http:^^www.cs.wisc.edu^~olvi^uwmp^cancer.html
来自「This data set contains WWW-pages collect」· HTML 代码 · 共 404 行 · 第 1/2 页
HTML
404 行
Date: Thu, 07 Nov 1996 19:23:09 GMTServer: NCSA/1.5Content-type: text/htmlLast-modified: Tue, 10 Sep 1996 22:05:24 GMTContent-length: 15138<HTML><HEAD><TITLE>Machine Learning for Cancer Diagnosis and Prognosis</TITLE></HEAD><BODY><H1>Machine Learning for Cancer Diagnosis and Prognosis</H1><P><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><IMG SRC="http://www.cs.wisc.edu/~olvi/uwmp/92_5622_www.gif" align=middle><P>This page describes various linear-programming-based machine learningapproaches which have been applied to the diagnosis and prognosis ofbreast cancer. This work is the result of a collaboration at theUniversity of Wisconsin-Madison between<!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><A HREF="http://www.cs.wisc.edu/~olvi/olvi.html">Prof. Olvi L. Mangasarian </A>of the Computer Sciences Department and<!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><A HREF="http://www.biostat.wisc.edu/people/profiles/people/WOLBERG,_WILLIAM_H.html">Dr. William H. Wolberg</A>of the departments of Surgery and Human Oncology.<P><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><A HREF="http://www.cs.wisc.edu/~street/press_rel.html">Here</A> is a copy of thepress release distributed at the American Cancer Society Science Writers seminar inMarch of 1994. It provides a good overview of this research.<P><HR><H2>Table of Contents</H2><UL> <LI> <!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><A HREF="#diag">Diagnosis</A> <LI> <!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><A HREF="#prog">Prognosis</A> <LI> <!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><!WA6><A HREF="#bib">Bibliography</A> <LI> <!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><!WA7><A HREF="#pop">Citation in the Popular Press</A> <LI> <!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><!WA8><A HREF="#lref">Local Related Links</A> <LI> <!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><!WA9><A HREF="#oref">Other Related Links</A></UL><HR><A NAME="diag"><H2>Diagnosis</H2></a>This work grew out of the desire by Dr. Wolberg to accurately diagnosebreast masses based solely on a Fine Needle Aspiration (FNA). Heidentified nine visually assessed characteristics of an FNA sample which he consideredrelevant to diagnosis. In collaboration with Prof. Mangasarian andtwo of his graduate students, Rudy Setiono and <!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><!WA10><A HREF="http://www.rpi.edu/~bennek">Kristin Bennett</A>, aclassifier was constructed using the multisurface method (MSM) of pattern separation on these nine features thatsuccessfully diagnosed 97% of new cases. The resulting data set iswell-known as the <!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><!WA11><AHREF="ftp://ftp.cs.wisc.edu/math-prog/cpo-dataset/machine-learn/cancer1/">Wisconsin Breast Cancer Data.</A><P>The image analysis work began in 1990 with the addition of <!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><!WA12><A HREF="http://www.cs.wisc.edu/~street/street.html">Nick Street</A>to the research team. The goal was to diagnose the sample based on adigital image of a small section of the FNA slide. The results ofthis research have been consolidated into a software system known as <b>Xcyt</b>, which is currently used by Dr. Wolberg in his clinicalpractice. The diagnosis process is now performed as follows:<UL> <LI> An FNA is taken from the breast mass. This material is thenmounted on a microscope slide and stained to highlight the cellularnuclei. A portion of the slide in which the cells arewell-differentiated is then scanned using a digital camera and aframe-grabber board. <LI> The user then isolates the individual nuclei using <b>Xcyt</b>.Using a mouse pointer, the user draws the approximate boundary ofeach nucleus. Using a computer vision approach known as "snakes",these approximations then converge to the exact nuclear boundaries.This interactive process takes between two and five minutes per slide.<!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><!WA13><A HREF="http://www.cs.wisc.edu/~street/xcyt1.gif">Here</A> is an image showing<b>Xcyt</b> in use. <LI> Once all (or most) of the nuclei have been isolated in thisfasion, the program computes values for each of ten characteristics ofeach nuclei, measuring size, shape and texture. The mean, standarderror and extreme values of these features are computed, resulting ina total of 30 nuclear features for each sample. <LI> Based on a training set of 569 cases, a linear classifier wasconstructed to differentiate benign from malignant samples. Thisclassifier consists of a single separating plane in the space of threeof the features: Extreme Value of Area, Extreme Value of Smoothness,and Mean Value of Texture. By projecting all the cases onto thenormal of this separating plane, approximate <!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><!WA14><A HREF="http://www.cs.wisc.edu/~street/xcyt3.gif">probability densities</A> ofthe benign and malignant points were constructed. These allow a simple Bayesiancomputation of probability of malignancy for new patients. Thesedensities are shown to the patient, allowing her to judge the"confidence" of her diagnosis by comparison to hundreds of previous samples.</UL>To date, this system has correctly diagnosed 176 consecutive newpatients (119 benign, 57 malignant). In only eight of those cases did<b>Xcyt</b> return a "suspicious" diagnosis (that is, an estimatedprobability of malignancy between 0.3 and 0.7). <P>A small subset of the source images used in this research can be found<!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><!WA15><a href = "http://www.cs.wisc.edu/~street/images"> here. </a> These are very goodtest cases for image segmentation or object recognition algorithms. If your petsegmentation algorithm can automatically identify all of the nuclei inthese images, please email me (street@cs.wisc.edu) and let's work together.<P><HR><A NAME="prog"><H2>Prognosis</H2></a>The second problem considered in this research is that of prognosis,the prediction of the long-term behavior of the disease. We haveapproached prognosis as a function-approximation problem, using inputfeatures -- including those computed by <b>Xcyt</b> -- to predict atime of recurrence in malignant patients, using right-censored data.Our solution is termed the Recurrence Surface Approximation method (RSA), and utilizes a linearprogram to construct a surface which predicts time of recurrence fornew patients. By examining the actual recurrence of those training caseswith similar predicted recurrence times, we can plot the probability ofdisease-free survival for various times (out to 10 years) for anindividual patient. This capability has been incorporated into<b>Xcyt</b> and an example is shown <!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><!WA16><A HREF="http://www.cs.wisc.edu/~street/xcyt4.gif">here.</A>These survival curves plot the probability of disease-free survival versus time (in years).The black disease-free survival curve represents all patients in ouroriginal study; the red curve represents the probability ofdisease-free survival for the sample case. This particular case thereforehas an above-average prognosis, with a probability of being disease-freeafter 10 years equal to about 80%.<P>The RSA procedure can also be used to compare the predictive power ofvarious prognostic factors. Our results indicate that precise,detailed cytological information of the type provided by <b>Xcyt</b>gives better prognostic accuracy than the traditional factors TumorSize and Lymph Node Status. If corroborated by other researchers,this result could remove the need for the often painful axillary lymph node surgery.<HR><A NAME="bib"><H2>Chronological Bibliography</H2></a>Linked papers are provided in postscript format; if you don't have a postscript viewer, you can download the file (e.g., shift-click in Netscape)and print it. Abstracts are ASCII text. To obtain papers which are notlinked, please contact the first author.<DL> <DT> <B> O.L. Mangasarian, R. Setiono and W.H. Wolberg.</B> <DD> Pattern Recognition via Linear Programming: Theory and Application to Medical Diagnosis. In <EM> Proceedings of the Workshop on Large-Scale Numerical Optimization</EM>, 1989, pages 22-31, Philadelphia, PA. SIAM. <DT> <B> O.L. Mangasarian and W. H. Wolberg.</B> <DD> Cancer Diagnosis via Linear Programming. <EM>SIAM News, </EM> Vol. 23, 1990, pages 1 & 18. <DT> <B> W.H. Wolberg and O.L. Mangasarian.</B> <DD> Multisurface Method of Pattern Separation for Medical Diagnosis Applied to Breast Cytology. <EM>Proceedings of the National Academy of Sciences, U.S.A., </EM> Vol. 87, 1990, pages 9193-9196. <DT> <B> W.N. Street.</B> <DD> Toward Automated Cancer Diagnosis: An Interactive System for Cell Feature Extraction. Technical Report 1052, Computer Sciences Department, ⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?