Week 3 - Aaron LaViolette

As Dr. Prince, my clinical adviser was still traveling until Friday, I spent most of this week with Dr. Auh and his residents in the body reading room. When I was watching Dr. Prince last week, most of the images he was looking at were magnetic resonance (MR) images, but with Dr. Auh, most images were computed tomography (CT) or ultrasonography (US). The first thing I realized is that this must have to do with time, since MR images take longer to acquire. I spoke with one of the residents about this and she said generally CT may only take about 15 minutes to do while MR can take a half hour to an hour and so if a "good enough" image can be gotten with CT, that is generally done. This made sense considering most of the images I saw Dr. Prince reading were for clinical research using MR while the images coming up to the reading room were generally less than an hour old (so in the reading room time was of the essence). I learned from this that the quickest modality that can accurately show the pathology is probably the best option clinically.

I also learned this week how patient age (and overall health) can come into play when the images are read. For example, one patient in their 90s had a nephrectomy, which also resulted in a partial colectomy due to complications in surgery. The patient reported abdominal pain triggering a CT to be ordered (as it was a worry it could be a from a complication of surgery). On the image there was of course the missing organs, but also lots of other pathologies (considering the age and clinical health of the patient didn’t seem too surprising). All these factors and patient age made it difficult to really pinpoint what was the cause of the pain. However, latter an image from the emergency room (ER) came in with an otherwise healthy patient in their 30s experiencing abdominal pain. The image had much fewer pathologies present except for signs of appendicitis making it easy to pinpoint the cause of the pain. Indeed, the resident commented on the how ER images, where the patient is usually in good health, are much easier to read than one of patients in poor health. This to me presented an interesting point. Generally, someone who is unhealthy will need to be imaged, but they are likely to have multiple pathologies present making diagnosis harder. In other words, medical technology exists to help access disease, but the more diseased someone is the harder it is to use the output from technology (i.e. read the image). Although this may seem somewhat obvious, it never really hit me until now.

Another case that made me think was one involving an inferior vena cava (IVC) filter. This is a device that surgically placed into a patient to stop thrombi (from the lower exterminates) from reaching the pulmonary system. They are used when patients cannot be put on blood thinners and can be permanent or temporary (the temporary ones require surgical removal). One of the residents was telling me that these temporary devices need to be removed in the prescribed times, because otherwise they can cause complications (such as digging into the wall of the lumen and surrounding organs). This didn’t surprise me, but what did surprises me is the resident’s comment that the filters being left in too long is a common problem. This was the case on the image being evaluated (which since the filter was metal showed up easily on CT). The resident told me this issue is trying to be overcome with a better computer tracking system for these patients. This made me think about the possibility of a design allowing degradation of the filter instead of removal, mitigating issues of bookkeeping and reluctance of the patient to have the removal done. I looked at the literature and there are some clever designs already in the works.

One MR case that did get read was a defecography. This is where the patient has a bowel movement within the MRI machine and a real time image is evaluated. This was a procedure I never hear of before, and I actually had to ask if the “defec-” referred to defecate (because although that’s what it sounded like I just didn’t believe it at first). I did have to ask how the patient manages to do this in the bore and was told a pillow is placed under the knees thereby raising the anus. The clinical purpose of this procedure is to access things such as pelvic floor relaxation and prolapse of different organs. What was interesting to see was that geometry was used for grading. First a reference line was drawn on the image and then additional lines were drawn between this and different parts of the patient’s anatomy. Measurement of these lines and a look-up table was used to grade the prolapse of different organs and pelvic floor relaxation. As an engineer I quite liked this idea (of using geometry, measurements and a look-up table) and started to imagine how you might be able to get a computer to be able to do this automatically.

At the suggestion of Dr. Prince, I also attended a lecture, which was part of a series on clinical statistics. The topic was propensity scores. This technique dealt with the fact that in using already generated data (instead of a clinical trial), the groups that receive and do not receive treatment may have other underlying factors that make the groups "too different” to believe that the only thing being looked at is treatment or not. Propensity scores would essentially allow all the "other factors" (co-variates) that are not treatment to be warped up into a score for each person. The idea is that matching the scores between the treated and untreated group, so that you are left with two groups with similar distributions of scores, would mean the two groups are probably equal in every way except the receiving of treatment. This was interesting for me to learn for two reasons. First although I have studied several courses on statistics and stochastic models, I have never been exposed to how statistics is done clinically. Second, generally large numbers are need for statistics to be meaningful, and this may allow for larger sample sizes to be used allowing for "better" statistical tests.

I also spent time this week reviewing MR physics again as well as reviewing the literature about polycystic kidney disease.