Method and apparatus for diagnosis of breast tumors

• US 5,260,871 A
• Filed: 07/31/1991
• Issued: 11/09/1993
• Est. Priority Date: 07/31/1991
• Status: Expired due to Fees

• Current Assignee: Mayo Foundation For Medical Education And Research (Mayo Clinic)
• Original Assignee: Mayo Foundation For Medical Education And Research (Mayo Clinic)
• Inventors: Goldberg, Victor
• Primary Examiner(s): Envall, Jr., Roy N.
• Assistant Examiner(s): Poinvil, Frantzy
• Application Number:
  US07/738,460
• Time in Patent Office: 832 Days
• Field of Search: 364/413.08, 364/413.01, 364/413.13, 364/413.02, 364/413.14, 364/513, 382/56, 382/15, 382/16, 382/36, 382/6
• US Class Current: 600/320
• CPC Class Codes: G06T 2207/30096 Tumor; Lesion
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G16H 30/40 for processing medical images, e.g. editing
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: Y10S 128/922 including image analysis
• CPC Class Codes: Y10S 128/925 Neural network

Abstract

An apparatus for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient. A region of interest is located and defined on the ultrasonic image, including substantially all of the candidate tissue and excluding substantially all the normal tissue. The region of interest is digitized, generating an array of pixels intensity values. A first features is generated from the arrays of pixels corresponding to the angular second moment of the pixel intensity values. A second feature is generated from the array of pixels corresponding to the inverse contrast of the pixel intensity values. A third feature is generated from the array of pixels corresponding to the short run emphasis of the pixel intensity values. The first, second and third feature values are provided to a neural network. A set of trained weights are applied to the feature values, which generates a network output between 0 and 1, whereby the output values tend toward 1 when the candidate tissue is malignant and the output values tend toward 0 when the candidate tissue is benign.

Claims

1. A method for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient, comprising:

digitizing selected ultrasonic images from a plurality of patients generating an array of pixel intensity values;

entering into a computer said digitized ultrasonic images and target data based on excisional biopsy results for each of said patients where the candidate tissue is identified as malignant or benign;

locating and defining a region of interest in said digitized ultrasonic images, which includes substantially all of the candidate tissue and excludes substantially all normal tissue;

generating a first feature value from said arrays of pixels corresponding to an angular second moment of said pixel intensity values;

generating a second feature value from said arrays of pixels corresponding to an inverse contrast of said pixel intensity values;

generating a third feature value from said arrays of pixels corresponding to a short run emphasis of said pixel intensity values;

applying a set of weights to said first, second, and third feature values for generating values between an upper and lower value, whereby said output values tend toward said upper value when the candidate tissue is malignant and said output values tend toward said lower value when the candidate tissue is benign;

calculating a deviation between said output values with said target data; and

altering said weights until said deviation is reduced to a predetermined value.

2. The method of claim 1 for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient, further comprising the steps of increasing said target data above said upper value for malignant candidate tissue to increase the rate at which said altering step reduces said deviation to said predetermined value.

3. A method for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient, comprising:

digitizing said ultrasonic image generating an array of pixel intensity values;

locating and defining a region of interest in said digitized ultrasonic image, which includes substantially all of the candidate tissue and excludes substantially all normal tissue;

generating a first feature value from said array of pixels corresponding to a angular second moment of said pixel intensity values;

generating a second feature value from said array of pixels corresponding to a inverse contrast of said pixel intensity values;

generating a third feature value from said array of pixels corresponding to a short run emphasis of said pixel intensity values; and

applying a set of weights to said first, second and third feature values for generating network output values between an upper and lower value, whereby said output values tend toward said upper value when the candidate tissue is malignant and said output values tend toward said lower value when the candidate tissue is benign.

4. An apparatus for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient, comprising:

means for digitizing selected ultrasonic images from a plurality of patients generating an array of pixel intensity values;

means for entering into a computer said digitized ultrasonic images and target data based on excisional biopsy results for each of said patients where the candidate tissue is identified as malignant or benign;

means for locating and defining a region of interest in said digitized ultrasonic images, which includes substantially all of the candidate tissue and excludes substantially all normal tissue;

first extraction means for generating a first feature value from said arrays of pixels corresponding to a angular second moment of said pixel intensity values;

second extraction means for generating a second feature value from said arrays of pixels corresponding to a inverse contrast of said pixel intensity values;

third extraction means for generating a third feature value from said arrays of pixels corresponding to a short run emphasis of said pixel intensity values;

an untrained neural network means coupled to said first, second and third feature value extraction means for determining a set of weights to said feature values for generating network output values between an upper and lower value, whereby said output values tend toward said upper value when the candidate tissue is malignant and said output values tend toward said lower value when the candidate tissue is benign;

means for calculating a deviation between said output values with said target data; and

means for altering said weights until said deviation is reduced to a predetermined value.

5. The apparatus of claim 4 for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient, further comprising means for increasing said target data above said upper value for malignant candidate tissue to increase the rate at which said means for altering reduces said deviation to said predetermined value.

6. An apparatus for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient, comprising:

means for digitizing said ultrasonic image generating an array of pixel intensity values;

means for locating and defining a region of interest in said digitized ultrasonic image, which includes substantially all of the candidate tissue and excludes substantially all normal tissue;

first extraction means for generating a first feature value from said array of pixels corresponding to a angular second moment of said pixel intensity values;

second extraction means for generating a second feature value from said array of pixels corresponding to a inverse contrast of said pixel intensity values;

third extraction means for generating a third feature value from said array of pixels corresponding to a inverse contrast of said pixel intensity values;

a trained neural network means coupled to said first, second and third feature network means coupled to said first, second and third feature value extraction means for applying a set of weights to said feature values for generating network output values between an upper and lower value, whereby said output values tend toward said upper value when the candidate tissue is malignant and said output values tend toward said lower value when the candidate tissue is benign.