Method, device and system to determine a frequency function of a basilar membrane of cochleae to tune cochlear implants

1 . A method comprising:

inputting, at a computing device, a clinically available scan, of a cochlea of a given format, to a machine learning engine trained to: output higher resolution cochlear images of a resolution higher than the given format using input based on clinically available cochlear scans of the given format;

determining, at the computing device, using the machine learning engine, a higher resolution image of the cochlea;

determining, at the computing device, from the higher resolution image, a number of turns of the cochlea; and

one or more of determining and outputting, at the computing device, a frequency function of a basilar membrane of the cochlea that is dependent on an angle of the basilar membrane, by inputting the number of turns into a generic predetermined frequency function dependent on the number of turns and the angle.

2 . The method of claim 1 , wherein the clinically available scan comprises a preoperative clinically available scan, and the higher resolution image comprises a preoperative higher resolution image, the method further comprising:

inputting, at the computing device, a postoperative clinically available scan of the cochlea of the given format to the machine learning engine, the postoperative clinically available scan including an image of a cochlear implant inserted into the cochlea;

determining, at the computing device, using the machine learning engine, a postoperative higher resolution image of the cochlea and the cochlear implant;

determining, at the computing device, from the postoperative higher resolution image, respective positions of electrodes of the cochlear implant; and

tuning the electrodes of the cochlear implant based on the respective positions of the electrodes and the frequency function.

3 . The method of claim 2 , wherein determining, at the computing device, the respective positions of the electrodes of the cochlear implant comprises:

aligning the postoperative clinically available scan with the preoperative higher resolution image; and

determining the respective positions of the electrodes of the cochlear implant at the preoperative higher resolution image.

4 . The method of claim 1 , further comprising training the machine learning engine in a training mode using:

as training input to the machine learning engine, one or more of the clinically available cochlear scans of the given format and upsampled versions of the clinically available cochlear scans; wherein the given format comprises one or more of helical computerized tomography (CT) scans and cone beam CT scans; and

as training output to the machine learning engine, corresponding higher resolution cochlear scans, the corresponding higher resolution cochlear scans comprising one or more of corresponding higher resolution cochlear images and corresponding higher resolution cochlear segmentations, the corresponding higher resolution cochlear scans based on one or more of synchrotron radiation-phase contrast imaging scans and micro-CT scans; wherein the upsampled versions of the clinically available cochlear scans have a same resolution as the corresponding higher resolution cochlear scans.

5 . The method of claim 1 , wherein the generic predetermined frequency function comprises:

f

BM

=

2

(

0.0285

N

-

0.1978

)

⁢

θ

+

168

+

c

12

wherein N comprises the number of turns, θ comprises the angle in degrees as measured from a cochlear round window, f BM comprises a frequency response of the basilar membrane at the angle, and c comprises:

c=− 1.26-2.33 cos(0.0059θ)−6.84 sin(0.0059θ); or

wherein the generic predetermined frequency function is generated from synchrotron radiation-phase contrast imaging scans of a plurality of cochleae and basilar membranes.

6 . The method of claim 1 , wherein determining, from the higher resolution image, the number of turns of the cochlea comprises one or more of:

segmenting the higher resolution image;

identifying edges of the cochlea in the higher resolution image;

performing shape analysis on the higher resolution image; and

performing geometric analysis on the higher resolution image.

7 . The method of claim 1 , wherein determining, from the higher resolution image, the number of turns of the cochlea comprises analyzing the higher resolution image to determine estimations of one or more of:

a modiolar axis of the cochlea;

a basal plane of the cochlea;

a length of a hook of the cochlea;

a position of a helicotrema of the cochlea; and

a respective position of a round window of the cochlea.

8 . The method of claim 1 , wherein the clinically available scan further comprises a preoperative clinically available scan for planning insertion of a cochlear implant into the cochlea, and the higher resolution image comprises one or more of a preoperative higher resolution image and a preoperative higher resolution segmentation, the preoperative clinically available scan, the preoperative higher resolution image, and the preoperative higher resolution segmentation including one more of:

a temporal bone;

an external auditory canal;

ossicles;

a tympanic membrane;

an inner ear;

a round window;

a facial nerve;

a chorda tympani nerve; a

a sigmoid sinus;

a carotid artery; and

a tegmen.

9 . The method of claim 1 , wherein inputting the clinically available scan of the cochlea to the machine learning engine comprises, prior to the inputting: generating an upsampled version of the clinically available scan, the upsampled version of the clinically available scan used as input to the machine learning engine, and the upsampled version having a same resolution as one or more of the higher resolution image and respective higher resolution cochlear scans used to train the machine learning engine.

10 . The method of claim 1 , wherein the higher resolution image comprises a higher resolution segmentation of the cochlea.

11 . A device comprising:

a controller configured to:

input a clinically available scan, of a cochlea of a given format, to a machine learning engine trained to: output higher resolution cochlear images of a resolution higher than the given format using input based on clinically available cochlear scans of the given format;

determine, using the machine learning engine, a higher resolution image of the cochlea;

determine, from the higher resolution image, a number of turns of the cochlea; and

one or more of determine, and output, a frequency function of a basilar membrane of the cochlea that is dependent on an angle of the basilar membrane, by inputting the number of turns into a generic predetermined frequency function dependent on the number of turns and the angle.

12 . The device of claim 11 , wherein the clinically available scan comprises a preoperative clinically available scan, and the higher resolution image comprises a preoperative higher resolution image, the controller is further configured to:

input a postoperative clinically available scan of the cochlea of the given format to the machine learning engine, the postoperative clinically available scan including an image of a cochlear implant inserted into the cochlea;

determine, using the machine learning engine, a postoperative higher resolution image of the cochlea and the cochlear implant;

determine, from the postoperative higher resolution image, respective positions of electrodes of the cochlear implant; and

tune the electrodes of the cochlear implant based on the respective positions of the electrodes and the frequency function.

13 . The device of claim 12 , wherein the controller is further configured to determine the respective positions of the electrodes of the cochlear implant by:

aligning the postoperative clinically available scan with the preoperative higher resolution image; and

determining the respective positions of the electrodes of the cochlear implant at the preoperative higher resolution image.

14 . The device of claim 11 , wherein the controller is further configured to train the machine learning engine in a training mode using:

as training input to the machine learning engine, one or more of the clinically available cochlear scans of the given format and upsampled versions of the clinically available cochlear scans; wherein the given format comprises one or more of helical computerized tomography (CT) scans and cone beam CT scans; and

as training output to the machine learning engine, corresponding higher resolution cochlear scans, the corresponding higher resolution cochlear scans comprising one or more of corresponding higher resolution cochlear images and corresponding higher resolution cochlear segmentations, the corresponding higher resolution cochlear scans based on one or more of synchrotron radiation-phase contrast imaging scans and micro-CT scans; wherein the upsampled versions of the clinically available cochlear scans have a same resolution as the corresponding higher resolution cochlear scans.

15 . The device of claim 11 , wherein the generic predetermined frequency function comprises:

f

BM

=

2

(

0.0285

N

-

0.1978

)

⁢

θ

+

168

+

c

12

wherein N comprises the number of turns, θ comprises the angle in degrees as measured from a cochlear round window, f BM comprises a frequency response of the basilar membrane at the angle, and c comprises:

c

=

-

1.26

-

2.33

cos

⁡

(

0.0059

θ

)

-

6.84

sin

⁡

(

0.0059

θ

)

;

 or

wherein the generic predetermined frequency function is generated from synchrotron radiation-phase contrast imaging scans of a plurality of cochleae and basilar membranes.

16 . The device of claim 11 , wherein the controller is further configured to determine, from the higher resolution image, the number of turns of the cochlea by one or more of:

segmenting the higher resolution image;

identifying edges of the cochlea in the higher resolution image;

performing shape analysis on the higher resolution image; and

performing geometric analysis on the higher resolution image.

17 . The device of claim 11 , wherein the controller is further configured to determine, from the higher resolution image, the number of turns of the cochlea comprises analyzing the higher resolution image to determine estimations of one or more of:

a modiolar axis of the cochlea;

a basal plane of the cochlea;

a length of a hook of the cochlea;

a position of a helicotrema of the cochlea; and

a respective position of a round window of the cochlea.

18 . The device of claim 11 , wherein the clinically available scan further comprises a preoperative clinically available scan for planning insertion of a cochlear implant into the cochlea, and the higher resolution image comprises one or more of a preoperative higher resolution image and a preoperative higher resolution segmentation, the preoperative clinically available scan, the preoperative higher resolution image, and the preoperative higher resolution segmentation including one more of:

a temporal bone;

an external auditory canal;

ossicles;

a tympanic membrane;

an inner ear;

a round window;

a facial nerve;

a chorda tympani nerve; a

a sigmoid sinus;

a carotid artery; and

a tegmen.

19 . The device of claim 11 , wherein inputting the clinically available scan of the cochlea to the machine learning engine comprises, prior to the inputting: generating an upsampled version of the clinically available scan, the upsampled version of the clinically available scan used as input to the machine learning engine, and the upsampled version having a same resolution as one or more of the higher resolution image and respective higher resolution cochlear scans used to train the machine learning engine.

20 . The device of claim 11 , wherein the higher resolution image comprises a higher resolution segmentation of the cochlea.