# US 2008/0306898 A1 · Method for reliable classification of samples in clinical diagnostics using an improved method of classification

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2008/0306898 A1
- Document ID: US20080306898A1
- Status: Published application
- Kind: A1
- Application number: 12/218,519
- Filed: 2008-07-15
- Published: 2008-12-11
- Earliest claimed: 2006-03-31
- Assignees: Biodesix, Inc.
- Inventors: Maxim Tsypin; Heinrich Roder
- Web page: https://patentgrep.com/patent/US20080306898A1

## CPC

- G16C 20/70: Machine learning, data mining or chemometrics
- G06F 18/2414: {Smoothing the distance, e.g. radial basis function networks [RBFN]}
- G06F 18/24147: {Distances to closest patterns, e.g. nearest neighbour classification}
- G16C: COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A system for classification of a test object using a training set comprising a plurality of objects, each of which is assigned as a member of a class. Collectively, the objects in the training set are members of at least two classes. A computer system is configured as a probabilistic classifier. The classifier estimates the probability of the test object being a member of each of the classes in the training set. The probabilistic classifier estimates the probability with reference to the class assignments of the objects in the training set which are neighbors to the test object within a defined region within the training set. The probabilistic classifier takes into account the situation in which there is an imbalance in the number of objects in the different classes in the training set. Additionally, the probabilistic classifier does not require any knowledge of the probability distribution function of the classes in the training set.

## Claims

### Claim 1

A system for classification of a test object, comprising:
a machine-readable storage device storing a training set comprising a plurality of objects, each of the objects being assigned as a member of a class, collectively the objects in the training set comprising members of at least two classes; and
a computer system coupled to the machine-readable storage device configured as a probabilistic classifier, the probabilistic classifier estimating the probability of the test object as being a member of each of the classes in the training set, wherein the probabilistic classifier estimates the probabilities with reference to the class assignments of objects in the training set which are neighbors to the test object within a defined region within the training set; and
wherein the training set comprises objects being members of two classes, class 1 and class 2, wherein classes 1 and 2 in the training set have N1 and N2 numbers of objects respectively, in the training set and wherein the computer system is programmed to determine the probability P(class 1) of the object being a member of class 1 according to the relation:
P

(
class


1
)
=
k
1
+
1
k
1
+
k
2
+
2
·
F
1
2

(
1
,
k
2
+
1
;
k
1
+
k
2
+
3
;
1
-
N
1
N
2
)
.
where 2F1 (1, k2+1; k1+k2+3; 1−N1/N2) is a calculation of the hypergeometric function 2F1 and where k1 represents the number of neighbors of class 1 within the defined region and k2 represents the number of neighbors in class 2 within the defined region.

### Claim 2 (depends on 1)

The system of claim 1, wherein the defined region within the training set comprises a sphere ω centered about the test instance, and wherein there are k1 neighbors of class 1 within ω and there are k2 neighbors in class 2 within ω.

### Claim 3 (depends on 2)

The system of claim 2, wherein the size of sphere ω is selected such that probability density for classes 1 and 2 are approximately constant within sphere ω.

### Claim 4 (depends on 1)

The system of claim 1, where the computer system is further configured to compare the estimated probabilities for each of the classes to determine if the difference in estimated probability between the class having the highest probability estimate and the class having the next highest probability estimate is greater than a user-supplied parameter.

### Claim 5 (depends on 4)

The system of claim 4, wherein the user-supplied parameter is 0.1.

### Claim 6 (depends on 1)

The system of claim 1, wherein the probabilistic classifier comprises a probabilistic k nearest neighbor classifier.

### Claim 7 (depends on 1)

The system of claim 1, wherein the defined region comprises a set of k objects within the training set which are the k closest objects to the test object.

### Claim 8 (depends on 1)

The system of claim 1, wherein the objects comprise the intensities of one or more features and/or one or more peaks in mass spectrometry data.

### Claim 9 (depends on 8)

The system of claim 8, wherein the objects in the training set comprise test data obtained from two classes of patients: 1) a first class of patients exhibiting an adverse health condition and 2) a second class of patients that do not exhibit the adverse health condition, and wherein the classes assigned to the objects correspond to the two classes of patients.

### Claim 10 (depends on 8)

The system of claim 8, wherein the objects in the training set comprise the data obtained from two classes of patients: 1) a first class of patients that derive benefit from a particular therapy for treatment of a health condition and 2) a second class of patients that either do not derive benefit, or suffer adverse effects from the particular therapy for treatment of the medical condition, and wherein the classes assigned to the objects correspond to the two classes of patients.

### Claim 11 (depends on 10)

The system of claim 10, wherein the health condition comprises cancer.

### Claim 12 (depends on 11)

The system of claim 11, wherein the cancer comprises non-small cell lung cancer.

### Claim 13

A method for classifying a test object, comprising the steps of:
accessing a machine-readable storage device storing a training set comprising a number of objects, each of the objects being assigned as a member of a class, collectively the objects in the training set comprising members of at least two classes; and
estimating the probability of an object as being a member of each class in the training set, the estimating being performed with reference to the class assignments of the objects in the training set which are neighbors to the test object within a defined region within the training set, and
classifying the test object as belonging to the class with the highest estimated probability;
wherein the training set comprises objects being members of two classes, class 1 and class 2, wherein classes 1 and 2 have numbers of objects N1 and N2, respectively, in the training set, and wherein the probability P(class 1) of the test instance being a member of Class 1 is estimated according to the relation:
P

(
class


1
)
=
k
1
+
1
k
1
+
k
2
+
2
·
F
1
2

(
1
,
k
2
+
1
;
k
1
+
k
2
+
3
;
1
-
N
1
N
2
)
.
where 2F1 (1, k2+1; k1+k2+3; 1−N1/N2) is a calculation of the hypergeometric function 2F1 and where k1 represents the number of neighbors of class 1 within the defined region and k2 represents the number of neighbors in class 2 within the defined region.

### Claim 14 (depends on 13)

The method of claim 13, further comprising the step of comparing the estimates of probabilities for each of the classes to determine if the difference of estimated probability between the class having the highest probability estimate and the class having the next highest probability estimate is greater than a user-supplied parameter.

### Claim 15 (depends on 14)

The method of claim 14, wherein the user-supplied parameter is 0.1

### Claim 16 (depends on 13)

The method of claim 13, wherein the defined region within the training set comprises a sphere ω centered about the test object, and wherein there are k1 neighbors of class 1 within ω and there are k2 neighbors in class 2 within ω.

### Claim 17 (depends on 16)

The method of claim 16, wherein the size of sphere c is selected such that probability density for classes 1 and 2 are approximately constant within sphere ω.

### Claim 18 (depends on 13)

The method of claim 13, wherein the defined region comprises a set of k objects within the training set which are the k closest objects to the test instance.

### Claim 19 (depends on 13)

The method of claim 13, wherein the objects comprise the intensities of one or more features and/or one or more peaks in mass spectrometry data.

### Claim 20 (depends on 19)

The method of claim 19, wherein the objects in the training set comprise test data obtained from two classes of patients: 1) a first class of patients exhibiting an adverse health condition and 2) a second class of patients that do not share the adverse health condition, and wherein the classes assigned to the objects correspond to the two classes of patients.

### Claim 21 (depends on 19)

The method of claim 19, wherein the objects in the training set comprise the data obtained from two classes of patients: 1) a first class of patients that derive benefit from a particular therapy for treatment of a health condition and 2) a second class of patients that either do not derive benefit, or suffer adverse effects from the particular therapy for treatment of the health condition, and wherein the classes assigned to the objects correspond to the two classes of patients.

### Claim 22 (depends on 21)

The method of claim 21 wherein the health condition comprises cancer.

### Claim 23 (depends on 22)

The method of claim 22, wherein the cancer comprises non-small cell lung cancer.

### Claim 24

A method of classification of a test instance comprising the steps of:
creating and storing in a machine-readable memory a training set comprising a plurality of objects, the objects being assigned a class label, the objects being members of either class 1 or a class 2;
obtaining a test instance to be classified;
classifying the test instance with a computer system configured as a probabilistic classifier by estimating the probability that the test instance is a member of classes 1 and 2,
wherein the probability P(class 1) of the test instance being a member of Class 1 is estimated according to the relation:
P

(
class


1
)
=
k
1
+
1
k
1
+
k
2
+
2
·
F
1
2

(
1
,
k
2
+
1
;
k
1
+
k
2
+
3
;
1
-
N
1
N
2
)
.
where 2F1 (1, k2+1; k1+k2+3; 1−N1/N2) is a calculation of the hypergeometric function 2F1 and where k1 represents the number of neighbors of class 1 within a defined region in the training set and k2 represents the number of neighbors in class 2 within the defined region, N1 is the number of objects in class 1 in the training set, and N2 is the number of objects in class 2 in the training set.

### Claim 25 (depends on 24)

The method of claim 24, further comprising the steps of
conducting mass spectroscopy on a test sample, wherein the test instance comprises a data set representing a spectrum obtained from the mass spectroscopy.

### Claim 26 (depends on 25)

The method of claim 25, wherein the test sample comprises a sample obtained from a human subject.

### Claim 27 (depends on 24)

The method of claim 24, further comprising the steps of estimating the probability that the test instance is a member of classes 1 and 2, and comparing the estimates of probabilities for each of the classes to determine if the difference of estimated probability between the class having the highest probability estimate and the class having the next highest probability estimate is greater than a user-supplied parameter.

### Claim 28

A system for classification of a test object, comprising:
a machine-readable storage device storing a training set comprising a plurality of objects, each of the objects being assigned as a member of a class, collectively the objects in the training set comprising members of at least two classes, including class 1 and class 2; and
a computer system coupled to the machine-readable storage device configured as a probabilistic classifier, the probabilistic classifier estimating the probability of the test object as being a member of a class in the training set, wherein the probabilistic classifier estimates the probability with reference to the class assignments of objects in the training set which are neighbors to the test object within a defined region within the training set; and
wherein classes 1 and 2 have N1 and N2 objects, respectively, in the training set, wherein N1 and N2 are not the same.

### Claim 29 (depends on 28)

The system of claim 28, wherein the probabilistic classifier estimates the probability of the test object as being a member of each of the classes in the training set and classifies the test object as belonging to the class with the highest estimated probability.

### Claim 30 (depends on 29)

The system of claim 29, there are only two classes of objects in the training set, classes 1 and 2, and wherein the probability P(class 1) of the test instance being a member of Class 1 is estimated according to the relation:
P

(
class


1
)
=
k
1
+
1
k
1
+
k
2
+
2
·
F
1
2

(
1
,
k
2
+
1
;
k
1
+
k
2
+
3
;
1
-
N
1
N
2
)
.
where 2F1 (1, k2+1; k1+k2+3; 1-N1/N2) is a calculation of the hypergeometric function 2F1 and where k, represents the number of neighbors of class 1 within a defined region in the training set and k2 represents the number of neighbors in class 2 within the defined region, N1 is the number of objects in class 1 in the training set, and N2 is the number of objects in class 2 in the training set.

### Claim 31 (depends on 29)

The system of claim 29, wherein there are at least three classes in the training set and wherein the probability p(class i) of the test instance being a member of class i in the training set is approximated as being proportional to the expression:
1
N
i

(
k
i
+
1
)
,
while


∑
i
=
1
N


classes

p

(
class


i
)
=
1

### Claim 32 (depends on 29)

The system of claim 29, wherein there are at least three classes in the training set and wherein the ratio of the probabilities of the test instance being a member of class i in the training set and class j in the training set is approximated by the relation:
p

(
class


i
)
p

(
class


j
)
=
(
k
i
+
1
)
(
k
j
+
1
)
·
N
j
N
i
where Nj is the number of objects in class j and Ni is the number of objects in class i.

### Claim 33 (depends on 28)

The system of claim 28, wherein the objects in the training set comprise the intensities of one or more features and/or one or more peaks in mass spectrometry data.

### Claim 34 (depends on 33)

The system of claim 33, wherein the objects in the training set comprise test data obtained from two classes of patients: 1) a first class of patients exhibiting an adverse health condition and 2) a second class of patients that do not exhibit the adverse health condition, and wherein the classes assigned to the objects correspond to the two classes of patients.

### Claim 35 (depends on 33)

The system of claim 33, wherein the objects in the training set comprise the data obtained from two classes of patients: 1) a first class of patients that derive benefit from a particular therapy for treatment of a health condition and 2) a second class of patients that either do not derive benefit, or suffer adverse effects from the particular therapy for treatment of the medical condition, and wherein the classes assigned to the objects correspond to the two classes of patients.

### Claim 36 (depends on 35)

The system of claim 35, wherein the health condition comprises cancer.

### Claim 37 (depends on 36)

The system of claim 36, wherein the cancer comprises non-small cell lung cancer.
