Patient and consumer data de-identification
The present technology provides solutions for anonymizing patient data and in particular, for anonymizing potentially unique patient-identifying information while retaining clinically relevant information, for example, that can be used to illustrate the results or outcomes of a clinical intervention. In some aspects, a process of the disclosed technology can include steps for receiving patient data, wherein the patient data comprises a patient image corresponding with a patient, receiving treatment data, wherein the treatment data is based on a treatment goal for the patient, automatically determining, using one or more semantic controls, one or more anonymization parameters for the patient image; and generating, based on the one or more anonymization parameters, an anonymized patient image. Systems and machine-readable media are also provided.
1 . A computer-implemented method comprising:
receiving patient data, wherein the patient data comprises a patient image corresponding with a patient;
receiving treatment data, wherein the treatment data is based on a treatment goal for the patient and specifies a clinical intervention to be performed on the patient, including portions of the patient's anatomy to be modified by the clinical intervention;
automatically determining, using one or more semantic controls, one or more anonymization parameters for the patient image based at least in part on the treatment data, wherein the one or more anonymization parameters define first pixel regions of the patient image that are to be altered, wherein the first pixel regions include one or more facial features;
generating, based on altering at least the first pixel regions of the patient image that are defined by the one or more anonymization parameters, an anonymized patient image, wherein the anonymized patient image:
comprises one or more modified facial features corresponding to the one or more facial features, wherein the one or more modified facial features comprise at least one of a modified skin color, a modified image geometry, or a modified prominence of at least one of the one or more facial features; and
preserves one or more clinically relevant regions of the patient image in view of the treatment data, the one or more clinically relevant regions of the patient image comprising second pixel regions of the patient image depicting the portions of the patient's anatomy to be modified by the clinical intervention; and
causing the anonymized patient image to be presented on a display device.
2 . The computer-implemented method of claim 1 , wherein the anonymized patient image preserves artifacts related to at least one of repositioning one or more teeth of the patient or repositioning jaw positions of the patient.
3 . The computer-implemented method of claim 1 , wherein the one or more anonymization parameters are further based on the patient data.
4 . The computer-implemented method of claim 1 , wherein the first pixel regions are to be modified to obfuscate an identity of the patient.
5 . The computer-implemented method of claim 1 , wherein the one or more anonymization parameters for the patient image specify the one or more clinically relevant regions in the patient image that are not to be modified in the patient image.
6 . The computer-implemented method of claim 1 , wherein the one or more anonymization parameters specify one or more image characteristics to be applied to generate the anonymized patient image.
7 . The computer-implemented method of claim 6 , wherein the one or more image characteristics include color parameters, blur filter parameters, geometric parameters, or a combination thereof.
8 . The computer-implemented method of claim 1 , wherein the one or more semantic controls are based on a user input.
9 . The computer-implemented method of claim 1 , wherein the anonymized patient image comprises global changes and changes to specific image regions of the patient image.
10 . The computer-implemented method of claim 9 , wherein the global changes comprise at least one of changes to skin color or changes that alter aspects of a demographic identity of the patient.
11 . The computer-implemented method of claim 1 , wherein generating the anonymized patient image comprises:
determining key points in the patient image that correspond to body landmarks;
generating a clinically relevant sub-image from the patient image based on the key points, the clinically relevant sub-image comprising the one or more clinically relevant regions of the patient image that are not to be modified, wherein the clinically relevant sub-image extends a specified number of pixels beyond the key points;
generating a color image from the patient image, the color image comprising color data for regions of the patient image;
processing the patient image and the color image using a generative machine learning model to output an initial anonymized patient image; and
combining the initial anonymized patient image with the clinically relevant sub-image using at least one of replacement, averaging values, or alpha channel blurring.
12 . The computer-implemented method of claim 11 , wherein the generative machine learning model comprises a generative adversarial network (GAN).
13 . A system, comprising:
at least one memory; and
at least one processor coupled to the at least one memory, the at least one processor configured to:
receive patient data, wherein the patient data comprises a patient image corresponding with a patient;
receive treatment data, wherein the treatment data is based on a treatment goal for the patient and specifies a clinical intervention to be performed on the patient, including portions of the patient's anatomy to be modified by the clinical intervention;
automatically determine, using one or more semantic controls, one or more anonymization parameters for the patient image based at least in part on the treatment data, wherein the one or more anonymization parameters define first pixel regions of the patient image that are to be altered, wherein the first pixel regions include one or more facial features;
generate, based on altering at least the first pixel regions of the patient image defined by the one or more anonymization parameters, an anonymized patient image, wherein the anonymized patient image;
comprises one or more modified facial features corresponding to the one or more facial features, wherein the one or more modified facial features comprise at least one of a modified skin color, a modified image geometry, or a modified prominence of at least one of the one or more facial features; and
preserves one or more clinically relevant regions of the patient image in view of the treatment data, the one or more clinically relevant regions of the patient image comprising second pixel regions of the patient image depicting the portions of the patient's anatomy to be modified by the clinical intervention; and
cause the anonymized patient image to be presented on a display device.
14 . The system of claim 13 , wherein the one or more anonymization parameters are further based on the patient data.
15 . The system of claim 13 , wherein the first pixel regions are to be modified to obfuscate an identity of the patient.
16 . The system of claim 13 , wherein the one or more anonymization parameters for the patient image specify the one or more clinically relevant regions in the patient image that are not to be modified in the patient image.
17 . The system of claim 13 , wherein the one or more anonymization parameters specify one or more image characteristics to be applied to generate the anonymized patient image.
18 . The system of claim 17 , wherein the one or more image characteristics include color parameters, blur filter parameters, geometric parameters, or a combination thereof.
19 . The system of claim 13 , wherein the one or more semantic controls are based on a user input.
20 . A non-transitory computer-readable storage medium comprising at least one instruction for causing a computer or processor to:
receive patient data, wherein the patient data comprises a patient image corresponding with a patient;
receive treatment data, wherein the treatment data is based on a treatment goal for the patient and specifies a clinical intervention to be performed on the patient, including portions of the patient's anatomy to be modified by the clinical intervention;
automatically determine, using one or more semantic controls, one or more anonymization parameters for the patient image based at least in part on the treatment data, wherein the one or more anonymization parameters define first pixel regions of the patient image that are to be altered, wherein the first pixel regions include one or more facial features;
generate, based on altering at least the first pixel regions of the patient image defined by the one or more anonymization parameters, an anonymized patient image, wherein the anonymized patient image:
comprises one or more modified facial features corresponding to the one or more facial features, wherein the one or more modified facial features comprise at least one of a modified skin color, a modified image geometry, or a modified prominence of at least one of the one or more facial features; and
preserves one or more clinically relevant regions of the patient image in view of the treatment data, the one or more clinically relevant regions of the patient image comprising second pixel regions of the patient image depicting the portions of the patient's anatomy to be modified by the clinical intervention; and
cause the anonymized patient image to be presented on a display device.
21 . The non-transitory computer-readable storage medium of claim 20 , wherein the one or more anonymization parameters are further based on the patient data.
22 . The non-transitory computer-readable storage medium of claim 20 , wherein the first pixel regions are to be modified to obfuscate an identity of the patient.
23 . The non-transitory computer-readable storage medium of claim 20 , wherein the one or more anonymization parameters for the patient image specify the one or more clinically relevant regions in the patient image that are not to be modified in the patient image.
24 . The non-transitory computer-readable storage medium of claim 20 , wherein the one or more anonymization parameters specify one or more image characteristics to be applied to generate the anonymized patient image.
25 . The non-transitory computer-readable storage medium of claim 24 , wherein the one or more image characteristics include color parameters, blur filter parameters, geometric parameters, or a combination thereof.
26 . The non-transitory computer-readable storage medium of claim 20 , wherein the one or more semantic controls are based on a user input.