Radiology

Dr. James Thrall, chief of radiology, Massachusetts General Hospital, talks about the digital divide in medicine and how some hospitals have begun to cross it. The process of reading breast, colon, and other high-volume scans "manually"i.e., by a trained radiologist reading them one after anothercan become tedious and error prone if done for long periods of time. By contrast, computers never tire, seldom complain, and can now be fantastically accurate based on "training" from analysis of massive historical data:

Departments of radiology that have crossed the divide behave differently than those that haven't. We used to think that reaching the divide was the goal, because we could envision all that could be better. In reality, the goal is the exploitation of data, not the collection...We currently save all radiological images, of everything, indefinitely. Why? Because of the mining value...We recently went beyond our 100-millionth digital image captured, and at the rate we're going we will have a billion images before long. We're adding a terabyte of disk space every other week, because we need massive numbers of cases in order to get good inferences and correlations. [X-rays and other radiological data can easily be many megabytes of data per image or even from a single "slice" of part of an anatomyheavy data to store and more hefty to sift and search for patterns.]

Radiology, and medicine in general, is gravitating toward computer-aided diagnostics. However, doctors need years of digitized data to make this process effective. This is particularly true in radiology. "We can show that 5 or 10 years later what was predicted at diagnosis time was actually true," continues Dr. Thrall. "There is a time enhancement of the verity of the diagnosis now as a result of the medical image histories." An important characteristic of Inescapable Data is to have enough data and over a long enough time period that valuable inferences can be discovered. The data, at least radiological data, is now being collected.

We expect to see a huge increase in computer-aided diagnostic usage coming, made possible by having petabytes of historical image data and analysis already collected. Note that a petabyte is equal to 1,024 terabytesan amount of data that would have filled a warehouse with disk drives just a few years ago. It can actually be argued that the data becomes the hypothesis. "The data is screaming at you that there is something going on here in addition to whatever hypothesis you started with. You have all the data. Now figure it out," exclaims Dr. Thrall. In fact, with massive repositories, some purely data-oriented analysis techniques can distill meaningful patterns without specific medical background. A hope of the Inescapable Data world is that unrelated industries will derive techniques and tools applicable to other industries, as we saw with bioinformatics analysis being used for e-mail spam prevention.

Radiology data is an Inescapable Data natural given its heft and pure digital format. Although sharing and having easy-to-exploit formats allows for better care, some worry that it raises privacy and security concerns. To that point, Dr. Thrall states, "With paper or film records, there was no security; folders of records were left lying around, and anyone could observe them. At least with digital forms, you need authentication, passwords, and access activities are logged and audited."

Other hospital patient records are not quite as far along as radiological data. "The history of hospitals tracking patient data and charts is abysmal," continues Dr. Thrall. "When the data was on paper and film, there was simply no practical way of sharing it. Now, even if it's in digital form, sharing remains elusive. I can travel to Europe and find a bank machine that effectively has my bank records and allows me access to my money. If in that same town I have an accident, there is no ability to get access to my medical records." Barriers to interinstitutional sharing still exist, but like airline reservations and electronic banking, the industry will mature and learn that connectivity at the patient record level allows them to leverage the world rather than just their locality.

We will get to the point where there is a massive fluidity of all patient data so that if you get sick in a different state or city you can get to the information. Today, individual hospitals have crossed the divide, but are not linked, an opportunity for the Inescapable Data world.