When we think about medical devices, one automatic relates them to words like "precision", "shiny", and "high tech". Although it is always desirable to have 3D printed hardwares, maticulously  engineered novel materials, and preferably sexy industrial designs like those you may have seen in 'Resident Evil', the toys in your kids' playroom may just be as useful as those expensive devices. Surprise! In the recent years, the Lego bricks have become the heroes of many applications in medical imaging and surgical technologies. 

Several features of the Lego bricks contribute to their success in the medical fields. First of all, the manufacture of each block is quite accurate and consistent,which really helps with the construction of any device that has a high requirement on precision. For those who are interested in how this is achieved, here is a video on the secrete of Lego making.

 
Ok..Let's continue.
The second major feature of Lego is the modular nature of the toy and diverse pieces, which offer great flexibility in making devices in different shapes and dimensions. Even better, the Lego robot series provide electronic components and the ability to program different actions, enabling the researchers to add intelligent kinetic inputs in scenarios like simulating organic movements without the cost of complicated machinery. Third, the material of Lego bricks are great for medical imaging applications, such as ultrasound, MRI and CT. Lastly, there are many freely available software, including the official 'Lego Digital Designer' to help the design process.

Now let's take a look at a couple of the medical applications of the Lego bricks.

Image distortion correction
MRI can be subject to image distortion just like your regular cameras. It can greatly affect the measurements of brain volumes. Such distortions may come from different sources, but in short, they are the results of 'not-nicely-behaving' magnetic fields, which may largely be characteristic for each individual MRI scanner. By mapping these distortions using a phantom, we can recover the 'true' images. One solution applies one such phantom made of a bucket of Lego bricks. The Lego phantom is scanned inside an MRI scanner, and a correction field is produced by registering the distorted image to the digital model of the phantom. As a result, any upcoming scans are corrected with this pre-calculated correction field. The images below are taken from the article "Improved Precision in the Measurement of Longitudinal Global and Regional Volumetric Changes via a Novel MRI Gradient Distortion Characterization and Correction Techniques" by Fonov et al. (2010). The image on the left demonstrates the 3D distortion field while the one on the right shows the alignment of the distorted scan (in red) with the digital model (in green).
 ​
​

Ultrasound calibration​
In some applications, such as brain surgeries, ultrasound images are spatially tracked to provide real-time updates of tissue deformation. As a result, there is a need to establish the spatial transformation between the 2D ultrasound image and the motion tracking device attached to the transducer so that the pixels in the 2D image can be accurately mapped to the 3D world. This procedure is called ultrasound calibration. Historically, such a calibration procedure has always been achieved through the use of a phantom. To obtain the most accurate construction, there have been a long list of different phantom designs and mathematical formulations. These calibration phantoms can be terribly tricky to manufacture, and may be quite expensive as well. Alternatively, phantoms made of Lego bricks can be used.

One solution was proposed by the Perk Lab at Queen's University. The lego bricks provide a frame for the fishing wires to form a Z-shape configuration. By matching the features of the wires shown in the US scans (typically as bright dots) and the known configurations, one can compute the calibration matrix. 

I have also proposed a solution that applies a phantom made of Lego bricks. However, instead of matching features of fishing wires, scans of the phantom are directly registered to its digital model. As a result, there is no need to explicitly (and typically with manual labor...) identify the fishing wires in the US images!

There are many more applications of Lego bricks in medicine (Just type in  'LEGO' in pubmed, you will see). Although it is often regarded as merely a toy in the playroom, it can definitely show more potentials, and it really depends on how imaginative we are!

I have previously posted a blog about human brain atlases. 
To give an access to a broader audience, my colleague Ian Gerard and I have created a video of the brief history of mapping the human brain anatomy.

check it out!


Some updates:

This video was entered for the 2014 MICCAI Educational Challenge, and we won the 1st prize in the expert panel awards category! If you happen to be in the MICCAI community or just medical/surgical image processing field, I do encourage you to participate the future competitions. 

If you happen to have to deal with medical device testing and validation, it may be difficult to escape from either getting the real body tissues from animal or human (yes, tissues from our own species) or making a fake one that resembles the real deal.

For the purpose of brain surgery, I have bought real animal brains in the past, and the experience really was not as nice as I thought. Hunting for the freshest possible brains which preserve the best mechanical and imaging properties in the city of Montreal was quite a journey, and often it won't serve a long term for investigation. In rare occasions, I have heard at some more fancy courses/demos, real human heads were used for training, but how rare will you get into this type of situation? 
​
Another convenient option will be making a fake brain. If you search the relevant key words in Youtube, it is likely that you will get many tutorials on how to make fake jello brains for special occasions. Surprisingly, for disposable brain phantoms, these tutorials will very likely to work for the scientific needs. However, if you intend to keep the brain for longer time and better simulate the mechanical and medical imaging properties, we need to do some more sophisticated cooking. I happened to need to make a brain phantom this week, and let me show you how to cook half a realistic brain for science!

In general, I followed the recipe introduced in this paper [1], and the associated medical images (MRI, CT and Ultrasound) can be found in here. Instead of getting the brain mold from the dollar store, I did get a fancier version  - a 3D printed mold made from one of the most famous brains in the field, the Colin27 template. Yet, the dollar store mold works fine if you don't have access to brain data or a 3D printer. If you ask anyone in the brain imaging field, nearly everyone has heard about Colin27. I suppose now he is a bit of a rock star in the field, not for the music, but for his brain (check out his interview here). The main material for the brain is called PVA-C or polyvinyl alcohol cryolgel, which is commonly used in making body tissue phantoms. When the gel is frozen and thawed (each process for 12 hours), it will be transformed into a material that feels much like a piece of meat (minus the sticky part and smell). Depending on the concentration and the freezing-thawing cycles, the mechanical properties can be manipulated to resemble different body tissues. For our brain phantom, we choose a PVA concentration of 6% with one freezing-thawing cycle. In addition, we will also put some "implants" to resemble different tissue types and to mark the internal location of the phantom in the MRI/US images, and for the implants, we chose a PVA concentration of 8% with two freezing-thawing cycles. To obtain the US contrast, talcum powder (or simply baby powder..ask your friends with babies) was used. For the implants,  5% concentration is used, and 1% was applied for the main brain phantom. If you also need to have some CT contrast, certain concentration of BaSO4 is required. Although CuSO4 may be required for MRI contrast, often the MRI scanner will be able to differentiate PVA with different concentrations. These will really depend on your specific needs.  

OK. Now let's make the implants first. You may be creative on this. I personally made some spheres, stars and hearts. For the spheres, I used the marbles as the models, and made the mold with the mold putty, which is a really nice material for fast (only 5 min for hardening) and easy casting work, and lasts for a long time for re-uses. Otherwise, using silicone is the standard choice...

After the implants are ready, make the gel with the appropriate PVA concentration for the main brain. Put the implants inside and fill the brain mold with the gel. To save your time and material, it is suggested to fill the mold with water first, and measure the volume of the water to know how much PVA you will need to make the phantom. However, if you don't want to wet your mold, some couscous will work the same. 

The most difficulty I had with the cooking is really dissolving the PVA crystals into water. To melt the crystals while not burning them into some yellow colour, you should prepare your water to be between 70 ~90 degrees Celsius, and keep stirring. In this sense,  a nice water bath and a thermometer will be very instrumental. Also, watch out for the loss of H2O through the cooking process. Too much water loss may mean that your concentration will go up, resulting in a harder material. 

Now let's take a look what the brain will look like in the ultrasound image.

We can clearly identify the implants in the brain phantom when imaged with ultrasound, as well as the cortex. In a real brain, there are many different structures, and their imaging properties vary as well. If you need to make even more realistic brain, more complex casting techniques are the only limit you will face in production. However, if you are interested, you are more than welcome to give it a try.

Reference:
[1] Chen SJ, Hellier P. Marchal M., Gaufrait JY, Carpenteir R, Morandi X, and Collins DL, "An anthropomorphic polyvinyl alcohol brain phantom based on Colin27 for use in multimodal imaging," Medical Physics, 39(1):554-61, 2012.

For the longest time, our ancestors all over the world have always believed that the central control unit of our bodies is the beating heart. I suppose whenever there is crucial emotional roller-coaster, heart rate is often the first thing we notice despite the poor brain is working seriously hard to regulate the entire body to adjust to sudden changes. Well..the traditional definition of death using pulse and heart beat may have also played a major part in the old belief.

When we started to realize that the lump of meat sitting in our skull is the real control centre for all our daily activities, many attempts have been made in the past to provide a comprehensive map for navigating this vital structure of ours. One of the earliest, and personally, the most intriguing one, is from the "science" of Phrenology developed by the German physician Franz Joseph Gall in 1796. It is a method of measuring the skulls in relation to the brain regions underneath the skull territory that are responsible for specific functions or characters of the human being. Although today, we might more or less put the sculpture of Phrenology head (yes, I have one in my living room as well) in the oddity section together perhaps with vodoo dolls and electrical therapy chairs, it did provide some preliminary directions on the separation of brain regions and functions. In the modern world, the general divisions of the brain lobes are also named after the skull bones they are under. However, deep under the cortex, there is a whole lot of world going on carrying relay signals between different sub-regions, which will require much more in-depth investigation. 

For most people who have studied the human brain in greater details besides the general brain structure divisions, the more elaborate and detailed brain maps, such as the Talairach atlas and/or Schaltenbrand atlas would probably ring a bell. These brain maps are based on the more advanced tissue staining technology and arduous sketches of the European physicians. The general idea here is that the brain is cut into thin slices, and each slice is chemically stained, and outlined for each shown structure. Later on, a few variants of these two atlases were published as well, but the general names are still kept the same. For both of them, the establishment of the coordinate system for the brain are based on two important brain structures, the anterior (AC) and posterior commissures (PC), which, during the time before the invention of MRI, were important landmarks in CT images for the early generation of neuroscientists and surgeons to navigate the brain. For the historical reasons, the habit of using these two structures to define brain map coordinate remains until today even more advanced imaging and computer tools have become increasingly available. With the two landmarks and according to the rule of symmetry, the brain is then investigated under a grid system, or often referred to as "stereotactic space". For many deep brain structures, the idea (and the wishful thinking) is that for all human brains, without slicing open the living brain, the location of the structure is a fixed or scale-up distance in 3D walking from the landmarks. This will help tremendously in the operations as we know more and more about the "functional surgeries". Despite the similar attempts to define the coordinate system in both atlases, one should always be warned that the central lines connecting the two commissure points, which define the anterior-posterior direction in the "stereotactic space",  are not exactly the same. For the Talairach atlas, the line connects the superior tip of the anterior commissure and the inferior tip of the posterior commissure, while for the Schaltendbran atlas, the line connects the centroids of these two structures. Although the differences can be rather minute, they change quite much how the brain is oriented in the drawings, and thus the way you navigate yourself in the brain when using these brain maps. I have seen quite a few not-so-careful physicians mixing them up. Also, the brain donor for Talairach atlas is an old French lady. Since one brain cannot be sliced in different directions twice, some sketches were estimated from the slicing directions. However, for Schaltenbrand atlas, there are 3 different brain donors for each direction of slicing. As a result, the measurement may not be very consistent across different slicing directions. 

Unfortunately, as you may have guessed, everyone's brain is different, and the generalization that everyone's brain is highly similar and can be navigated by the grid system from these two atlases may not be ideal, particularly when different diseases and aging process can impact the shape of the brain structures. And very importantly...it is still just a piece of drawing! How can this adapt to the new 3D MRI?

While you can now buy a 3D digitized ensemble of Talairach and Schaltenbrand atlases (ask your nearby medical school's library) for your professional use or purely out of curiosity, there are many other versions of more modern digitized 3D brain maps. They are often color-coded and stored as formats that can be directly morphed to match the individual brain anatomy.  With the better and stronger MRI scanners, these atlases will often accompany their MRI templates. The most famous ones in the brain imaging field are the MNI305 atlas (http://www.bic.mni.mcgill.ca/ServicesAtlases/MNI305) and ICBM152 atlas (http://www.bic.mni.mcgill.ca/ServicesAtlases/ICBM152NLin2009). The early versions of these atlases were manually aligned to fit the grid system of the Talairach atlas with AC-PC line aligned. The later versions applied more sophisticated image processing techniques to make multiple brain MRIs into one that has the averaged shape of a large population. I have also made my humble share of brain maps for the Parkinson's disease population. If you interested, you may check them out here. 

One latest example of the brain map is the project called "Big Brain" developed by the team of Dr. Alan Evans at Montreal Neurological Institute and their collaborators. It was a hyped version of the previous works with more advanced technology. You may zoom-in for incredible details of the brain till the cellular level. The registration is free for use. You can give it a try here. It is mostly for educational purposes so far, application in clinical use is still limited.

From almost superstitious set-up to the cellular view of the human brain map, it was quite a big journey, right? As we add more and more technology and discoveries into the map, the information will only become richer and richer. What would you expect the next map to be like?

Great idea for the medical students, but not sure if the lack of anatomical variability and texture simulation will do the justice.

Deep brain stimulation (DBS) has been an alternative therapy for treating various neuro-degenerative diseases, such as Parkinson's disease. Usually the electrode shall be inserted at a specific site (i.e. subthalmic nucleus or STN for the case of Parkinson's disease), which is often surrounded by many delicate and vital structures to the healthy function of our brain. The accuracy of the electrode placement is crucial to the surgical outcome, and the understanding of the circuitry of the human brain. However, there is always a question regarding whether the stimulation electrode has been placed at the desired site inside the brain.  Some believe that in some cases migrated or misplaced electrode can still result in positive surgical outcome, and this may raise the question of whether the conventional stimulation sites are the unique solution to the problem. 

I ran across an interesting  demonstration of the DBS procedure made by Dr. Jeremie Dequidt from INRIA (http://dequidt.plil.net/publications.php). The accompanied article will soon appear in the proceeding of MICCAI 2011.

Although the simulation may disclose some possible scenario during the DBS procedure, validation of the accuracy of the model should be further verified. However, the implementation of the validation can be highly complicated.

So....besides the old controller...how do we play the Mario game?