Neural network for cell image analysis for identification of abnormal cells

• US 6,463,438 B1
• Filed: 06/03/1994
• Issued: 10/08/2002
• Est. Priority Date: 06/03/1994
• Status: Expired due to Fees

• Current Assignee: Laboratory Corporation Of America Holdings
• Original Assignee: UroCor, Inc.
• Inventors: Hu, Ying, Ashenayi, Kaveh, Veltri, Robert W., OʼDowd, Gerard J.
• Primary Examiner(s): Shah, Sanjiv
• Application Number:
  US08/253,933
• Time in Patent Office: 3,049 Days
• Field of Search: 395/22, 395/23, 395/21, 364/413.02, 364/413.1, 607/14, 128/736, 706/15, 706/16, 706/20, 706/13, 706/28, 706/31
• US Class Current: 706/15
• CPC Class Codes: G06T 2207/10056 Microscopic image
• CPC Class Codes: G06T 2207/30024 Cell structures in vitro; Tissue sections in vitro
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 20/695 Preprocessing, e.g. image segmentation
• CPC Class Codes: G06V 20/698 Matching; Classification

Abstract

A neural network is used in a system to detect abnormalities in cells, including cancer in bladder tissue cells. The system has an image analysis system for generating data representative of imaging variables from an image of stained cells. The set of data is provided to a neural network which has been trained to detect abnormalities from known tissue cells with respect to the data from the same set of imaging variables. A conventional sigmoid-activated neural network, or alternatively, a hybrid neural network having a combination of sigmoid, gaussian and sinusoidal activation functions may be utilized. The trained neural network applies a set of weight factors obtained during training to the data to classify the unknown tissue cell as normal or abnormal.

Claims

1. A program storage device having a computer readable program code embodied therein for classifying an unknown tissue cell, said program storage device comprising:

a first computer readable code, said first computer readable code implementing a neural network having a plurality of neurons, said neural network having a plurality of interconnections coupling said plurality of neurons;

a second computer readable code, said second computer readable code having a plurality of weight factors representative of said interconnections, said weight factors having been derived by training said neural network on a training data set representative of a plurality of imaging variables from a known tissue cell, said imaging variables comprising an area, an average intensity, a shape, a texture, and a DNA content; and

a third computer readable code, said third computer readable code capable of applying a sample data set from said unknown tissue cell representative of said plurality of imaging variables to said weight factors of said second computer readable code for classifying said tissue cell as normal or abnormal.

2. The program storage device of claim 1, wherein each of said neurons has an activation function, said activation comprising sinusoid, gaussian, and sigmoid functions.

3. The program storage device of claim 2, wherein said neural network has a plurality of said activation functions.

4. The program storage device of claim 2, wherein said activation function is flattened.

5. The program storage device of claim 1, wherein each of said plurality of neurons comprises:

a fourth computer readable code representative of a plurality of input ports, each of said input ports capable of storing an input value and a corresponding weight value selected from said weight factors;

a fifth computer readable code representative of an accumulator, said accumulator capable of generating an output by summing each multiplication of said input value and said corresponding weight value of each of said plurality of input ports; and

a sixth computer readable code representative of a function generator having a plurality of activation functions, said function generator capable of applying one of said plurality of activation functions to the output of said accumulator to generate an output value for said neuron.

6. A program storage device readable by a machine, tangibly embodying a program of instructions executable by the machine to perform a method of detecting the presence of an abnormality in an unknown tissue cell, said method comprising the steps of:

applying a sample data set from said unknown tissue cell to a neural network, said sample data set being representative of a plurality of imaging variables from said unknown tissue cell, said imaging variables comprising an area, an average intensity, a shape, a texture, and a DNA content, said neural network having been trained to detect the presence of said abnormality on a training data set from a known tissue cell representative of said plurality of imaging variables; and

classifying said tissue cell as normal or abnormal based on said training of said neural network.

7. An apparatus for classifying an unknown tissue cell, said apparatus comprising:

a plurality of data buffers, said data buffers capable of storing data corresponding to a plurality of imaging variables from said unknown tissue cell, said imaging variables comprising an area, an average intensity, a shape, a texture and a DNA content;

a neural network, said neural network coupled with said plurality of data buffers, said neural network having been trained on a training data representative of said imaging variables from a known tissue cell, said neural network classifying said unknown tissue cell as normal or abnormal based on said training of said neural network.