Pneumothorax detection

1. A processor-implemented method for determining a likelihood of pneumothorax of a patient, the method comprising:

assessing a digital image of a chest x-ray of the patient;

applying a standard detection pipeline to the digital image;

applying a confounding factor detector to the digital image; and

applying a high-resolution detection pipeline to the digital image, wherein the high-resolution detection pipeline comprises a deep sulcus detector and an apical pneumothorax detector that operate on one or more patches of the digital image, wherein the deep sulcus detector classifies the one or more patches with regard to whether or not the one or more patches have a deep sulcus pneumothorax present, and wherein the apical pneumothorax detector classifies the one or more patches with regard to whether or not the one or more patches have an apical pneumothorax present.

2. The processor-implemented method according to claim 1 , wherein applying a standard detection pipeline to the digital image comprises:

down sampling the digital image;

applying an artificial intelligence (AI) machine learning (ML) algorithm to identify a region of pleural air from the down sampling; and

applying an AI ML algorithm to generate a probability of the down sampling containing pneumothorax.

3. The processor-implemented method according to claim 1 , wherein applying a confounding factor detector to the digital image comprises:

applying an artificial intelligence (AI) machine learning (ML) algorithm to identify a foreign object on the digital image;

applying an AI ML algorithm to identify emphysema or bullae on the digital image;

applying an AI ML algorithm to identify a chest tube on the digital image.

4. The processor-implemented method according to claim 1 , wherein applying a high-resolution detection pipeline to the digital image comprises:

applying an artificial intelligence (AI) machine learning (ML) algorithm to segment the digital image into one or more patches each corresponding to one or more areas of a lung of the patient;

extracting each of the one or more areas of the lung;

applying an AI ML algorithm to correlate each of the one or more patches to a specific area of the lung;

applying an AI ML algorithm to extract a first subset of the one or more patches which correspond to a base of the lung;

applying an AI ML algorithm to identify a deep sulcus in the first subset;

applying an AI ML algorithm to extract a second subset of the one or more patches which correspond to an apex of the lung; and

applying an AI ML algorithm to identify an apical pneumothorax in the second subset.

5. The processor-implemented method according to claim 1 , further comprising:

applying an artificial intelligence (AI) machine learning (ML) algorithm using the assessment of the digital image, using output from the standard detection pipeline, using output from the confounding factor detector and using output from the high-resolution detection pipeline to identify a likelihood of pneumothorax of the patient.

6. The processor-implemented method according to claim 5 , further comprising:

applying an AI ML algorithm to classify the pneumothorax.

7. The processor-implemented method according to claim 5 , wherein the confounding factor detector comprises one or more artificial intelligence AI ML algorithms trained to identify features, in the digital image, corresponding to shadows or air pockets that mimic pneumothorax but which are not pneumothorax, and wherein applying the AI ML algorithm to identify a likelihood of pneumothorax of the patient comprises reducing a likelihood of pneumothorax of the patient in response to the confounding factor detector identifying features in the digital image corresponding to shadows or air pockets that mimic pneumothorax but which are not pneumothorax.

8. A computer system for determining a likelihood of pneumothorax of a patient, the computer system comprising:

one or more processors, one or more computer-readable memories, one or more computer-readable tangible storage medium, and program instructions stored on at least one of the one or more tangible storage medium for execution by at least one of the one or more processors via at least one of the one or more memories, wherein the computer system is capable of performing a method comprising:

program instructions to assess a digital image of a chest x-ray of the patient;

program instructions to apply a standard detection pipeline to the digital image;

program instructions to apply a confounding factor detector to the digital image; and

program instructions to apply a high-resolution detection pipeline to the digital image, wherein the high-resolution detection pipeline comprises a deep sulcus detector and an apical pneumothorax detector that operate on one or more patches of the digital image, wherein the deep sulcus detector classifies the one or more patches with regard to whether or not the one or more patches have a deep sulcus pneumothorax present, and wherein the apical pneumothorax detector classifies the one or more patches with regard to whether or not the one or more patches have an apical pneumothorax present.

9. The computer system according to claim 8 , wherein program instructions to apply a standard detection pipeline to the digital image comprises:

program instructions to down sample the digital image;

program instructions to apply an artificial intelligence (AI) machine learning (ML) algorithm to identify a region of pleural air from the down sampling; and

program instructions to apply an AI ML algorithm to generate a probability of the down sampling containing pneumothorax.

10. The computer system according to claim 8 , wherein program instructions to apply a confounding factor detector to the digital image comprises:

program instructions to apply an artificial intelligence (AI) machine learning (ML) algorithm to identify a foreign object on the digital image;

program instructions to apply an AI ML algorithm to identify emphysema or bullae on the digital image;

program instructions to apply an AI ML algorithm to identify a chest tube on the digital image.

11. The computer system according to claim 8 , wherein program instructions to apply a high-resolution detection pipeline to the digital image comprises:

program instructions to apply an artificial intelligence (AI) machine learning (ML) algorithm to segment the digital image into one or more patches each corresponding to one or more areas of a lung of the patient;

program instructions to extract each of the one or more areas of the lung;

program instructions to apply an AI ML algorithm to correlate each of the one or more patches to a specific area of the lung;

program instructions to apply an AI ML algorithm to extract a first subset of the one or more patches which correspond to a base of the lung;

program instructions to apply an AI ML algorithm to identify a deep sulcus in the first subset;

program instructions to apply an AI ML algorithm to extract a second subset of the one or more patches which correspond to an apex of the lung; and

program instructions to apply an AI ML algorithm to identify an apical pneumothorax in the second subset.

12. The computer system according to claim 8 , further comprising:

program instructions to apply an artificial intelligence (AI) machine learning (ML) algorithm using the assessment of the digital image, using output from the standard detection pipeline, using output from the confounding factor detector and using output from the high-resolution detection pipeline to identify a likelihood of pneumothorax of the patient.

13. The computer system according to claim 12 , further comprising:

program instructions to apply an AI ML algorithm to classify the pneumothorax.

14. The computer system according to claim 12 , wherein the confounding factor detector comprises one or more artificial intelligence AI ML algorithms trained to identify features, in the digital image, corresponding to shadows or air pockets that mimic pneumothorax but which are not pneumothorax, and wherein applying the AI ML algorithm to identify a likelihood of pneumothorax of the patient comprises reducing a likelihood of pneumothorax of the patient in response to the confounding factor detector identifying features in the digital image corresponding to shadows or air pockets that mimic pneumothorax but which are not pneumothorax.

15. A computer program product for determining a likelihood of pneumothorax of a patient, the computer program product comprising:

one or more computer-readable tangible storage medium and program instructions stored on at least one of the one or more tangible storage medium, the program instructions executable by a processor, the program instructions comprising:

program instructions to assess a digital image of a chest x-ray of the patient;

program instructions to apply a standard detection pipeline to the digital image;

program instructions to apply a confounding factor detector to the digital image; and

program instructions to apply a high-resolution detection pipeline to the digital image, wherein the high-resolution detection pipeline comprises a deep sulcus detector and an apical pneumothorax detector that operate on one or more patches of the digital image, wherein the deep sulcus detector classifies the one or more patches with regard to whether or not the one or more patches have a deep sulcus pneumothorax present, and wherein the apical pneumothorax detector classifies the one or more patches with regard to whether or not the one or more patches have an apical pneumothorax present.

16. The computer program product according to claim 15 , wherein program instructions to apply a standard detection pipeline to the digital image comprises:

program instructions to down sample the digital image;

program instructions to apply an artificial intelligence (AI) machine learning (ML) algorithm to identify a region of pleural air from the down sampling; and

program instructions to apply an AI ML algorithm to generate a probably probability of the down sampling containing pneumothorax.

17. The computer program product according to claim 15 , wherein program instructions to apply a confounding factor detector to the digital image comprises:

program instructions to apply an artificial intelligence (AI) machine learning (ML) algorithm to identify a foreign object on the digital image;

program instructions to apply an AI ML algorithm to identify emphysema or bullae on the digital image;

program instructions to apply an AI ML algorithm to identify a chest tube on the digital image.

18. The computer program product according to claim 15 , wherein program instructions to apply a high-resolution detection pipeline to the digital image comprises:

program instructions to apply an artificial intelligence (AI) machine learning (ML) algorithm to segment the digital image into one or more patches each corresponding to one or more areas of a lung of the patient;

program instructions to extract each of the one or more areas of the lung;

program instructions to apply an AI ML algorithm to correlate each of the one or more patches to a specific area of the lung;

program instructions to apply an AI ML algorithm to extract a first subset of the one or more patches which correspond to a base of the lung;

program instructions to apply an AI ML algorithm to identify a deep sulcus in the first subset;

program instructions to apply an AI ML algorithm to extract a second subset of the one or more patches which correspond to an apex of the lung; and

program instructions to apply an AI ML algorithm to identify an apical pneumothorax in the second subset.

19. The computer program product according to claim 15 , further comprising:

program instructions to apply an artificial intelligence (AI) machine learning (ML) algorithm using the assessment of the digital image, using output from the standard detection pipeline, using output from the confounding factor detector and using output from the high-resolution detection pipeline to identify a likelihood of pneumothorax of the patient; and

program instructions to apply an AI ML algorithm to classify the pneumothorax.

20. The computer program product according to claim 19 , wherein the confounding factor detector comprises one or more artificial intelligence AI ML algorithms trained to identify features, in the digital image, corresponding to shadows or air pockets that mimic pneumothorax but which are not pneumothorax, and wherein applying the AI ML algorithm to identify a likelihood of pneumothorax of the patient comprises reducing a likelihood of pneumothorax of the patient in response to the confounding factor detector identifying features in the digital image corresponding to shadows or air pockets that mimic pneumothorax but which are not pneumothorax.