President Obama and others argue that better science will save time, money, and lives. Some doubt the argument, but it’s true. Here’s a promising example that popped up this past week. It’s this March 4 ScienceDaily article on successful use of ultrasound and computer algorithms to distinguish between benign and malignant masses in the breast. The new technique appears to be much easier for the woman, as well as being cheaper,faster and safer than conventional approaches. The method draws heavily from today’s strong computer power. Key parts of the article are pasted below, but it’s well worth a full read for both the science and some insight into the changes being brought by today’s computing power and creative thinking by a bright PhD student.
"ScienceDaily (Mar. 4, 2011) — Recent research by doctoral student Sevan Goenezen holds the promise of becoming a powerful new weapon in the fight against breast cancer. His complex computational research has led to a fast, inexpensive new method for using ultrasound and advanced algorithms to differentiate between benign and malignant tumors with a high degree of accuracy.
Goenezen’s research offers the hope of dramatically reducing the need for invasive, uncomfortable, and stress-inducing biopsies, and perhaps even replacing mammograms. It uses a new technique to analyze images captured with a noninvasive, radiation-free ultrasound device, locate tumors, and determine if the tumor is malignant. The only required equipment is a specific type of ultrasound machine — which generally costs around $10,000, far less than X-ray equipment — and a common personal computer. Thanks to these new algorithms, results can be computed in less than five minutes on a high-end PC.
This new technique uses ultrasound images of breast tissue to infer the mechanical properties of the tissue as it is compressed. The structure of collagen fibers within malignant tissues is very different from the collagen fiber structure in benign tissue. This method quantifies the non-linear behavior of the tumor tissue to determine whether it is cancerous.
In a clinical study, Goenezen used this strategy to analyze 10 data sets, five of which were from patients with benign tumors, and five with malignant tumors. The system correctly diagnosed nine out of the 10 patients. The lone error was a false positive, meaning the system indicated the tumor was malignant when it was actually benign."