Page 24 - Vibrational Spectroscopic Imaging for Biomedical Applications

2 Cha pte r O n e

promising results for reliable epithelial and stromal recognition. This
approach also has the advantage of providing insight into impor-
tant biochemical properties of tissues, and sensitivity analysis of
the method is straightforward. We extend the classification to
include spatial measures of disease and propose that the highly
accurate results may lead to specific applications in areas of clin-
ical need.

1.1 Introduction
1
As 98 percent of localized breast cancers are treated effectively, all
women over 40 years old are recommended for annual mammogra-
2
phy screening and over 1.6 million breast biopsies are performed
3
each year to investigate screening abnormalities by removing a small
sample of tissue for further analysis. A manual examination of micro-
4
scopic structure (histology) within the biopsy to determine the cancer
type and grade forms the gold standard of diagnoses for most cancers. 5
Histologic examinations involve extensive human interpretation,
7
6
making consistency difficult and second opinions necessary. Further,
patients often wait days or weeks to receive a pathology report follow-
8
ing a tumor biopsy. Although 80 percent of these biopsies are eventu-
9
ally diagnosed as benign, this extended waiting period is associated
10
with substantial distress in all biopsy patients. When a report is
available, intra- and interobserver variability in diagnosis and treat-
11
ment recommendations ranges from 1 to 43 percent. Illustrative of
these concerns is a study of 481 breast cancer patients from 1982 to
2000 at a regional cancer center that revealed that 73 percent of breast
ductal carcinoma in situ (DCIS) patients were referred by a general
pathologist for review by an expert pathologist. After expert patholo-
gist‘s review, 43 percent of these cases received different treatment
recommendations and 29 percent of these cases had a change in assess-
12
ment of cancer recurrence risk. A separate study found that 52 percent
of patients referred to a multidisciplinary breast cancer review board at
a university hospital for a second evaluation received a change in
surgical treatment recommendation. This delay and variability in
13
tumor diagnosis may impact studies that guide basic science and
clinical decision making. Clearly, the process is suboptimal; improve-
ments in cancer diagnosis and prognosis prediction are of wide
15
interest to clinicians, pathologists, health insurance companies, 16
14
and the general public. 17
The root cause of these problems in cancer diagnoses, leading to
complications in treatment and research, is the inability to univer-
sally provide rapid, accurate, and reproducible histologic determi-
nations. Technology to address these needs for cancer histopathology
can, thus, prove to be of central importance to cancer research and
treatment. Imaging-based technology for this purpose is especially