System and method for enhancing an object of interest in noisy medical images

1. A real time medical viewing system for enhancing an object of interest in a sequence of medical images, comprising:

alarm detection means ( 20 ) for detecting a set of alarms (A t ) of said object of interest in a medical image (I t ) at time t,

temporal tracking means ( 30 ) comprising:

track creation sub-means ( 70 ) for iteratively creating a set of tracks (T t ) at time t by associating alarms of said set of alarms (A t ) with tracks (T t−1 ) at time t−1, a track (T j t−1 ) at time t−1 comprising a predicted alarm (Ap j t ) at time t and an association of an alarm (A i t ) at time t with said predicted alarm (Ap j t ) at time t forming a track (T j t ) at time t,

decision sub-means ( 80 ) for choosing a track (T j t ) from among said set of tracks (T t ) at time tin accordance with a merit criterion (M i,j t ), the alarm (A i t ) of said track (T j t ) at time t corresponding to a detection of said object of interest in said medical image at time t.

2. A real time medical viewing system as claimed in claim 1 , wherein said track creation means ( 70 ) are able to associate, with an alarm (A i t ) at time t, the track (T j t−1 ) at time t−1 that minimizes a matching error (E i,j t ).

3. A real time medical viewing system as claimed in claim 2 , wherein an alarm at time t is described by a vector of characteristics, and said matching error (E i,j t ) is evaluated on the basis of distances between components of the vectors of characteristics (VC) of said alarm (A i t ) at time t and said predicted alarm (Ap j t ) of said track at time t−1.

4. A real time medical viewing system as claimed in claim 3 , wherein said vector of characteristics (VC) of said alarm (A i t ) comprises a barycenter position (c i t ), a radius (r i t ), and an angle (α i t ), and said matching error (E i,j t ) comprises a barycenter error, a radius error, and an angle error.

5. A real time medical viewing system as claimed in claim 2 , wherein said matching error (E i,j t ) is a fuzzy error (E Fzi,j t ).

6. A real time medical viewing system as claimed in claim 1 , wherein an alarm (A i t ) is given a strength measure (s i t ), and said decision sub-means ( 80 ) are able to calculate said merit criterion (M i,j t ), for an association of said alarm (A i t ) with a track (T j t−1 ) at time t−1, as a combination of a temporally integrated alarm strength measure ( S i t ) and a temporally integrated matching error (Ē i,j t ).

7. A real time medical viewing system as claimed in claim 1 , wherein said temporal tracking means ( 30 ) comprise virtual track creation sub-means ( 110 ) for creating a virtual track (TV j t ) at time t from an orphan track (OT j t ) at time t−1 or from an orphan alarm at time t.

8. A real time medical viewing system as claimed in claim 1 , wherein said temporal tracking means ( 30 ) comprise prediction sub-means ( 100 ) for predicting a track (Tp j t ) at time t from a track (T j t ) at time t−1.

9. A real time medical viewing system as claimed in claim 1 , wherein said predicted alarm (Ap j t ) at time t corresponds to the alarm (A i t−1 ) of said track at time t−1.

10. A real time medical viewing system as claimed in claim 1 , wherein said temporal tracking means ( 30 ) comprise contender penalizing sub-means ( 120 ) for penalizing losing tracks at time t, that originate from the same track at time t−1 as a winning track (Tw j t ) at time t.

11. An image processing method for use in a real-time medical viewing system capable of detecting an object of interest in a sequence of medical images and comprising:

a step ( 20 ) of detecting a set of alarms (A t ) of said objects of interests in a medical image (I t ) at time t,

a step ( 30 ) of temporal tracking comprising the sub-steps of:

iteratively creating ( 70 ) a set of tracks (T t ) at time t by associating alarms of said set of alarms (A t ) with tracks (T t−1 ) at time t−1, a track (T j t−1 ) at time t−1 comprising a predicted alarm (Ap j t ) at time t and an association of an alarm (A i t ) at time t with said predicted alarm (Ap j t ) at time t forming a track (T j t ) at time t,

choosing ( 80 ) a track (T j t ) from among said set of tracks at time tin accordance with a merit criterion (M i,j t ), the alarm (A i t ) of said track at time t corresponding to a detection of said object of interest in said medical image at time t.

12. A computer program product comprising a computer-readable medium having encoded thereon a set of instructions for carrying out on a computer an image processing method comprising:

detecting a set of alarms (A t ) of an object of interest in a medical image (I t ) at a time t;

iteratively creating a set of tracks (T t ) at time t by associating alarms of said set of alarms (A t ) with tracks (T t−1 ) at a time t−1, each track (T j t−1 ) at time t−1 comprising a predicted alarm (Ap j t ) at time t and an association of an alarm (A i t ) at time t with said predicted alarm (Ap j t ) at time t forming a track (T j t ) at time t; and

choosing a track (T j t ) from among said set of tracks (T t ) at time t in accordance with a merit criterion (M i,j t ) and penalizing losing tracks at time t, that originate from the same track at time t−1 as a winning track (Tw j t ) at time t, the alarm (A i t ) of said track (T j t ) at time t corresponding to a detection of said object of interest in said medical image at time t.

13. A medical examination imaging apparatus comprising a viewing system as claimed in one of claims 1 to 9 .

14. The method of claim 11 , wherein said steps are carried out on a computer.