System, user interface and method for interactive negative explanation of machine-learning localization models in health care applications

1. A method for assessing a machine learning model providing a prediction as to the disease state of a patient from a 2D or 3D image, comprising the steps of:

a) presenting an image with a risk score or classification associated with the prediction, wherein the image is further augmented with highlighting to indicate one or more regions in the image which affected the prediction produced by the machine learning model;

b) providing a user interface tool for highlighting one or more different regions of the image,

c) receiving a user input highlighting one or more different regions of the image;

d) subjecting the highlighted one or more different regions to inference by the machine learning model; and

e) presenting the results of the inference on the one or more different regions to the user via the display.

2. The method of claim 1 , further comprising the steps of receiving user input highlighting still further different regions of the image and performing steps d) and e) for such still further different regions.

3. The method of claim 1 , wherein the prediction is associated with a risk score, and wherein step e) comprises the step of generating a new risk score.

4. The method of claim 1 , wherein the machine learning model comprises a convolutional neural network trained to recognize the presence of cancer cells in the image.

5. The method of claim 1 , wherein the image comprises a magnified digital image of a tissue specimen.

6. The method of claim 1 , wherein the image comprises a 2D radiological image.

7. The method of claim 6 , wherein the image is selected from the group of images consisting of a mammogram and a chest X-ray.

8. The method of claim 1 , wherein the image comprises a 3D radiological image obtained from either nuclear magnetic resonance (NMR) or computed tomography (CT).

9. The method of claim 1 , wherein the tool comprises a mouse associated with the display.

10. The method of claim 1 , wherein the display is touch sensitive and wherein the tool comprises graphics processing which record positions on the display which are touched directly or indirectly by the user and translate such positions to locations within the image.

11. A workstation configured to assess a machine learning model providing a prediction as to the disease state of a patient from a 2D or 3D image of the patient or a sample obtained therefrom, wherein the machine learning model produces a prediction of the disease state from the image, wherein the workstation comprises:

a) display for displaying the image of the patient or a sample obtained therefrom along with a risk score or classification associated with the prediction, wherein the image is further augmented with highlighting to indicate one or more regions in the image which affected the prediction produced by the machine learning model;

b) a user interface tool by which the user may highlight on the display one or more different regions of the image, wherein the user invokes the tool to thereby highlight the one or more different regions;

wherein the display is further configured to present the results of inference performed by the machine learning model on the one or more different regions highlighted by the user.

12. The workstation of claim 11 , wherein the workstation further comprises a processing unit in communication with the display and wherein the machine learning model is implemented in the processing unit.

13. The workstation of claim 11 , wherein the machine learning model comprises a convolutional neural network trained to recognize the presence of cancer in the image.

14. The workstation of claim 11 , wherein the image comprises a magnified digital image of a tissue specimen.

15. The workstation of claim 11 , wherein the image comprises a 2D radiological image.

16. The workstation of claim 15 , wherein the image is selected from the group of images consisting of a mammogram and a chest X-ray.

17. The workstation of claim 11 , wherein the image comprises a 3D radiological image obtained from either nuclear magnetic resonance or computed tomography.

18. The workstation of claim 11 , wherein the tool comprises a mouse associated with the display.

19. The workstation of claim 11 , wherein the display is touch sensitive and wherein the tool comprises graphics processing which record positions on the display which are touched directly or indirectly by the user and translate such positions to locations within the image.

20. An interface for a workstation configured to assess a machine learning model providing a prediction as to the disease state of a patient from a 2D or 3D image of the patient or a sample obtained therefrom, wherein the machine learning model produces a prediction of the disease state from the image, wherein the interface comprises:

a display for displaying the image of the patient or a sample obtained therefrom along with a risk score or classification associated with the prediction, wherein the image is further augmented with highlighting to indicate one or more regions in the image which affected the prediction produced by the machine learning model; and

a user interface tool by which the user may highlight on the display one or more different regions of the image which the user deems to be suspicious for the disease state, wherein the user invokes the tools to thereby highlight the one or more different regions.