, The Cleveland Clinic
, Dartmouth College
Pages: pp. 14-17
Modern computing in medicine might well have started with the early radiation therapy treatment planning systems. 1-3 These systems did the lengthy and laborious calculations needed to assure that the radiation dose distributions delivered to patients reasonably matched the physicians' prescriptions. Furthermore, results of these calculations strongly depend on the mathematical-physical dose-deposition model; accurate parametric radiation beam modeling; accurate modeling of beam modifiers such as shielding blocks, wedges, and compensators; and a sufficiently accurate and detailed patient data set. Early treatment planning systems used simple attenuation dose-deposition models, few data points to describe the radiation beam profile, and geometric wedge modeling. They also modeled the patient as a cylinder of water. More modern systems use much better semiempirical dose-deposition models and increasingly accurate data including high-resolution (CT) computerized tomography images of the patient's structures.
In diagnostic medicine, computation became a critical component with the introductions of CT, 4 ultrasonography, 5 nuclear medicine imaging, 6 and magnetic-resonance imaging. 7 At first, computing's role was in data acquisition and image computation from the data and then image processing. Diagnostic medicine computing today includes automatic segmentation of—and disease diagnosis from—those images.
Other areas of computing in medicine include electronic charting and picture archiving and communications systems. Each requires massive databases and interoperability among applications from many vendors. Solutions to the interoperability issues are underway with the Digital Image Communications in Medicine (DICOM) 8 and Health Level Seven (HL-7) standards. Teleradiology 9 is another emerging computing-in-medicine application that enables consultations over the Internet. It requires significant attention to bandwidth-conserving techniques and to the associated computer power required to encrypt and compress and decompress in real time.
The four articles in this issue illustrate some of the applications of modern high-end computing in medicine. Christine L. Hartmann Siantar and Dewey Garrett show us a glimpse of the next step in specificity and accuracy in radiation therapy treatment planning using Monte Carlo dose calculations for each beam for each target for each patient. Although not yet as fast as traditional semiempirical methods, the Monte Carlo techniques offer greater accuracy and hope for better disease control. In diagnostic medicine, Maryellen L. Giger's article describes how advances in computer vision and artificial intelligence can aid in disease diagnosis. At the very least, these techniques can draw the attention of a human expert to a subtle shading on an image.
Scott L. Delp and J. Peter Loan show how we can create mathematical models of the musculoskeletal system. This lets us predict the effects of surgery on walking and locomotion. Idith Haber, Dimitris N. Metaxas, and Leon Axel take a similar look at internal organs, specifically the heart. Their work lets us reconstruct the heart and how it beats, perhaps helping us predict and determine diseases and their effects.
Computers have greatly influenced medicine in the last 50 years—the revolution will be in computation and the underlying mathematics, which will dramatically alter the practice of medicine over the next 50 years. 10
This issue brings together four articles that represent the two fundamental applications of computational systems to medicine: computer-aided diagnosis and surgery. Although we've made much progress in CAD and CAS, we're still a long way away from automated diagnosis and remote surgery.
But as mathematics ultimately improves and our knowledge of the human body increases, we will be able to better predict the outcomes of surgery, the responses to medicine, and possibly the long-term effects of certain lifestyle choices such as smoking, alcohol, and obesity. A generalized approach that creates a flexible model of the human will let us superimpose information from various sources on a model of a specific patient. The articles in this issue lay down the framework for this revolution in medicine to occur.
Hopefully, we'll be able to one day predict the outcome of therapies for a given patient, rather than predict the outcome from accumulated outcomes from other patients in the population, as we do today. This would let us more rapidly alter our therapies in a positive manner as researchers introduce new therapies in surgery and medicine.