Xcyt provides accurate, repeatable diagnosis and prognosis of breast
cancer cytology. The program combines techniques from computer
vision / image analysis, mathematical programming, inductive machine
learning, and statistics. The result is a system that diagnoses fine
needle aspirates as well or better than the best cancer centers, and
as well as any system we know of short of surgical biopsy. The
prognostic system provides precise, individualized predictions of
disease-free survival without the need for axillary lymph node
dissection.
We invite you to try out the program on our sample images. If you like what you see, please contact us regarding collaborative research. Use of Xcyt is free, and will stay that way until somebody makes us a better offer.
Standard disclaimer: the predictive results of Xcyt are reliable when applied to identically-prepared samples, as documented in our various publications. However, the results are only advisory, and should be interpreted by a qualified physician. We only evaluate and report past performance; we do not guarantee future performance.
Accuracy summary
The diagnostic system using the current classifier has been in clinical use at the University of Wisconsin Hospitals and Clinics since 1993. It's estimated accuracy at the time of deployment was 97.5%. It has correctly diagnosed 98.9% of the 273 new cases on which it has been tested in the past six years.
The prognostic system has been demonstrated to separate good from poor prognoses as well or better than the traditional prognostic factors of axillary lymph node status and tumor size.
Fine needle aspriate preparation
There are several methods for preparing FNAs which produce different morphology. Xcyt reliability requires that samples be prepared in the same manner as the calibrating slides.
Aspriates were expressed on a slide and spread between this and a similar slide placed face-to-face on the first. Then, the slides were rapidly separated in a horizontal direction and immediately fixed in 95% ethanol. The slides were stained with hematoxylin and eosin.
Clinical use of Xcyt
As with any other laboratory test, Xcyt results must be interpreted in the total clinical context. Remember that there are false positives and false negatives although the overall accuracy exceeds 97%. Particular attention should be given to samples where the estimated probability of malignancy is between 0.30 and 0.70. All lesions that are not removed should be followed clinically in 3 and 12 months.
The overall accuracy of FNA in diagnosing breast masses is enhanced by using visual examination in conjunction with Xcyt. False negative results occur in small-cell lobular cancers that are particularly difficult for Xcyt to diagnose. Visual inspection will reveal small spherical clusters of cells. Additionally, intracytoplasmic vacuoles may be present.
False positive results occur most frequently in fibroadenomas. Regardless of the cytologic diagnosis, an excisional biopsy with frozen section is usually indicated prior to definitive breast surgery both to confirm the cytologic diagnosis and to determine the need for and the extent of axillary node dissection. Of specific interest is whether the cancer is in situ or invasive, or whether it is of a special indolent type such as tubular.