Method for predicting the future occurrence of clinically occult or non-existent medical conditions

• US 5,862,304 A
• Filed: 10/31/1990
• Issued: 01/19/1999
• Est. Priority Date: 05/21/1990
• Status: Expired due to Fees

• Current Assignee: Board of Regents of the University of Texas System (University of Texas System)
• Original Assignee: Board of Regents of the University of Texas System (University of Texas System)
• Inventors: Clark, Gary M., McGuire, William L., Ravdin, Peter M.
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US07/607,120
• Time in Patent Office: 3,002 Days
• Field of Search: 364/413.1, 364/513, 395/22
• US Class Current: 706/21
• CPC Class Codes: G06N 3/065 Analogue means
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: Y10S 706/924 Medical

Abstract

A method is presented for evaluating data to predict the future occurrence of a medical condition that is presently clinically occult or which has not yet occurred. Specifically, the method uses a neural network to analyze and interpret DNA flow cytometric histograms. A first set of DNA histograms taken from tumors from patients having known relapse rates are used to train the neural network, an then the trained network is applied to predict the relapse rates of patients using DNA histograms of tumors from those patients. Prognosis according to this method can be performed using only diploid histograms, using only aneuploid histograms, or using a combination of diploid and aneuploid histograms.

Claims

1. A method of characterizing DNA cytophotometric data to predict the future occurrence of a target medical condition that is presently clinically occult or non-existing, comprising:

providing a neural network;

data compressing DNA cytophotometric data generated from cells of patients having known occurrence or non-occurrence of said target medical condition to produce a first set of DNA cytophotometric data;

training said neural network using said first set of DNA cytophotometric data; and

predicting future occurrence of said target medical condition for at least one additional patient using DNA cytophotometric data obtained from cells of said at least one additional patient and said trained neural network.

2. A method of predicting the relapse of cancer that is presently clinically occult or non-existing, comprising:

providing a neural network;

data compressing a plurality of DNA flow cytometric histograms generated from tumor cells of patients having known cancer relapse rates to produce a first set of histograms;

training said neural network using said first set of DNA flow cytometric histograms;

obtaining at least one DNA flow cytometric histogram from tumor cells of a patient having an unknown cancer relapse rate; and

predicting relapse of cancer in said patient having an unknown relapse rate using said at least one DNA flow cytometric histogram and said trained neural network.

3. The method of claim 2, said first set of DNA flow cytometric histograms including diploid histograms and aneuploid histograms.

4. The method of claim 2, said first set of DNA flow cytometric histograms including only diploid histograms.

5. The method of claim 2, said first set of DNA flow cytometric histograms including only aneuploid histograms.

6. The method of claim 2, said cancer comprising breast cancer.

7. A method of predicting the future occurrence of breast cancer that is presently occult or non-existing, comprising:

providing a neural network;

data compressing DNA cytometric data generated from cells of patients having known occurrence of breast cancer to produce a first set of DNA cytometric data;

training said neural network using said first set of DNA cytometric data and first sets of known data, each of said first sets of known data including a predetermined number of prognostic input variables, and corresponding known breast cancer occurrence, said prognostic input variables chosen according to capability to predict occurrence of breast cancer; and

predicting future occurrence of breast cancer for a second set of DNA cytometric data and second sets of data using said trained neural network, each of said second sets of data including only said predetermined number of prognostic input variables.

8. The method of claim 7, said neural network comprising a back-propagation class neural network.

9. The method of claim 7, said training step comprising conditioning said first sets of known data using a back-propagation of errors neural network training algorithm.

10. The method of claim 7, further comprising quantizing each of said predetermined number of input variables before said training step.

11. The method of claim 10, said quantizing step comprising:

defining a range for each of said input variables;

dividing each of said ranges into subranges; and

determining a subrange within which each of said input variables falls.

12. The method of claim 7, said prognostic input variables comprising proposed treatments for breast cancer.

13. The method of claim 7, said prognostic input variables comprising progesterone receptor status, tumor size, cathepsin D protein level and HER-2/neu protein level.