Deep learning modeling using health screening images

1. A method for improving health screening using artificial intelligence (AI), comprising:

receiving a user image by an AI system, wherein the AI system includes:

a first AI model trained using labeled training images from prior health screenings to predict a medical condition, wherein the first AI model generates a first set of feature vectors based on the user image;

a second AI model trained using labeled training images from current health screenings to classify images, wherein the second AI model generates a second set of feature vectors based on the user image;

a feature selection algorithm that selects feature vectors from at least one of the first set of feature vectors or the second set of feature vectors to generate selected feature vectors; and

a classification AI model that processes the selected feature vectors to generate a classification output that comprises a prediction score indicative of a likelihood of the medical condition; and

receiving the classification output from the AI system.

2. The method of claim 1 , wherein the second AI model is trained by, after the first AI model is trained to predict the medical condition, using transfer learning to apply the weights from the first AI model to initialize the second AI model.

3. The method of claim 1 , wherein the images from the current health screenings are images from one or more first users who received positive diagnoses within a first time period from the current health screenings being taken.

4. The method of claim 1 , wherein the images from the prior health screenings are images from one or more second users who received negative screenings for at least a second time period following the prior health screenings being taken and later received positive diagnoses.

5. The method of claim 1 , wherein the user image is a digital x-ray.

6. The method of claim 1 , wherein the first AI model is a convolutional neural network.

7. The method of claim 1 , wherein the second AI model is a convolutional neural network.

8. A method for improving breast cancer screening using artificial intelligence (AI), comprising:

receiving a user image by an AI system, the AI system including:

a first AI model trained using labeled training images including images from prior mammograms to predict cancer, wherein the first AI model generates a first set of feature vectors based on the user image;

a second AI model trained using labeled training images including images from current mammograms to classify mammogram images, wherein the second AI model generates a second set of feature vectors based on the user image, and wherein the second AI model is initialized using the weights of the first AI model using transfer learning;

a feature selection algorithm that selects feature vectors from at least one of the first set of feature vectors or the second set of feature vectors to generate selected feature vectors; and

a classification AI model that processes the selected feature vectors to generate a classification output that comprises a prediction score indicative of a current breast cancer diagnosis or a likelihood of the user to develop breast cancer in the future; and

receiving the classification output.

9. The method of claim 8 , wherein the user image is a digital mammogram.

10. The method of claim 9 , wherein the current mammograms are one or more medical images from one or more first users who received positive cancer diagnoses within a first time period from the medical images being taken.

11. The method of claim 10 , wherein the prior mammograms are one or more further medical images from one or more second users who received negative cancer screenings for at least a second time period following the medical image being taken and later received positive diagnoses.

12. The method of claim 8 , wherein the first AI model and the second AI model are convolutional neural networks.

13. A method for training an artificial intelligence (AI) system for improved health screening, the method comprising:

receiving training images, wherein each of the training images is linked in a group, and wherein the group includes training images of one training user;

registering grouped training images, wherein registering grouped training images involves aligning a first training image so that the same region of 2D space in the first training image in the group can be compared to a same region of 2D space in a second training image in the group;

obtaining a first set of feature vectors from a first AI model trained to predict cancer using training images from prior mammograms;

obtaining a second set of feature vectors from a second AI model trained to classify images using training images from current mammograms;

applying a feature selection algorithm to the first set of feature vectors and the second set of feature vectors;

inputting feature vectors selected using the feature selection algorithm into a classification AI model; and

providing an output, wherein the output includes a first prediction score indicating a likelihood of future breast cancer development and a second classification score indicating likelihood of current cancer.

14. The method of claim 13 , wherein the current mammograms are one or more medical images from one or more first users who received positive diagnoses within a first time period from the medical images being taken.

15. The method of claim 14 , wherein the prior mammograms are one or more further medical images from one or more second users who received negative screenings for a second time period following the medical images being taken and later received positive diagnoses.

16. The method of claim 13 , wherein the second AI model is trained by, after the first AI model is trained to predict cancer, using transfer learning to apply the weights from the first AI model to initialize the second AI model.

17. The method of claim 13 , wherein the first AI model and the second AI model are convolutional neural networks.

18. The method of claim 13 , wherein the classification AI model is a Long Short-Term Memory.

19. A system comprising:

a memory; and

a processor in communication with the memory, the processor being configured to perform operations comprising:

receiving training images, wherein each of the training images is linked in a group, and wherein the group includes training images of one training user;

obtaining a first set of feature vectors from a first AI model trained to predict cancer using training images from prior mammograms;

obtaining a second set of feature vectors from a second AI model trained to classify images using training images from current mammograms;

applying a feature selection algorithm to the first set of feature vectors and the second set of feature vectors;

inputting feature vectors selected using the feature selection algorithm into a classification AI model; and

providing an output, wherein the output includes a first prediction score indicating a likelihood of future breast cancer development and a second classification score indicating likelihood of current cancer.

20. The system of claim 19 , wherein the second AI model is trained by, after the first AI model is trained to classify images, using transfer learning to apply the weights from the first AI model to initialize the second AI model.

21. The system of claim 19 , wherein the current mammograms are one or more medical images from one or more first users who received positive cancer diagnoses within a first time period from the medical images being taken.

22. The system of claim 19 , wherein the prior mammograms are one or more further medical images from one or more second users who received negative cancer screenings for at least a second time period following the medical image being taken and later received positive diagnoses.

23. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform operations, the operations comprising:

receiving training images, wherein each of the training images is linked in a group, and wherein the group includes training images of one training user;

obtaining a first set of feature vectors from a first AI model trained to predict cancer using training images from prior mammograms;

obtaining a second set of feature vectors from a second AI model trained to classify images using training images from current mammograms;

applying a feature selection algorithm to the first set of feature vectors and the second set of feature vectors;

inputting feature vectors selected using the feature selection algorithm into a classification AI model; and

providing an output, wherein the output includes a first prediction score indicating a likelihood of future breast cancer development and a second classification score indicating likelihood of current cancer.

24. The computer program product of claim 23 , wherein the current mammograms are one or more medical images from one or more first users who received positive cancer diagnoses within a first time period from the medical images being taken.

25. The computer program product of claim 23 , wherein the prior mammograms are one or more further medical images from one or more second users who received negative cancer screenings for at least a second time period following the medical image being taken and later received positive diagnoses.