This is the first of a two-part series on cancer imaging. This week's article describes how radiologists more effectively image anatomy to produce more effective diagnoses and improved patient care. Next week's article will discuss new contrast agents that can help illustrate function and aid in drug delivery.

Biomedical imaging of cancer patients has become essential to state-of-the-science oncology practice. Over the last few decades, radiology and imaging sciences have been riding the same wave of discovery and innovation as has the rest of cancer medicine. Imaging advances have powered, demonstrated, and increased understanding of carcinogenesis at the cellular and molecular levels.

"Imaging in the 21st century will follow the historical cycle of innovation in radiology that dates back to Roentgen's X-ray discovery 110 years ago," predicts NIH Director Dr. Elias Zerhouni. Dr. Zerhouni, who was chairman of the radiology department at Johns Hopkins University School of Medicine before coming to NIH, spoke at CCR Grand Rounds on February 15. "You begin with a biological question as a signpost. Then you marry technological advances in scanners and computers to invent new medical devices to approach that question and further interrogate the biology, which suggests how to tailor novel therapeutics. We are heading into a clinical world that will rely profoundly on very powerful, minimally invasive therapeutics. We're at that frontier right now and in vivo cellular and molecular imaging beckon us forward."

One of the cardinal tenets of biomedical research holds that "biology is messy." At this month's Sixth Annual National Forum for Biomedical Imaging in Oncology - which focused on the challenge of transforming inherently subjective pictures of biology into repeatable, valid, and quantifiable datasets - Dr. Daniel Sullivan, director of NCI's Cancer Imaging Program (CIP), explained why "segmentation is the mother of all problems."

"The major challenge for oncologists when looking at a structural image is to be able to define what is tumor and what is not," says Dr. Sullivan. "The field or concept of fuzzy mathematics is one approach that can be used to try to pin down the edges of a tumor." Radiologists use highly skilled mental algorithms to perform this task, he notes, and have been remarkably effective, although also notoriously variable, in these judgments. Still, radiologists have done better than early computer attempts to accomplish the same diagnostic task, he adds.

But, once X-ray data became digitized with computerized tomography (CT) in the 1970s, the urge to apply computer algorithms proved irresistible. "You could capture the complexity of reality in a way that just wasn't available before," says Dr. Zerhouni.

Researchers began to dream of the statistical power that could be generated from the images drawn from hundreds of thousands of cancer patients around the country. But before they could begin to put this potential data into a meaningful scientific context, they needed to face the "messy biology" problem and also avoid "comparing apples to oranges," Dr. Sullivan says. The randomized controlled clinical trial is the ultimate scientific framework for turning valid data into meaningful information, he continues, but the obstacles relating to using data derived from thousands of individual, human radiologists reading thousands of images of different tumors from many different machines and technologies were formidable.

NCI addressed this dilemma by supporting the American College of Radiology Imaging Network (ACRIN). Dr. Carl C. Jaffe, chief of NCI's Diagnostic Imaging Branch, helps to oversee the unique national cooperative group that has been conducting clinical trials in imaging since 1999. ACRIN's trials address the four major uses of imaging in cancer care: screening, diagnosis and staging, image-guided treatment, and measuring response to treatment. Underlying all is the goal of generating information that will lengthen and improve the quality of patients' lives.

The main idea, explains Dr. Jaffe, "is to establish quality control over the process" by developing a model that allows you to aggregate data from widely scattered sites into a truly meaningful process "that rigorously tests biological response in a verifiable, repeatable way."

"We're not using machines to actually read the films," continues Dr. Jaffe. "However, we calibrate the machines and systematize the radiologist's approach to the image." Once captured, an image's digital data can be processed by algorithms that further sort and standardize it. "We use blinding and randomization to present the films to the radiologists, and always have a second independent reading, sometimes a third," he notes. "In essence, ACRIN provides a core facility that embodies the rigor and scientific control you need for a national study with geographically distant sites. From all of this quality assurance comes the beginning of quantifiable data in oncology."

While CIP is only one of the six major programs within NCI's Division of Cancer Treatment and Diagnosis, imaging is a crucial part of NCI's long-term strategy. In NCI's budget proposal for fiscal year 2006, imaging was designated an area of "extraordinary opportunity" for the institute. CIP includes other major projects besides ACRIN, including imaging technology, molecular imaging, and image-guided intervention.

"We're almost like the early explorers," Dr. Zerhouni observes. "We're beginning to explore the frontiers of genetics and biology at the molecular level." He is confident that imaging science will be the primary guide to that terrain because it increasingly enables researchers to look directly into the metabolizing cell. "More and more, the business of clinical medicine will be to deliver treatment drugs and nano-size tools to highly specific sites in the body," such as tumor tissues, he says. "We will use imaging to guide us to the target and then to capture the relevant biological response in real time."

By Addison Greenwood