IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0441130
(2012-04-06)
|
등록번호 |
US-8873816
(2014-10-28)
|
발명자
/ 주소 |
|
출원인 / 주소 |
|
대리인 / 주소 |
Pearl Cohen Zedek Latzer Baratz LLP
|
인용정보 |
피인용 횟수 :
7 인용 특허 :
78 |
초록
▼
A system and method may identify pathologies such as red pathologies in in-vivo images. Candidate pathology regions may be identified by identifying red regions. Features indicative of the probability of pathology in candidate regions may be identified. An image score for an image may be identified
A system and method may identify pathologies such as red pathologies in in-vivo images. Candidate pathology regions may be identified by identifying red regions. Features indicative of the probability of pathology in candidate regions may be identified. An image score for an image may be identified based on one or more identified features, the image score indicative of existence in the image of at least one candidate region with high probability of pathology. Calculating an image score may include calculating a candidate score for at least one identified candidate region based on features, the candidate score indicative of the probability of pathology being imaged in said candidate region, where the image score corresponds to the candidate score of the candidate region with the highest probability of pathology in the image.
대표청구항
▼
1. A method for identification of red pathologies in in-vivo images, the method comprising: receiving an image captured in-vivo;identifying candidate pathology regions in the image by identifying red regions in the image;identifying one or more features indicative of the probability of pathology in
1. A method for identification of red pathologies in in-vivo images, the method comprising: receiving an image captured in-vivo;identifying candidate pathology regions in the image by identifying red regions in the image;identifying one or more features indicative of the probability of pathology in said candidate regions by applying on the image a set of analyses, each analysis for identifying a feature indicative of the probability of pathology, wherein said set of analyses comprises at least one analysis for identifying a physical feature causing false identification of apparent red regions in the image, and identification of said feature in positional relation to a candidate region in the image is indicative of low probability of pathology depicted by said candidate, wherein said physical feature is a bubble, and said at least one analysis comprises identifying whiteness features in the perimeter of a candidate region; andcalculating an image score for said image based on said one or more identified features, said image score indicative of existence in the image of at least one candidate region with high probability of pathology, wherein calculating an image score comprises calculating a candidate score for at least one identified candidate region based on said features, said candidate score indicative of the probability of pathology being imaged in said candidate region, wherein said image score corresponds to the candidate score of the candidate region with the highest probability of pathology in the image. 2. The method according to claim 1, wherein said identification of red regions is according to a red criterion value in image area elements, said red criterion value is determined according to location of color of an image area element in a color space relative to pre-calculated border in said color space, said border distinguishing between a red portion and a non-red portion of said color space, wherein the distance of said color from said border corresponds to said red criterion value. 3. The method according to claim 2, wherein said identification of candidate pathology regions comprises selection of identified red regions according to multiple thresholds of red criterion values applied on the image. 4. The method according to claim 2, wherein said set of analyses comprises an analysis for identifying absolute and relative mean red criteria features, said analysis comprising calculating an interior mean red criterion value within a candidate region, a surrounding mean red criterion value in immediate surroundings of said candidate region and an external mean red criterion value in the image outside the candidate region, and comparing the interior mean red criterion value to the surrounding mean red criterion value and to the external mean red criterion value. 5. The method according to claim 2, wherein said set of analyses comprises a calculation of standard deviation of red criterion value within a candidate region. 6. The method according to claim 1, wherein said identification of whiteness features comprises assigning whiteness criterion values to area elements of the image, wherein the whiteness criterion value for an area element of the image is determined according to a color brightness value and differences between color components within the area element of the image. 7. The method according to claim 6, wherein said at least one analysis comprises calculating the median whiteness criterion value and the 95th percentile whiteness criterion value in the perimeter of a candidate region. 8. A method for identification of red pathologies in in-vivo images, the method comprising: receiving an image captured in-vivo;identifying candidate pathology regions in the image by identifying red regions in the image;identifying one or more features indicative of the probability of pathology in said candidate regions by applying on the image a set of analyses, each analysis for identifying a feature indicative of the probability of pathology, wherein identifying one or more features comprises calculating a distance score for at least a first candidate region by estimating a distance between the first candidate region and at least one additional candidate region in the image; andcalculating an image score for said image based on said one or more identified features, said image score indicative of existence in the image of at least one candidate region with high probability of pathology, wherein calculating an image score comprises calculating a candidate score for at least one identified candidate region based on said features, said candidate score indicative of the probability of pathology being imaged in said candidate region, wherein said image score corresponds to the candidate score of the candidate region with the highest probability of pathology in the image. 9. A method for identification of red pathologies in in-vivo images, the method comprising: receiving an image captured in-vivo;identifying candidate pathology regions in the image by identifying red regions in the image;identifying one or more features indicative of the probability of pathology in said candidate regions by applying on the image a set of analyses, each analysis for identifying a feature indicative of the probability of pathology, wherein said set of analyses comprises at least one analysis for identifying a physical feature causing false identification of apparent red regions in the image, and identification of said feature in positional relation to a candidate region in the image is indicative of low probability of pathology depicted by said candidate, wherein said physical feature is a fold in an in vivo tissue, and wherein said at least one analysis comprises identifying correlation of color saturation distribution within the candidate region; andcalculating an image score for said image based on said one or more identified features, said image score indicative of existence in the image of at least one candidate region with high probability of pathology, wherein calculating an image score comprises calculating a candidate score for at least one identified candidate region based on said features, said candidate score indicative of the probability of pathology being imaged in said candidate region, wherein said image score corresponds to the candidate score of the candidate region with the highest probability of pathology in the image. 10. The method according to claim 1, wherein calculating said candidate score is performed by assigning a feature value for each identified feature in relation to said candidate region and calculating a vector of values for said identified features in a multidimensional space, each dimension represents an identified feature, wherein said candidate score is determined according to position of said vector in said multidimensional space relative to pre-calculated border in said multidimensional space, said border distinguishing between pathological portion and non-pathological portion of said space, wherein the distance of said vector from said border corresponds to said candidate score. 11. The method according to claim 1, wherein said set of analyses comprises an analysis of shape of a candidate region for determining the similarity of said shape to an ellipse and the degree of oblongness of said shape. 12. A system for identification of red pathologies in in vivo images, the system comprising: a memory storing in-vivo images; anda processor to: identify candidate pathology regions by identifying red regions in an image captured in vivo;identify one or more features indicative of the probability of pathology in said candidate regions by applying on the image a set of analyses, each analysis for identifying a feature indicative of the probability of pathology, wherein identifying one or more features comprises calculating a distance score for at least a first candidate region by estimating a distance between the first candidate region and at least one additional candidate region in the image;calculate a candidate score for one or more identified candidate region based on said features, said candidate score indicative of the probability of pathology being imaged in said candidate region; andcalculate an image score for said image based on said one or more identified features, said image score indicative of existence in the image of at least one candidate region with high probability of pathology, wherein said image score corresponds to the candidate score of the candidate region with the highest probability of pathology in the image. 13. The system according to claim 12, wherein the processor is to identify a physical feature causing false identification of apparent red regions in the image, and wherein identification of said feature in positional relation to a candidate region in the image is indicative of low probability of pathology of said candidate. 14. The system according to claim 12, wherein the processor is to determine a red criterion value in image area elements, said red criterion value is determined according to location of color of an image area element in a color space relative to pre-calculated border in said color space, said border distinguishing between a red portion and a non-red portion of said color space, wherein the distance of said color from said border corresponds to said red criterion value. 15. The system according to claim 14, wherein the processor is to perform analysis for identifying absolute and relative mean red criteria features, said analysis comprising calculating an interior mean red criterion value within a candidate region, a surrounding mean red criterion value in immediate surroundings of said candidate region and an external mean red criterion value in the image outside the candidate region, and comparing the interior mean red criterion value to the surrounding mean red criterion value and to the external mean red criterion value. 16. The system according to claim 12, wherein the processor is to identify whiteness features in the perimeter of a candidate region. 17. The system according to claim 13, wherein the processor is to identify a tissue fold within a candidate region. 18. The method according to claim 9, wherein said at least one analysis comprises identifying within a candidate region color brightness values below a certain threshold, and said identifying correlation of color saturation distribution within the candidate region is with a predetermined fold template. 19. The method according to claim 1, wherein identifying one or more features comprises calculating a distance score for at least a first candidate region by estimating a distance between the first candidate region and at least one additional candidate region in the image. 20. A system for identification of red pathologies in in vivo images, the system comprising: a memory storing in-vivo images; anda processor to:receive an image captured in-vivo;identify candidate pathology regions in the image by identifying red regions in the image;identify one or more features indicative of the probability of pathology in said candidate regions by applying on the image a set of analyses, each analysis for identifying a feature indicative of the probability of pathology, wherein said set of analyses comprises at least one analysis for identifying a physical feature causing false identification of apparent red regions in the image, and identification of said feature in positional relation to a candidate region in the image is indicative of low probability of pathology depicted by said candidate, wherein said physical feature is a bubble, and said at least one analysis comprises identifying whiteness features in the perimeter of a candidate region; andcalculate an image score for said image based on said one or more identified features, said image score indicative of existence in the image of at least one candidate region with high probability of pathology, wherein calculating an image score comprises calculating a candidate score for at least one identified candidate region based on said features, said candidate score indicative of the probability of pathology being imaged in said candidate region, wherein said image score corresponds to the candidate score of the candidate region with the highest probability of pathology in the image. 21. A system for identification of red pathologies in in vivo images, the system comprising: a memory storing in-vivo images; anda processor to: receive an image captured in-vivo;identify candidate pathology regions in the image by identifying red regions in the image;identify one or more features indicative of the probability of pathology in said candidate regions by applying on the image a set of analyses, each analysis for identifying a feature indicative of the probability of pathology, wherein said set of analyses comprises at least one analysis for identifying a physical feature causing false identification of apparent red regions in the image, and identification of said feature in positional relation to a candidate region in the image is indicative of low probability of pathology depicted by said candidate, wherein said physical feature is a fold in an in vivo tissue, and wherein said at least one analysis comprises identifying correlation of color saturation distribution within the candidate region; andcalculating an image score for said image based on said one or more identified features, said image score indicative of existence in the image of at least one candidate region with high probability of pathology, wherein calculating an image score comprises calculating a candidate score for at least one identified candidate region based on said features, said candidate score indicative of the probability of pathology being imaged in said candidate region, wherein said image score corresponds to the candidate score of the candidate region with the highest probability of pathology in the image.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.