Device for non-invasive detection of skin problems associated with diabetes mellitus
A medical diagnostic apparatus includes a controller; a user interface; a platform having a horizontal surface; and at least one visible light image sensor positioned below the horizontal surface that is capable of producing a diagnostic visible light image of a bottom portion of a foot or feet positioned on the horizontal surface. At least a portion of the horizontal surface is transparent to visible light. The apparatus may further include at least one infrared image sensor positioned below the horizontal surface that is capable of producing a thermal image of the first target area. The apparatus may further include sensors positioned above the platform capable of producing diagnostic visible light images or thermal images of a top portion of the user's foot or feet. The apparatus may further include a weight measurement system coupled to the platform.
1. A method for automated diagnosis of a diabetic foot condition, the method comprising:
capturing, by one or more image capture devices of a medical diagnostic apparatus, optical image data of a target area of a foot;
collecting, by a touch sensitivity testing device of the medical diagnostic apparatus, physical touch sensitivity data for the target area of the foot;
transmitting, by the medical diagnostic apparatus, the optical image data and the physical touch sensitivity data to an analysis engine, wherein the analysis engine comprises a neural network classifier pipeline; and
outputting, by the analysis engine, one or more indications of a diabetic foot condition,
wherein the neural network classifier pipeline of the analysis engine automatically performs classification steps on the optical image data to generate the one or more indications of a diabetic foot condition, and wherein the classification steps performed by the neural network classifier pipeline include:
detecting presence, position, and orientation of the foot;
detecting presence and location of a skin abnormality on the foot; and
classifying the detected skin abnormality as a diabetic foot ulcer.
2. The method of claim 1 further comprising checking the optical image data for one or more of image quality, lighting conditions, or body positioning.
3. The method of claim 1 further comprising, prior to collecting the physical touch sensitivity data, confirming the position of the foot based at least in part on the optical image data.
4. The method of claim 1 , wherein the touch sensitivity testing device comprises a servo-actuated monofilament probe.
5. The method of claim 1 , wherein the neural network classifier pipeline of the analysis engine comprises a deep convolutional neural network.
6. The method of claim 1 further comprising outputting, by the analysis engine, a binary classification for medical follow-up in a case where the analysis engine determines that the neural network classifier pipeline cannot accurately classify the detected skin abnormality based on the optical image data.
7. A non-transitory computer-readable medium having stored therein instructions configured to cause one or more computing devices to cause steps to be performed for automated diagnosis of a diabetic foot condition, the steps comprising:
capturing, by one or more image capture devices of a medical diagnostic apparatus, optical image data of a target area of a foot;
collecting, by a touch sensitivity testing device of the medical diagnostic apparatus, physical touch sensitivity data for the target area of the foot;
transmitting, by the medical diagnostic apparatus, the optical image data and the physical touch sensitivity data to an analysis engine, wherein the analysis engine comprises a neural network classifier pipeline; and
outputting, by the analysis engine, one or more indications of a diabetic foot condition,
wherein the neural network classifier pipeline of the analysis engine automatically performs classification steps on the optical image data to generate the one or more indications of a diabetic foot condition, and wherein the classification steps performed by the neural network classifier pipeline include:
detecting presence, position, and orientation of the foot;
detecting presence and location of a skin abnormality on the foot; and
classifying the detected skin abnormality as a diabetic foot ulcer.
8. The non-transitory computer-readable medium of claim 7 , the steps further comprising checking the optical image data for one or more of image quality, lighting conditions, or body positioning.
9. The non-transitory computer-readable medium of claim 7 , the steps further comprising, prior to collecting the physical touch sensitivity data, confirming the position of the foot based at least in part on the optical image data.
10. The non-transitory computer-readable medium of claim 7 , wherein the touch sensitivity testing device comprises a servo-actuated monofilament probe.
11. The non-transitory computer-readable medium of claim 7 , wherein the neural network classifier pipeline of the analysis engine comprises a deep convolutional neural network.
12. The non-transitory computer-readable medium of claim 7 , the steps further comprising causing the analysis engine to output a binary classification for medical follow-up in a case where the analysis engine determines that the neural network classifier pipeline cannot accurately classify the detected skin abnormality based on the optical image data.
13. A computer system comprising one or more computing devices programmed to perform steps for automated diagnosis of a diabetic foot condition, the steps comprising:
causing one or more image capture devices of a medical diagnostic apparatus to capture optical image data of a target area of a foot;
causing a touch sensitivity testing device of the medical diagnostic apparatus to collect physical touch sensitivity data for the target area of the foot;
causing the medical diagnostic apparatus to transmit the optical image data and the physical touch sensitivity data to an analysis engine, wherein the analysis engine comprises a neural network classifier pipeline; and
obtaining one or more indications of a diabetic foot condition from the analysis engine,
wherein the neural network classifier pipeline of the analysis engine automatically performs classification steps on the optical image data to generate the one or more indications of a diabetic foot condition, and wherein the classification steps performed by the neural network classifier pipeline include:
detecting presence, position, and orientation of the foot;
detecting presence and location of a skin abnormality on the foot; and
classifying the detected skin abnormality as a diabetic foot ulcer.
14. The computer system of claim 13 , the steps further comprising checking the optical image data for one or more of image quality, lighting conditions, or body positioning.
15. The computer system of claim 13 , the steps further comprising, prior to collecting the physical touch sensitivity data, confirming the position of the foot based at least in part on the optical image data.
16. The computer system of claim 13 , wherein the touch sensitivity testing device comprises a servo-actuated monofilament probe.
17. The computer system of claim 13 , wherein the neural network classifier pipeline of the analysis engine comprises a deep convolutional neural network.
18. The computer system of claim 13 , wherein the one or more computing devices are further programmed to cause the analysis engine to output a binary classification for medical follow-up in a case where the analysis engine determines that the neural network classifier pipeline cannot accurately classify the detected skin abnormality based on the optical image data.