Granted Patent B2

US 12,315,213 · App. 17/818,791 · Granted May 27, 2025

Inventors: Hong Yang (Beijing, CN); Chunqi Wang (Beijing, CN); Zhen Zhou (Beijing, CN)

Assignee: CANON MEDICAL SYSTEMS CORPORATION

G06V10/42G06T7/0014G06T7/74G06V10/764G06V10/778G06T2207/20076G06T2207/20081G06T2207/30004G06V2201/03

Abstract

Claims (51)

1. A medical image processing apparatus for detecting a plurality of feature points in a medical image, the medical image processing apparatus comprising:

processing circuitry configured to

detect, for each of a plurality of target points corresponding to the feature points, a reference point having a spatial correlation with the target point in the medical image,

generate a plurality of candidate points corresponding to the target point for each of the target points by using a detection model,

score, for each of the target points, a position score for the candidate points based on a position feature indicating a spatial position relationship between the target point and the reference point in the medical image, and select, for each of the target points, a candidate point having the position score satisfying a preset condition,

increase a detection rate of the feature points in the medical image by scoring, for each of a plurality of candidate point combinations obtained by combining the candidate points among the target points, a structural score for the candidate point combination based on a structural feature indicating a spatial structural relationship between the target points in the medical image, and select, for each of the candidate point combinations, a candidate point combination having the structural score satisfying a preset condition, and

output, via a display, the feature points in the medical image based on the selected candidate point and candidate point combination.

2. The medical image processing apparatus according to claim 1 , wherein the processing circuitry is further configured to

select a plurality of candidate points having the position score satisfying the condition from the generated candidate points,

select a candidate point combination having the structural score satisfying the condition from a plurality of candidate point combinations obtained by combining the selected candidate points, and

output a plurality of candidate points included in the selected candidate point combination as the feature points.

3. The medical image processing apparatus according to claim 1 , wherein the processing circuitry is further configured to

select a candidate point combination having the structural score satisfying the condition from a plurality of candidate point combinations obtained by combining the generated candidate points,

select a plurality of candidate points having the position score satisfying the condition from a plurality of candidate points included in the selected candidate point combination, and

output the selected candidate points as the feature points.

4. The medical image processing apparatus according to claim 1 , wherein the processing circuitry is further configured to

calculate a position feature distribution indicating a probability distribution of the position feature and a structural feature distribution indicating a probability distribution of the structural feature, based on a learning data set of the detection model,

score the position score for the candidate points based on the position feature and the position feature distribution corresponding to the position feature, and

score the structural score for the candidate point combination based on the structural feature and the structural feature distribution corresponding to the structural feature.

5. The medical image processing apparatus according to claim 1 , wherein the processing circuitry is further configured to

calculate a combined position feature distribution indicating a probability distribution of a combined position feature in which a plurality of spatial position relationships are combined as the position feature and a combined structural feature distribution indicating a probability distribution of a combined structural feature in which a plurality of spatial structural relationships are combined as the structural feature, based on a learning data set of the detection model,

score the position score for the candidate points based on the combined position feature and the combined position feature distribution corresponding to the combined position feature, and

score the structural score for the candidate point combination based on the combined structural feature and the combined structural feature distribution corresponding to the combined structural feature.

6. The medical image processing apparatus according to claim 1 , wherein the processing circuitry is further configured to

learn a position feature classifier based on the position feature and a spatial position relationship between an abnormal point and the reference point in the medical image and learns a structural feature classifier based on the structural feature, a spatial structural relationship between a plurality of abnormal points in the medical image, and a spatial structural relationship between the target points and the abnormal points in the medical image,

score the position score for the candidate points based on the position feature classifier, and

score the structural score for the candidate point combination based on the structural feature classifier.

7. The medical image processing apparatus according to claim 1 , wherein the processing circuitry is further configured to

generate a visual feature of each of the candidate points for each of the target points by using the detection model,

score an individual total score as the position score for the candidate points based on the position feature and the visual feature, and

score a combined total score as the structural score for the candidate point combination based on the structural feature and the individual total score.

8. The medical image processing apparatus according to claim 4 , wherein the processing circuitry is further configured to update the position feature distribution and the structural feature distribution by online learning based on the position score and the structural score.

9. The medical image processing apparatus according to claim 6 , wherein the processing circuitry is further configured to update position feature classifier and the structural feature classifier by online learning based on the position score and the structural score.

10. A medical image processing method for detecting a plurality of feature points in a medical image, the medical image processing method comprising:

detecting, for each of a plurality of target points corresponding to the feature points, a reference point having a spatial correlation with the target point in the medical image;

generating a plurality of candidate points corresponding to the target point for each of the target points by using a detection model;

scoring, for each of the target points, a position score for the candidate points based on a position feature indicating a spatial position relationship between the target point and the reference point in the medical image, and selecting, for each of the target points, a candidate point having the position score satisfying a preset condition;

increasing a detection rate of the feature points in the medical image by scoring, for each of a plurality of candidate point combinations obtained by combining the candidate points among the target points, a structural score for the candidate point combination based on a structural feature indicating a spatial structural relationship between the target points in the medical image, and selecting, for each of the candidate point combinations, a candidate point combination having the structural score satisfying a preset condition; and

outputting, via a display, the feature points in the medical image based on the selected candidate point and candidate point combination.

11. The medical image processing method according to claim 10 , wherein

the selecting the candidate point includes selecting a plurality of candidate points having the position score satisfying the condition from the generated candidate points,

the selecting the candidate point combination includes selecting a candidate point combination having the structural score satisfying the condition from the selected candidate points, and

the outputting the feature points includes outputting a plurality of candidate points included in the selected candidate point combination as the feature points.

12. The medical image processing method according to claim 10 , wherein

the selecting the candidate point combination includes selecting a candidate point combination having the structural score satisfying the condition from the generated candidate points,

the selecting the candidate point includes selecting a plurality of candidate points having the position score satisfying the condition from a plurality of candidate points included in the selected candidate point combination, and

the outputting the feature points includes outputting the selected candidate points as the feature points.

13. The medical image processing method according to claim 10 , further comprising:

calculating a position feature distribution indicating a probability distribution of the position feature and a structural feature distribution indicating a probability distribution of the structural feature, based on a learning data set of the detection model, wherein

the selecting the candidate point includes scoring the position score for the candidate points based on the position feature and the position feature distribution corresponding to the position feature, and

the selecting the candidate point combination includes scoring the structural score for the candidate point combination based on the structural feature and the structural feature distribution corresponding to the structural feature.

Assignments (1)

Priority Claims (2)

Continuity (1)

References Cited (16)