This project is the basis for my thesis for my masters degree at UConn. My main goal with the project was initially obtaining the data, which was more challenging than anticipated. Hospital PACS systems are generally designed to store lots of imaging data and make it available for 30 days or so. Then the images go into an archive and can be retrieved when necessary... which generally means one patient at a time. PACS is not designed to retrieve large batches of data, like the 900+ patients we got IRB approval to analyze. The 125 or so patients that I did get will be very useful for future imaging studies though.

I was interested in active appearance models (AAMs) as a method to extract 3D valve models from the CT data because of a project I did in my first semester in a pattern recognition class. I actually used active shape models (ASMs), which are very similar to AAMs, to analyze 2D brain images. It worked fairly well and was able to extract the features from most target images. It wasn't terribly robust because it didn't take into account the texture information like AAMs do. I was hoping to extend the 2D ASMs into 3D AAMs for my thesis.

Transforming the 2D statistical shape model (SSMs) algorithm, of which AAMs and ASMs are subset, to 3D is not complicated. A decision must be made to use polar coordinates, quaternions, or another method of rotation in 3D space, but the general equations and algorithm are very similar. The difficult part is in creating the training data. Creating landmarks on 2D training data might involve picking 25, 50, or more points for each training image. The position of the points is generally the same in each image, so loading a 'master' set of points onto each image and moving the points individually to each landmark is relatively easy.

When trying the 2D approach in 3D it gets a lot more complicated. A simple solution, which has been done in the past is to reslice the volume and apply 2D AAMs to each slice. The drawback to this method is that the 3D objects are of different heights and can occupy different numbers of slices. This can be overcome by normalizing the heights, however the internal structures will appear on different slices. So that method is out.

The method I initially chose was to reslice the CT volumes along the long axis of the aortic valve (the valve were interested in), and establish a set of landmark points that describe the important details of the valve. The points would be defined individually for each of the training datasets and manually moved in a 3D viewer to correct locations. The problem with this is the size of the landmark set. Two hundred points might be necessary to accurately describe the valve shape in 3D, however this is an enormous number of points and would extremely time consuming to click them all by hand.

So now I search for better landmarking methods, and better methods are not limited to using AAMs. I may continue with AAMs, but if I find a simpler feature extraction method, I will probably experiment with it.

This is one of the best gifts I've ever received (back in 2007)... a used public payphone. It's an Intellicall UltraTel, made sometime in the mid 1990s. These older Ultratel payphones are AC powered instead of phone-line powered like most modern payphones. Even the older Western Electric payphones made by Ma Bell were line powered, so this phone is unique in that respect. Unfortunately, when the unit was taken out of service, the power cable was cut, leaving the AC transformer somewhere in the 970 area code.

I had to order a new AC transformer, handset cradle, and battery and eventually got the phone powered up and got a dial tone. Then I had to have Intellicall program it remotely, and now it now makes and receives calls. It's an interesting phone... every time we get a call, the phone's coin mechanism clicks to release any coins if there happened to be any in there. The phone also dials 111 every time we answer a call on it. Unfortunately I had to disconnect the phone line from it because my daughter learned to walk... and she learned to press buttons and take the handset off its cradle. Now all she hears is a tinny female voice saying "Please hang up and try again". As my daughter gets older, I'll be able to connect the phone again and use it as a regular phone. I also think as she gets older, this will be one of the only payphones she'll ever see and use.

For Rachel's 3rd Hanukkah, I decided to make wooden dominoes. I chose Gabon ebony for the dominoes and rosewood for the box to hold them. Working with Tropical hardwoods was quite an experience because they are so much denser than any North American hardwood. Tropical woods hold an edge much better than cherry or maple, and combined with its near black color, made it perfect for small dominoes.

The Olin Center was started back in 2003, with it's first recorded scans in April. The archiving and retrieval method at the time was a simple web interface that had access to only the past 30 days worth of data. The rest of the data was archived to DVD. After a couple years it became tough to find the data you wanted because you'd need to sift through 200 DVDs to find the MRI scans you wanted. It became more challenging if you needed data from 100 subjects. So in 2005 I built 4 servers to house the MRI data and created a simple web interface to allow people to search and download the directly to our analysis servers. The whole system was dubbed "All Data Online" or the adoserver for short.

In 2007 I rebuilt the system, distributing the data between 2 servers. At the time, there was approximately 6.5TB of data stored on the servers, all instantly searchable and downloadable. Earlier this year, I needed to rebuild the system and placed the entire system on a single server... The trend from 4 to 2 to 1 server is because of the tremendous drop in the price of disk space. It's just more affordable to have 14TB of space on a single server than 2 servers with 7TB. It's also easier to maintain. Since the amount of data transfer on and off the server is about 8GB/day, there's no bottleneck in keeping it in one place.

Now, at the end of 2009 after 7 years of collecting MRI data, our data is archived in triplicate on 1200 DVDs, we have more than 7500 MRI studies, and 170,000 series of MRI data, stored in 4 different formats. In total there are 12TB of data in approximately 68 million files, and the system has used 160 CPU-days to process and archive the data. There are also new reporting, auditing, and trend monitoring tools. Searches are faster and more comprehensive, with thumbnail image results. There have been approximately 40,000 data requests since the system was created, though the counter was reset with each new system.