, when in practice it samples a much higher frequency. Until her if you sampled information at X frequency, you actually got X/2.
Therefore, it does not lose information, but simply scatters the samples, perhaps at a non-uniform time distance, but as needed.
If this great theory of hers (she is a member of the Israeli Young Academy of Sciences - a super rare privilege) is now being applied to such a practical application - I do not know. For this you need to read the scientific articles. If this is indeed the case, the information will not be lost, it will be identified and sampled more densely in time as needed.
We are used to thinking in absolute terms and in these terms many things seem impossible such as for example - fear of losing information. But in terms of a random signal, statistics always wins and otherwise lost signals can be sampled.
I'm not a fan of Professor Elder. There are non-scientific aspects of her work that I dislike. As a scientist she is determined and innovative.
So slide the pictures. Beautiful. But what if there is a small tumor that is only visible in one section? Smoothed and deleted. Wait for the first claim and then no doctor will dare to use the slide.
There are enough MRI machines in the western world for all needs. I read that even in Israel the problem was probably only financial (who pays whom) and has already been resolved.
But I have a suggestion to increase the utilization of the devices to close to 100%: from my experience, the preparation time is close to the time of the test itself. If beds are built on rails on both sides of the device, one patient can be prepared while another is examined. You need a dividing wall and screens that will hide visibility through the device. Simple and cheap.
The problem of the decoding bottleneck remains. If the education system continues to collapse, there will be no deciphering doctors and no operating technicians. So bring Chinese or Indians. Basically, there won't be any doctors ordering tests either. So no problem.