Autosegmentation/autocontouring system and method for use with three-dimensional radiation therapy treatment planning

• US 5,859,891 A
• Filed: 03/07/1997
• Issued: 01/12/1999
• Est. Priority Date: 03/07/1997
• Status: Expired due to Term

• Current Assignee: IMPAC Medical Systems, Inc. (Elekta AB)
• Original Assignee: Computerized Medical Systems, Inc. (Elekta AB)
• Inventors: Hibbard, Lyn
• Primary Examiner(s): Porta, David P.
• Assistant Examiner(s): Bruce, David Vernon
• Application Number:
  US08/813,701
• Time in Patent Office: 676 Days
• Field of Search: 128/922, 600/427, 382/131, 382/132, 378/58, 378/62, 378/65
• US Class Current: 378/62
• CPC Class Codes: A61N 5/103 Treatment planning systems
• CPC Class Codes: G06T 2207/10081 Computed x-ray tomography [CT]
• CPC Class Codes: G06T 2207/20056 Discrete and fast Fourier transform, [DFT, FFT]
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 2207/30061 Lung
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/143 involving probabilistic approaches, e.g. Markov random field [MRF] modelling

Abstract

An autosegmentation/autocontouring method that may be used for quickly and accurately contouring the regions and boundaries around regions for the development of cross-sections that may be linearly disposed for the three-dimensional reconstruction of an image is described and claimed. The autosegmentation/autocontouring method includes at least four steps. The first step is to digitize a CT, MRI, or other suitable image and display it on a display screen. The two-dimensional image on the display screen will include gray-scale representations of the internal organs and tissue masses of the anatomical site through which the cross-section was made. The second step is to select the interior of a ROI and draw a polygon within the boundaries of the cross-sectional view of this ROI. This polygon could also be drawn in the interior of a cancerous mass or other diseased tissue. The third step of the method of the present invention is to expand the polygon in a novel manner by iteratively testing pixels of the image on the display outside of, but adjacent to, the pixels that the polygon currently subtends. Pixels will be added to the polygon if the value of a decision rule function has a predetermined value. The expansion of the polygon is continued until none of the pixels at the perimeter of the polygon can satisfy the decision rule. Once it is found that none of the perimeter pixels satisfy the decision rule, the perimeter of the polygon is considered the boundary of the ROI. And the fourth step is that the boundary of the ROI is computed and a contour is developed based on this boundary. This same process is repeated for other ROIs that the user may select.

Claims

1. An autocontouring method, comprising the steps of:

(a) producing a two-dimensional image of a structure with the structure having at least one object disposed therein and the object is substantially defined by boundary, with an area within the boundary being of a first class of material and an area outside of the boundary being of a second class of material;

(b) forming a polygon in the object such that the polygon is fully within the boundary of the object and the polygon subtends only a first class of material;

(c) testing a discrete portion of an area outside of, but adjacent to, the polygon according to a maximum a posteriori decision rule according to an Expression

space="preserve" listing-type="equation">if g.sub.i (X)>

g.sub.j (X), decide the first class of material, else decide the second class of material,

where,

g_i (X)=discriminant function for a class of material, denoted i;

g_j (X)=discriminant function for a class of material, denoted j;

(d) expanding the polygon to include the discrete portion tested at step (c) if it is determined that the discrete portion is of a first class of material;

(e) repeating steps (c) and (d) until the testing at step (c) determines that at each location outside of, and adjacent to, a polygon expanded according to steps (c) and (d), a discrete portion tested at step (c) is of second material; and

(f) computing a boundary of the object based on a shape of the expanded polygon when it is determined that at each location outside of, and adjacent to, the polygon expanded according to steps (c) and (d), the discrete portion tested at step (c) is of second material.

2. The method as recited in claim 1, wherein g_i (X) is determined according to the Expression ##EQU7## where, X=vector X,

M_i =vector of the mean property,

Σ

_i =matrix of covariances for features in class i,

Σ

_i^-1 =inverse matrix of matrix Σ

_i,

|Σ

_i |=determinant of a covariance matrix,

T=transpose of a resulting vector,

P(i)=a priori probability that class i is observed.

3. The method as recited in claim 1, wherein g_j (X) is determined according to the Expression ##EQU8## where, X=vector X,

M_j =vector of the mean property,

Σ

_j =matrix of covariances for features in class j,

Σ

_j^-1 =inverse matrix of matrix Σ

_j,

|Σ

_j |=determinant of a covariance matrix,

T=transpose of a resulting vector,

P(j)=a priori probability that class j is observed.

4. An autocontouring method, comprising the steps of:

(a) producing a two-dimensional image of a structure with the structure having at least one object disposed therein and the object is substantially defined by boundary, with an area within the boundary being of a first class of material and an area outside of the boundary being of a second class of material;

(b) forming a polygon in the object such that the polygon is fully within the boundary of the object and polygon subtends only a first class of material;

(c) testing an discrete portion of an area outside of, but adjacent to, the polygon according to a maximum a posteriori decision method according to an Expression

space="preserve" listing-type="equation">if P(i|X)>

P(j|X), decide class i, else decide class j

where,

P(i|X)=probability that class i is a correct class for observation X,

P(j|X)=probability that class j is a correct class for observation X,

(d) expanding the polygon to include the discrete portion tested at step (c) if it is determined that the discrete portion is of a first class of material;

(e) repeating steps (c) and (d) until the testing at step (c) determine that at each location outside of, and adjacent to, a polygon expanded according to steps (c) and (d), a discrete portion tested at step (c) is of second material; and

(f) computing a boundary of the object based on a shape of the expanded polygon when it is determined that at each location outside of, and adjacent to, the polygon expanded according to steps (c) and (d), the discrete portion tested at step (c) is of second material.

5. The method as recited in claim 4, wherein P(i|X) is determined according to an Expression ##EQU9## where, P(i|X)=probability of observing class i given vector X,

p(X|i)=probability of observing vector X given class i,

Σ

_k p(X|k)P(k)=sum over all classes k of products of a probability density of data vector X given class k and an a priori probability of observing class k,

P(i)=a priori probability that class i is observed,

P(j)=a priori probability that class j is observed.

6. The method as recited in claim 4, wherein P(j|X) is determined according to an Expression ##EQU10## where, P(j|X)=probability of observing class j given vector X,

p(X|j)=probability of observing vector X given class j,

Σ

_k p(X|k)P(k)=sum over all classes k of products of a probability density of data vector X given class k and an a priori probability of observing class k,

P(j)=a priori probability that class j is observed.