Surgical video retrieval based on preoperative images
A system includes a display, and a database including surgical videos, images of organs in a human body obtained from a medical imaging device, and images of disease in a human body obtained from the medical imaging device. A controller including a processor is coupled to memory, the database, and the display, and the memory stores information that when executed by the processor causes the system to perform operations. For example, the processor may determine first organ information from the images of the organs, and first disease information from the images of the disease. The processor my calculate a similarity score between the first organ information and the first disease information and second disease information and second organ information indexed to the surgical videos. The processor selects one or more of the surgical videos based on the similarity score, and displays the surgical videos on the display.
1. A system, comprising:
a controller including a processor coupled to memory, wherein the memory stores instructions that when executed by the processor causes the system to perform operations including:
receiving via a user interface a search query including an input medical image and weighted parameters specified by a user of the system;
receiving a plurality of first medical images different from the input medical image;
generating respective feature vectors describing the plurality of first medical images using a machine learning algorithm configured to receive pixel-level information of the plurality of first medical images included in the memory;
weighting positionally common elements included in the respective feature vectors associated with the plurality of first medical images and a corresponding feature vector associated with the input medical image based on the weighted parameters;
calculating similarity scores for the plurality of first medical images relative to the input medical image, wherein the similarity scores correspond to a comparison between the respective feature vectors associated with the plurality of first medical images and the corresponding feature vector associated with the input medical image after the weighting;
filtering the plurality of first medical images based on the similarity scores, wherein the filtering includes filtering out a first portion of the plurality of first medical images outside a similarity score threshold range for relevancy from the plurality of first medical images by selecting for display a second portion of the plurality of first medical images within the similarity score threshold range for relevancy; and
outputting via the user interface a relative rank for the second portion of the plurality of first medical images based on the similarity scores after the filtering.
2. The system of claim 1 , wherein the memory stores additional instructions that when executed by the processor causes the system to perform further operations including:
determining, with the processor, first organ information of a first organ and first disease information of a first disease included in each of the plurality of first medical images, wherein the first organ information includes at least one of a shape of the first organ, a size of the first organ, or a location of the first organ, and wherein the first disease information includes at least one of a shape of the first disease, a location of the first disease, a size of the first disease, or a stage of the first disease, and wherein the respective feature vectors further includes the first organ information and the first disease information,
wherein the corresponding feature vector further includes second disease information and second organ information.
3. The system of claim 2 , wherein the determining the first organ information includes segmenting the first organ for each of the plurality of first medical images using the machine learning algorithm.
4. The system of claim 1 , wherein the plurality of first medical images correspond to images obtained from a medical imaging device, the medical imaging device including one of a computed tomography (CT) scan machine, a magnetic resonance imaging (MRI) machine, an ultrasound machine, a positron emission tomography (PET) machine, or an X-ray machine.
5. The system of claim 1 , wherein the plurality of first medical images is included in a medical video database, and wherein the input medical image is included in a medical video.
6. The system of claim 1 , wherein the machine learning algorithm is trained via a triplet loss method for determining the similarity scores.
7. The system of claim 1 , wherein the machine learning algorithm includes a plurality of deep learning models, each trained to recognize one or more different types of diseases, and wherein at least one of the plurality of deep learning models is trained to recognize a first disease associated with first disease information.
8. The system of claim 1 , wherein the calculating the similarity scores includes using metadata including at least one of patient age, patient body mass index, type of disease, patient gender, or patient preexisting conditions.
9. The system of claim 1 , wherein the input medical image corresponds to one of a computed tomography (CT) image, a magnetic resonance imaging (MRI) image, ultrasound image, positron emission tomography image, or an X-ray image.
10. The system of claim 1 , wherein the machine learning algorithm is configured to receive inputs for each pixel included in the plurality of first medical images including x, y, or z position and one or more color values for a given image included in the plurality of first medical images.
11. The system of claim 1 , wherein at least one element included in the feature vectors have no tangible descriptors.
12. The system of claim 1 , wherein generating the respective feature vectors includes inputting the plurality of first medical images into the machine learning algorithm to determine positions of the plurality of first medical images within a feature space, wherein the respective feature vectors each include a corresponding one of the positions, and
wherein the calculating the similarity score includes determining distances from the positions of the plurality of first medical images within the feature space to a corresponding position within the feature space of the input medical image, and wherein the similarity scores are based, at least in part, on the distances.
13. A method, comprising:
receiving a search query including an input medical image and weighted parameters;
receiving, with a processor coupled to memory, a plurality of first medical images different from the input medical image;
generating, with the processor, respective feature vectors describing the plurality of first medical images using a machine learning algorithm configured to receive pixel-level information of the plurality of first medical images included in the memory,
weighting positionally common elements included in the respective feature vectors associated with the plurality of first medical images and a corresponding feature vector associated with the input medical image based on the weighted parameters;
calculating, with the processor, similarity scores for the plurality of first medical images relative to the input medical image, wherein the similarity scores correspond to a comparison between the respective feature vectors associated with the plurality of first medical images and the corresponding feature vector associated with the input medical image after the weighting;
filtering the plurality of first medical images based on the similarity scores, wherein the filtering includes filtering out a first portion of the plurality of first medical images outside a similarity score threshold range for relevancy from the plurality of first medical images by selecting for display a second portion of the plurality of first medical images within the similarity score threshold range for relevancy; and
outputting a relative rank for the second portion of the plurality of first medical images based on the similarity scores after the filtering.
14. The method of claim 13 , further comprising:
determining, with the processor, first organ information of a first organ and first disease information of a first disease included in each of the plurality of first medical images, wherein the first organ information includes at least one of a shape of the first organ, a size of the first organ, or a location of the first organ, and wherein the first disease information includes at least one of a shape of the first disease, a location of the first disease, a size of the first disease, or a stage of the first disease, and wherein the respective feature vectors further includes the first organ information and the first disease information,
wherein the corresponding feature vector further includes second disease information and second organ information.
15. The method of claim 14 , wherein the determining the first organ information includes segmenting the first organ for each of the plurality of first medical images.
16. The method of claim 13 , wherein the machine learning algorithm includes a plurality of deep learning models, each trained to recognize one or more different types of diseases, and wherein at least one of the plurality of deep learning models is trained to recognize a first disease associated with first disease information.
17. The method of claim 13 , wherein the calculating the similarity scores includes using metadata including at least one of patient age, patient body mass index, type of disease, patient gender, or patient preexisting conditions.
18. The method of claim 13 , wherein the plurality of first medical images and the input medical image each correspond to an image obtained from a medical imaging device, the medical imaging device including one of a computed tomography (CT) scan machine, a magnetic resonance imaging (MRI) machine, an ultrasound machine, a positron emission tomography (PET) machine, or an X-ray machine.
19. At least one non-transitory machine-accessible storage medium that provides instructions that, when executed by a machine, will cause the machine to perform operations comprising:
receiving via a user interface a search query including an input medical image and weighted parameters specified by a user of the system;
receiving a plurality of first medical images different from the input medical image;
generating respective feature vectors describing the plurality of first medical images using a machine learning algorithm configured to receive pixel-level information of the plurality of first medical images included in the memory;
weighting positionally common elements included in the respective feature vectors associated with the plurality of first medical images and a corresponding feature vector associated with the input medical image based on the weighted parameters;
calculating similarity scores for the plurality of first medical images relative to the input medical image, wherein the similarity scores correspond to a comparison between the respective feature vectors associated with the plurality of first medical images and the corresponding feature vector associated with the input medical image after the weighting;
filtering the plurality of first medical images based on the similarity scores, wherein the filtering includes filtering out a first portion of the plurality of first medical images outside a similarity score threshold range for relevancy from the plurality of first medical images by selecting for display a second portion of the plurality of first medical images within the similarity score threshold range for relevancy; and
outputting via the user interface a relative rank for the second portion of the plurality of first medical images based on the similarity scores after the filtering.
20. The at least one non-transitory machine-accessible storage medium of claim 16 , that provides additional instructions that, when executed by the machine, will cause the machine to perform further operations comprising:
determining first organ information of a first organ-and first disease information of a first disease included in each of the plurality of first medical images, wherein the first organ information includes at least one of a shape of the first organ, a size of the first organ, or a location of the first organ, and wherein the first disease information includes at least one of a shape of the first disease, a location of the first disease, a size of the first disease, or a stage of the first disease, and wherein the respective feature vectors further includes the first organ information and the first disease information,
wherein the corresponding feature vector further includes second disease information and second organ information.