Method and system for classifying tympanic membrane, and non-transitory computer-readable medium

Abstract

To provide a method and a system for classifying a tympanic membrane and a non-transitory computer-readable medium.SOLUTION: Disclosed herein are a system and a method for classifying a tympanic membrane by using a classifier. The classifier is a machine learning algorithm. A method for classifying the tympanic membrane includes the steps of: receiving, from an interrogation system, one or more datasets relating to the tympanic membrane; determining a set of parameters from the one or more datasets, wherein at least one parameter of the set of parameters is related to a dynamic property or a static position of the tympanic membrane; and outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters. The classification comprises one or more of a state, a condition, or a mobility metric of the tympanic membrane.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] (CROSS REFERENCE TO RELATED APPLICATIONS) This application claims the benefit of U.S. Provisional Application No. 62 / 796,762, filed Jan. 25, 2019, which application is incorporated herein by reference.

[0002] The present disclosure relates to systems and methods for characterizing tissue and associated materials, such as for diagnosis. Specifically, the present disclosure relates to measuring tissue mobility or location, such as tympanic membrane mobility or location, to diagnose and characterize otitis media (i.e., ear infection). An embodiment of the present disclosure involves machine learning using a training data set consisting of tympanic membrane optical and / or acoustic and / or pressure datasets, for example, to predict tympanic membrane mobility or location. A predictive model is generated using the training set and can be used to characterize a patient's tympanic membrane mobility or location, and thereby diagnose and characterize any otitis media. [Background technology]

[0003] Otitis media (OM) is a group of inflammatory disorders of the inner ear and is the cause of the most common pediatric medical problem commonly known as "ear infection." OM is defined by the presence of fluid effusion within the middle ear and consists of two major types: acute otitis media (AOM), otitis media with effusion (OME), chronic otitis media (COM), and chronic suppurative otitis media (CSOM). Rapid onset of infection, usually presenting with ear pain, is characteristic of AOM middle ear effusion (MEE), while OME is typically not associated with symptoms as MEE fluid is non-infectious.

[0004] The diagnosis of OM most commonly relies on otoscopy to visually assess the tympanic membrane (TM), or "tympanic cavity." The otoscope illuminates and magnifies the TM, allowing the physician to qualitatively interpret visual indicators associated with OM infection, such as color, transparency, and bulging. However, the accuracy of OME diagnosis by otoscopy varies due to equipment and physician skill, and pediatricians and otolaryngologists with extensive experience in OM have an estimated diagnostic accuracy in the range of 50-70% using otoscopy. Given the difficulty of properly diagnosing OM through otoscopy, only 40% of leading pediatricians are confident about their otoscopic findings, and AOM is frequently misdiagnosed.

[0005] Therefore, improved systems, devices, and methods for classifying the tympanic membrane and / or diagnosing otitis media are desirable.

[0006] References that may be relevant to the disclosure of this specification include U.S. Patent No. US5235510, US5331550, US6090044, US7058441, US7283653, US7771356, US8115934, US8798345, US8996098, US9053536, US94457, 13, US9536054, US9589374, US9636007, US9684960, US9750450, US9867528, US9922242, and US10013757, U.S. Publication Nos. US2003026470, US2004068167, US2006235725, US200 No. 6282009, No. US2007112273, No. US2009112100, No. US2012185275, No. US2013303941, No. US2014036 No. 054, No. US2017071509, No. US2017126943, No. US2017132367, No. US2018025210, No. US2018211380, Nos. US2018242860, and US2018260616, and International Publication Nos. WO2000 / 010451, WO2001 / 026026, WO2009 / 157825, WO2018 / 045269, WO2018 / 106005, WO2018 / 140014, and WO2018 / 198253.

[0007] The following non-patent publications may be relevant to the disclosure herein: Vertan, C. et al., "Eardrum color content analysis in video-otoscopy images for the diagnosis support of pediatric," Int. Symp. on Signals, Circuits and Systems, Bucharest, Romania, July 2011; Shaikh, N. et al., "Development of an algorithm for the diagnosis of otitis media," Academic Pediatrics, Volume 12, Number 3 (May-June 2012); Kuruvilla, A. et al., "Automated diagnosis of otitis media: vocabulary and grammar," International Journal of Biomedical Imaging: NA. Hindawl Limited. (2013); Rappaport, K. et al., "Comparative Assessment of a Smartphone Otoscope for the Diagnosis and Management of Acute Otitis Media," Provisional Section on Advances in Therapeutics and Technology Poster. Presentation(October 28,2013), Shie, CK et al., "A hybrid feature-based segmentation and classification system for the computer aided self-diagnosis of otitis media" Conf Proc IEEE Eng Med Biol Soc.,2014; 2014:4655-8, Somrak, M. et al.,「Tricorder:consumer medical device for discovering common medical conditions」Informatica 38.1:81(8). Slovenian Society Informatika.(Mar. 2014)、Myburgh,H. et al.,「Towards low cost automated smartphone-. and cloud-based otitis media diagnosis」Biomedical Signal Processing and Control 39(2018) 34-52, Umea University. 「Diagnosing ear infection using smartphone.」ScienceDaily. ScienceDaily, 30 March 2016. <www.sciencedaily.com / releases / 2016 / 03 / 160330102850.htm>, Senaras, C. et al, 「Detection of eardrum abnormalities using ensemble deep learning approaches」Proc. SPIE 10575, Medical Imaging 2018: Computer-Aided Diagnosis, 105751A (27 February 2018), Kasher, M.S., 「Otitis Media Analysis: An Automated Feature Extraction and Image Classification System」Helsink Metropolia University of Applied Sciences, Bachelor of Engineering, Degree Programme in Electronics, Bachelor’s Thesis (25.04.2018), and Tran, T. et al., 「Development of an Automatic Diagnostic Algorithm for Pediatric Otitis Media」Otology & neurotology: official publication of the American Otological Society, American Neurotology Society [and] European Academy of Otology and Neurotology 39.8: 1060-1065. (Sep 2018). [Prior Art Documents] [Patent documents]

[0008] [Patent Document 1] U.S. Patent No. 5,235,510 Summary of the Invention [Means for solving the problem]

[0009] The present disclosure relates to systems and methods for measuring tympanic membrane mobility and position to diagnose and characterize otitis media (i.e., ear infection). Machine learning may be used with a training data set consisting of tympanic membrane optical and / or ultrasound and / or pressure datasets to generate predictive models for tympanic membrane mobility and / or position.

[0010] Aspects of the present disclosure provide a method for classifying a tympanic membrane, which may include receiving one or more data sets related to the tympanic membrane from an interrogation system, determining a set of parameters from the one or more data sets, where at least one parameter of the set of parameters is related to a dynamic property or a static position of the tympanic membrane, and outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters, where the classification comprises one or more of a state, condition, or mobility metric of the tympanic membrane.

[0011] In some embodiments, the interrogation system comprises an imaging system, and the one or more datasets comprise one or more images of the tympanic membrane. In some embodiments, the classifier model comprises a machine learning algorithm. In some embodiments, the machine learning algorithm comprises one or more of linear regression, logistic regression, classification and regression tree algorithms, support vector machines (SVM), naive Bayes, K-nearest neighbors, random forest algorithms, boosting algorithms such as XGBoost and LightGBM, neural networks, convolutional neural networks, and recurrent neural networks. In some embodiments, the machine learning algorithm is a supervised learning algorithm, an unsupervised learning algorithm, or a semi-supervised learning algorithm.

[0012] In some embodiments, the one or more images of the tympanic membrane comprise one or more ultrasound images. In some embodiments, the one or more ultrasound images comprise images from a pneumatic ultrasound tympanoscope. In some embodiments, the one or more ultrasound images are measured in response to pneumatic excitation. In some embodiments, the one or more images of the tympanic membrane comprise one or more optical coherence tomography images. In some embodiments, the images comprise optical images.

[0013] In some embodiments, at least one parameter related to the dynamic property or static position of the membrane is responsive to the pneumatic excitation. In some embodiments, the pneumatic excitation comprises a puff of gas. In some embodiments, the pneumatic excitation has a frequency greater than 10 Hz. In some embodiments, the dynamic property of the tympanic membrane comprises an indication of membrane movement or membrane mobility, a minimum or maximum displacement of the tympanic membrane, an outlier displacement, a difference or ratio between the minimum and maximum displacement, a slope of the displacement or a slope of the difference or ratio between the minimum and maximum displacement with respect to the pressure of the pneumatic excitation, a response of the measured pressure to the applied pressure, a visual movement of the tympanic membrane in response to the pneumatic excitation, one or more statistical components generated from singular value decomposition, principal component analysis, and K-means clustering, and one or more of an ultrasound pulse echo amplitude or an ultrasound echo phase or a derivative thereof or a moving average thereof. In some embodiments, the dynamic property of the tympanic membrane is normalized to the pressure of the pneumatic excitation.

[0014] In some embodiments, the method further includes generating one or more of a rank readout, a categorical readout, or a continuous numeric output of the tympanic membrane mobility. In some embodiments, the rank readout comprises a numeric scale related to the degree of membrane mobility. In some embodiments, the numeric scale comprises a classification of 0 to 4+. In some embodiments, the categorical readout comprises an indication of the degree of membrane mobility as at least one of highly mobile, moderately mobile, semi-mobile, or non-mobile. In some embodiments, the categorical readout comprises a binary classification. In some embodiments, the continuous numeric output comprises one or more of a measured membrane displacement, a rate of membrane movement, or a rate of membrane recovery.

[0015] In some embodiments, the state or condition of the tympanic membrane comprises one or more of acute otitis media, acute otitis media with effusion, middle ear effusion, chronic otitis media, chronic suppurative otitis media, bacterial infection, viral infection, no effusion, and unknown classification.

[0016] In some embodiments, the one or more datasets comprise an m-mode ultrasound dataset. In some embodiments, the one or more datasets comprise an infrared image. In some embodiments, the one or more datasets comprise an air pressure dataset. In some embodiments, the one or more datasets comprise one or more optical images taken in response to an air pressure excitation. In some embodiments, the static position comprises an inflated membrane or a retracted membrane.

[0017] In another aspect, the present disclosure provides a system for classifying a tympanic membrane, the system comprising a computing system comprising a memory, the memory may comprise instructions for classifying the tympanic membrane, the computing system configured to execute the instructions to at least receive from an interrogation system one or more data sets related to the tympanic membrane, determine a set of parameters from the one or more data sets, at least one parameter of the set of parameters being related to a dynamic property or a static position of the tympanic membrane, and output a classification of the tympanic membrane based on a classifier model derived from the set of parameters, the classification comprising a state, condition, or mobility metric of the tympanic membrane.

[0018] In some embodiments, the interrogation system comprises an imaging system and the one or more data sets comprise one or more images of the tympanic membrane. In some embodiments, the system further comprises a pneumatic ultrasound tympanoscope.

[0019] In some embodiments, the classifier model comprises a machine learning algorithm. In some embodiments, the machine learning algorithm comprises one or more of linear regression, logistic regression, classification and regression tree algorithms, support vector machines (SVM), naive Bayes, K-nearest neighbors, random forest algorithms, boosting algorithms such as XGBoost and LightGBM, neural networks, convolutional neural networks, and recurrent neural networks. In some embodiments, the machine learning algorithm is a supervised learning algorithm, an unsupervised learning algorithm, or a semi-supervised learning algorithm.

[0020] In some embodiments, the one or more images of the tympanic membrane comprise one or more ultrasound images. In some embodiments, the one or more ultrasound images comprise images from a pneumatic ultrasound tympanoscope. In some embodiments, the one or more ultrasound images are measured in response to pneumatic excitation. In some embodiments, the one or more images of the tympanic membrane comprise one or more optical coherence tomography images. In some embodiments, the images comprise optical images.

[0021] In some embodiments, at least one parameter related to the dynamic property or static position of the membrane is in response to a pneumatic excitation. In some embodiments, the pneumatic excitation comprises a puff of gas. In some embodiments, the pneumatic excitation comprises a frequency greater than 10 Hz. In some embodiments, the dynamic property of the tympanic membrane comprises an indication of membrane movement or membrane mobility, a minimum or maximum displacement of the tympanic membrane, an outlier displacement, a difference or ratio between the minimum and maximum displacement, a slope of the displacement or a slope of the difference or ratio between the minimum and maximum displacement with respect to the pressure of the pneumatic excitation, a response of the measured pressure to the applied pressure, a visual movement of the tympanic membrane in response to the pneumatic excitation, one or more statistical components generated from singular value decomposition, principal component analysis, and K-means clustering, and one or more of an ultrasound pulse echo amplitude or an ultrasound echo phase or a derivative or a moving average thereof. In some embodiments, the dynamic property of the tympanic membrane is normalized to the pressure of the pneumatic excitation.

[0022] In some embodiments, the computing system is further configured to execute instructions to generate one or more of a rank readout, a categorical readout, or a continuous numeric output of the tympanic membrane mobility. In some embodiments, the rank readout comprises a numerical scale related to the degree of membrane mobility. In some embodiments, the numerical scale comprises a classification of 0 to 4+. In some embodiments, the categorical readout comprises an indication of the degree of membrane mobility as at least one of highly mobile, medium mobile, semi-mobile, or non-mobile. In some embodiments, the categorical readout comprises a binary classification. In some embodiments, the continuous numeric output comprises one or more of a measured membrane displacement, a rate of membrane movement, or a rate of membrane recovery.

[0023] In some embodiments, the state or condition of the tympanic membrane comprises one or more of acute otitis media, acute otitis media with effusion, middle ear effusion, chronic otitis media, chronic suppurative otitis media, bacterial infection, viral infection, no effusion, and unknown classification.

[0024] In some embodiments, the one or more datasets comprise an m-mode ultrasound dataset. In some embodiments, the one or more datasets comprise an infrared image. In some embodiments, the one or more datasets comprise an air pressure dataset. In some embodiments, the one or more datasets comprise one or more optical images taken in response to an air pressure excitation. In some embodiments, the static position comprises an inflated membrane or a retracted membrane.

[0025] In another aspect, the present disclosure provides a non-transitory computer readable medium comprising machine-executable code that, when executed by a computing system, implements a method of classifying a membrane, the method including receiving from an interrogation system one or more datasets related to the tympanic membrane; determining a set of parameters from the one or more datasets, where at least one parameter of the set of parameters is related to a dynamic characteristic or a static position of the tympanic membrane; and outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters, where the classification comprises a state, condition, or mobility metric of the tympanic membrane.

[0026] In some embodiments, the interrogation system comprises an imaging system, and the one or more datasets comprise one or more images of the tympanic membrane. In some embodiments, the method further includes the method of any aspect or embodiment disclosed herein.

[0027] In another aspect, a method of training a computer-implemented classifier is provided. The method may include receiving a set of parameters based on one or more datasets related to one or more tympanic membranes and one or more classified datasets related to the one or more tympanic membranes, where the classification comprises a state, condition, or mobility metric of the tympanic membrane, and the set of parameters comprises at least one parameter related to a dynamic characteristic or a static position of the one or more tympanic membranes, storing the set of parameters and the one or more classified datasets in a database, constructing a classifier model based on the set of parameters and the one or more classified datasets, where the classifier model is derived from the set of parameters, and where the classifier model outputs a classification based on a dataset of the one or more classified datasets, and using the classifier model to provide a classification of an unclassified dataset.

[0028] In some embodiments, the method further includes updating the database based on the second one or more classified data sets and updating the classifier model based on the second one or more classified data sets. In some embodiments, the method further includes using the classifier model to provide a classification of the unclassified data set by the method of classifying the tympanic membrane of any aspect or embodiment.

[0029] In another aspect, a system for training a computer-implemented classifier is provided, the system comprising a computing system comprising a memory, the memory may comprise instructions for training a dataset, the computing system configured to execute the instructions to at least receive a set of parameters based on one or more datasets associated with one or more tympanic membranes and one or more classified datasets associated with the one or more tympanic membranes, the classification comprising a state, condition, or mobility metric of the tympanic membrane, the set of parameters comprising at least one parameter related to a dynamic characteristic or a static position of the one or more tympanic membranes, store the set of parameters and the one or more classified datasets in a database, build a classifier model based on the set of parameters and the one or more classified datasets, the classifier model being derived from the set of parameters, the classifier model outputting a classification based on a dataset of the one or more classified datasets, and use the classifier model to provide a classification of an unclassified dataset.

[0030] In some embodiments, the system is configured to execute instructions to at least update the database based on the second one or more classified data sets and update the classifier model based on the second one or more classified data sets. In some embodiments, the system further comprises a system for classifying the tympanic membrane of any aspect or embodiment.

[0031] In another aspect, the disclosure provides a non-transitory computer-readable medium comprising machine-executable code that, when executed by a computing system, implements a method of training a computer-implemented classifier. The method may include receiving a set of parameters based on one or more datasets associated with one or more tympanic membranes and one or more classified datasets associated with the one or more tympanic membranes, where the classification comprises a state, condition, or mobility metric of the tympanic membrane, and the set of parameters comprises at least one parameter related to a dynamic characteristic or a static position of the one or more tympanic membranes, storing the set of parameters and the one or more classified datasets in a database, constructing a classifier model based on the set of parameters and the one or more classified datasets, where the classifier model is derived from the set of parameters, and where the classifier model outputs a classification, and using the classifier model to provide a classification of the unclassified dataset.

[0032] In some embodiments, the method further includes updating the database based on the second one or more classified data sets and updating the classifier model based on the second one or more classified data sets. In some embodiments, the method further includes using the classifier model to provide a classification of the unclassified data set by the method of classifying the tympanic membrane of any aspect or embodiment. The present invention provides, for example, the following: (Item 1) 1. A method for classifying a tympanic membrane, the method comprising: receiving one or more data sets related to the tympanic membrane from an interrogation system; determining a set of parameters from the one or more data sets, at least one parameter of the set of parameters being related to a dynamic property or a static position of the tympanic membrane; outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters; Including, The method, wherein the classification comprises one or more of a state, condition, or mobility metric of the tympanic membrane. (Item 2) 2. The method of claim 1, wherein the interrogation system comprises an imaging system and the one or more data sets comprise one or more images of the tympanic membrane. (Item 3) Item 10. The method of item 1, wherein the classifier model comprises a machine learning algorithm. (Item 4) Item 4. The method of item 3, wherein the machine learning algorithm comprises one or more of linear regression, logistic regression, classification and regression tree algorithms, support vector machines (SVM), Naive Bayes, K-nearest neighbors, random forest algorithms, boosting algorithms such as XGBoost and LightGBM, neural networks, convolutional neural networks, and recurrent neural networks. (Item 5) 4. The method of claim 3, wherein the machine learning algorithm is a supervised learning algorithm, an unsupervised learning algorithm, or a semi-supervised learning algorithm. (Item 6) 3. The method of claim 2, wherein the one or more images of the tympanic membrane comprise one or more ultrasound images. (Item 7) 7. The method of claim 6, wherein the one or more ultrasound images comprise images from a pneumatic ultrasound tympanoscope. (Item 8) 7. The method of claim 6, wherein the one or more ultrasound images are measured in response to pneumatic excitation. (Item 9) 3. The method of claim 2, wherein the one or more images of the tympanic membrane comprise one or more optical coherence tomography images. (Item 10) Item 3. The method of item 2, wherein the image comprises an optical image. (Item 11) 3. The method of claim 2, wherein the at least one parameter related to a dynamic property or a static position of the membrane is in response to a pneumatic excitation. (Item 12) Item 12. The method of item 11, wherein the pneumatic excitation comprises a gas blowing. (Item 13) 12. The method of claim 11, wherein the pneumatic excitation has a frequency greater than 10 Hz. (Item 14) The dynamic characteristic of the tympanic membrane may be Indication of membrane translocation or membrane mobility; a minimum or maximum displacement of the tympanic membrane; Outlier variance and the difference or ratio between the minimum and maximum displacement; the slope of the displacement or the slope of the difference or ratio between the minimum and maximum displacements versus the pressure of the pneumatic excitation; a response of the measured pressure to the applied pressure; and a visual movement of the tympanic membrane in response to air pressure excitation; one or more statistical components generated from singular value decomposition, principal component analysis, and K-means clustering; Ultrasonic pulse echo amplitude or ultrasonic echo phase or its derivative or its moving average Item 1. The method of item 1, comprising one or more of the following: (Item 15) 15. The method of claim 14, wherein the dynamic properties of the tympanic membrane are normalized to a pressure of a pneumatic excitation. (Item 16) 2. The method of claim 1, further comprising generating one or more of a rank readout, a categorical readout, or a continuous numerical output of the mobility of the tympanic membrane. (Item 17) 17. The method of claim 16, wherein the rank readout comprises a numerical scale related to the degree of membrane mobility. (Item 18) Item 18. The method of item 17, wherein the numerical scale has a classification of 0 to 4+. (Item 19) 17. The method of claim 16, wherein the category readout comprises an indication of a degree of membrane mobility as at least one of: highly mobile, medium mobile, semi-mobile, or non-mobile. (Item 20) Item 17. The method of item 16, wherein the categorical reading comprises a binary classification. (Item 21) 17. The method of claim 16, wherein the continuous numerical output comprises one or more of a measured membrane displacement, a rate of membrane movement, or a rate of membrane recovery. (Item 22) 2. The method of claim 1, wherein the state or condition of the tympanic membrane comprises one or more of the following classifications: acute otitis media, acute otitis media with effusion, middle ear effusion, chronic otitis media, chronic suppurative otitis media, bacterial infection, viral infection, no effusion, and unknown. (Item 23) 2. The method of claim 1, wherein the one or more datasets comprise an m-mode ultrasound dataset. (Item 24) Item 10. The method of item 1, wherein the one or more data sets comprise infrared images. (Item 25) 2. The method of claim 1, wherein the one or more data sets comprises an air pressure data set. (Item 26) 2. The method of claim 1, wherein the one or more data sets comprise one or more optical images taken in response to a pneumatic excitation. (Item 27) 2. The method of claim 1, wherein the static position comprises an inflated membrane or a retracted membrane. (Item 28) 1. A system for classifying a tympanic membrane, the system comprising a computing system having a memory, the memory comprising instructions for classifying the tympanic membrane, the computing system comprising at least: receiving one or more data sets related to the tympanic membrane from an interrogation system; determining a set of parameters from the one or more data sets, at least one parameter of the set of parameters being related to a dynamic property or a static position of the tympanic membrane; outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters; configured to execute the instructions to The system, wherein the classification comprises a state, condition, or mobility metric of the tympanic membrane. (Item 29) 30. The system of claim 28, wherein the interrogation system comprises an imaging system and the one or more data sets comprise one or more images of the tympanic membrane. (Item 30) 30. The system of claim 28, further comprising a pneumatic ultrasound tympanoscope. (Item 31) 29. The system of claim 28, wherein the classifier model comprises a machine learning algorithm. (Item 32) Item 32. The system of item 31, wherein the machine learning algorithm comprises one or more of linear regression, logistic regression, classification and regression tree algorithms, support vector machines (SVM), naive Bayes, K-nearest neighbors, random forest algorithms, boosting algorithms such as XGBoost and LightGBM, neural networks, convolutional neural networks, and recurrent neural networks. (Item 33) Item 32. The system of item 31, wherein the machine learning algorithm is a supervised learning algorithm, an unsupervised learning algorithm, or a semi-supervised learning algorithm. (Item 34) 30. The system of claim 29, wherein the one or more images of the tympanic membrane comprise one or more ultrasound images. (Item 35) 35. The system of claim 34, wherein the one or more ultrasound images comprise images from a pneumatic ultrasound tympanoscope. (Item 36) Item 35. The system of item 34, wherein the one or more ultrasound images are measured in response to pneumatic excitation. (Item 37) 30. The system of claim 29, wherein the one or more images of the tympanic membrane comprise one or more optical coherence tomography images. (Item 38) 30. The system of claim 29, wherein the image comprises an optical image. (Item 39) 30. The system of claim 29, wherein the at least one parameter related to a dynamic property or static position of the membrane is in response to a pneumatic excitation. (Item 40) 40. The system of claim 39, wherein the pneumatic excitation comprises a blowing of gas. (Item 41) 40. The system of claim 39, wherein the pneumatic excitation has a frequency greater than 10 Hz. (Item 42) The dynamic characteristic of the tympanic membrane may be Indication of membrane translocation or membrane mobility; a minimum or maximum displacement of the tympanic membrane; Outlier variance and the difference or ratio between the minimum and maximum displacement; the slope of the displacement or the slope of the difference or ratio between the minimum and maximum displacements versus the pressure of the pneumatic excitation; a response of the measured pressure to the applied pressure; and a visual movement of the tympanic membrane in response to air pressure excitation; one or more statistical components generated from singular value decomposition, principal component analysis, and K-means clustering; Ultrasonic pulse echo amplitude or ultrasonic echo phase or its derivative or its moving average 29. The system of claim 28, further comprising one or more of the following: (Item 43) Item 43. The system of item 42, wherein the dynamic properties of the tympanic membrane are normalized to a pressure of pneumatic excitation. (Item 44) 30. The system of claim 28, wherein the computing system is further configured to execute the instructions to generate one or more of a rank readout, a categorical readout, or a continuous numerical output of the tympanic membrane mobility. (Item 45) 45. The system of item 44, wherein the rank readout comprises a numerical scale related to the degree of membrane mobility. (Item 46) Item 46. The system of item 45, wherein the numerical scale has a classification of 0 to 4+. (Item 47) 45. The system of claim 44, wherein the category readout comprises an indication of a degree of membrane mobility as at least one of highly mobile, medium mobile, semi-mobile, or non-mobile. (Item 48) Item 45. The system of item 44, wherein the categorical readout comprises a binary classification. (Item 49) 45. The system of claim 44, wherein the continuous numerical output comprises one or more of a measured membrane displacement, a rate of membrane movement, or a rate of membrane recovery. (Item 50) 29. The system of claim 28, wherein the state or condition of the tympanic membrane comprises one or more of the following classifications: acute otitis media, acute otitis media with effusion, middle ear effusion, chronic otitis media, chronic suppurative otitis media, bacterial infection, viral infection, no effusion, and unknown. (Item 51) 30. The system of claim 28, wherein the one or more datasets comprise an m-mode ultrasound dataset. (Item 52) 30. The system of claim 28, wherein the one or more data sets comprise an infrared image. (Item 53) 30. The system of claim 28, wherein the one or more data sets comprise an air pressure data set. (Item 54) 30. The system of claim 28, wherein the one or more data sets comprise one or more optical images taken in response to a pneumatic excitation. (Item 55) 30. The system of claim 28, wherein the static position comprises an inflated membrane or a retracted membrane. (Item 56) A non-transitory computer readable medium comprising machine executable code that, when executed by a computing system, implements a method for classifying a membrane, the method comprising: receiving one or more data sets related to the tympanic membrane from an interrogation system; determining a set of parameters from the one or more data sets, at least one parameter of the set of parameters being related to a dynamic property or a static position of the tympanic membrane; outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters; Including, The classification comprises a state, condition, or mobility metric of the tympanic membrane. (Item 57) Item 57. The non-transitory computer-readable storage medium of item 56, wherein the interrogation system comprises an imaging system, and the one or more data sets comprise one or more images of the tympanic membrane. (Item 58) 57. The non-transitory computer-readable storage medium of item 56, further comprising the method of any one of items 2-27. (Item 59) 1. A method of training a computer-implemented classifier, the method comprising: receiving a set of parameters based on one or more datasets associated with one or more tympanic membranes and one or more classified datasets associated with the one or more tympanic membranes, the classification comprising a state, condition, or mobility metric of the tympanic membrane, and the set of parameters comprising at least one parameter related to a dynamic property or a static position of the one or more tympanic membranes; storing the set of parameters and the one or more classified data sets in a database; constructing a classifier model based on the set of parameters and the one or more classified datasets, the classifier model being derived from the set of parameters, the classifier model outputting a classification based on a dataset of the one or more classified datasets; using said classifier model to provide a classification of an unclassified data set; A method comprising: (Item 60) The method further comprises: updating the database based on a second one or more classified data sets; updating the classifier model based on the second one or more classified data sets; 60. The method of claim 59, further comprising: (Item 61) 61. The method of claim 59 or 60, further comprising using the classifier model to provide a classification of an unclassified dataset by the method for classifying the tympanic membrane of any one of claims 1-27. (Item 62) 10. A system for training a computer-implemented classifier, the system comprising a computing system having a memory, the memory comprising instructions for training the dataset, the computing system comprising at least: receiving a set of parameters based on one or more datasets associated with one or more tympanic membranes and one or more classified datasets associated with the one or more tympanic membranes, the classification comprising a state, condition, or mobility metric of the tympanic membrane, the set of parameters comprising at least one parameter related to a dynamic property or a static position of the one or more tympanic membranes; storing the set of parameters and the one or more classified data sets in a database; constructing a classifier model based on the set of parameters and the one or more classified datasets, the classifier model being derived from the set of parameters, the classifier model outputting a classification based on a dataset of the one or more classified datasets; using said classifier model to provide a classification of an unclassified data set; A system configured to execute the instructions to: (Item 63) Item 63. The system of item 62, wherein the system is configured to execute the instructions to at least update the database based on a second one or more classified data sets and update the classifier model based on the second one or more classified data sets. (Item 64) 64. The system of claim 62 or 63, further comprising a system for classifying the tympanic membrane of any one of items 28-55. (Item 65) A non-transitory computer readable medium comprising machine executable code that, when executed by a computing system, implements a method for training a computer-implemented classifier, the method comprising: receiving a set of parameters based on one or more datasets associated with one or more tympanic membranes and one or more classified datasets associated with the one or more tympanic membranes, the classification comprising a state, condition, or mobility metric of the tympanic membrane, and the set of parameters comprising at least one parameter related to a dynamic property or a static position of the one or more tympanic membranes; storing the set of parameters and the one or more classified data sets in a database; constructing a classifier model based on the set of parameters and the one or more classified datasets, the classifier model being derived from the set of parameters, the classifier model outputting a classification; and using said classifier model to provide a classification of an unclassified data set; A non-transitory computer readable medium comprising: (Item 66) 66. The non-transitory computer-readable medium of claim 65, wherein the method further comprises updating the database based on a second one or more classified data sets and updating the classifier model based on the second one or more classified data sets. (Item 67) 67. The non-transitory computer-readable medium of claim 65 or 66, wherein the method further comprises using the classifier model to provide a classification of an unclassified dataset by the method of classifying the tympanic membrane of any one of claims 1-28.

[0033] (Incorporated by reference) All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference. To the extent that the publications and patents or patent applications incorporated by reference conflict with the disclosure contained herein, the present specification is intended to supersede and / or precede any such conflicting material. [Brief description of the drawings]

[0034] The novel features of the invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description and the accompanying drawings (also referred to herein as "Figure" and "FIG.") that set forth illustrative embodiments, in which the principles of the invention are utilized.

[0035] [Figure 1] FIG. 1 is a flowchart of an exemplary method for characterizing the tympanic membrane, according to some embodiments.

[0036] [Diagram 2] FIG. 2 is a schematic diagram illustrating an example system 200 for classifying the tympanic membrane, according to some embodiments.

[0037] [Diagram 3] FIG. 3 is a schematic diagram illustrating an example of an interrogation system as disclosed herein.

[0038] [Figure 4] FIG. 4 illustrates an exemplary data set provided by an interrogation device of the present disclosure.

[0039] [Diagram 5] 5 and 5B are schematic diagrams of parameter sets derived from pressure, ultrasound, and optical measurements that may constitute input to a classifier model for outputting a classification, according to some embodiments.

[0040] [Figure 6] FIG. 6 is a schematic diagram illustrating an example classifier model 600 comprising a machine learning algorithm, according to some embodiments.

[0041] [Figure 7] 7A, 7B, and 7C are schematic diagrams illustrating example decision trees that may be generated by a machine learning algorithm, according to some embodiments.

[0042] [Figure 8] FIG. 8 is a flowchart of an exemplary method for training a computer-implemented classifier, according to some embodiments.

[0043] [Figure 9] FIG. 9 is a schematic diagram illustrating an example of a system for training a computer-implemented classifier.

[0044] [Figure 10]FIG. 10 is a schematic diagram illustrating an exemplary system for classifying tympanic membranes comprising a digital processing device and a display visible to a user, according to some embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0045] The methods, systems, and media as disclosed herein may improve existing methods of classifying tissues by providing improved classification (e.g., diagnosis) of the tissue. For example, the methods and systems provided herein may use machine learning methods to build a classifier that improves the classification of the tympanic membrane. Machine learning approaches may leverage large data sets to gain new insights into the data sets. The classifier model may improve the characterization of the tympanic membrane, which may lead to better patient outcomes. The classifier model may provide clinicians with more accurate information for otitis media management (e.g., tympanic membrane mobility information and / or tympanic membrane status or condition) while reducing variance due to the training and experience of the physician using pneumatic otoscopy. In some cases, the methods and systems provided herein may provide disease status directly from the trained and validated algorithm.

[0046] The methods, systems, and media as disclosed herein may improve existing methods of classifying tissues by utilizing parameters related to the mobility of the tympanic membrane. A normal tympanic membrane may move in response to pressure, and a lack or reduction in mobility is an indication of fluid in the middle ear, perforation, or tympanosclerosis, etc. The addition of mobility measurements may be important for diagnostic accuracy, since the predictive value of visible tympanic characteristics for OM diagnosis may vary. Thus, the additional information of mobility may provide an indication of the presence of effusion, even when the appearance of the tympanic membrane does not provide any indication of middle ear pathology.

[0047] Methods, systems, and media as disclosed herein may be used in combination with devices and methods for characterizing ductile film, surface, and subsurface properties, such as those described in, for example, commonly owned U.S. Patent No. 7,771,356 and U.S. Patent Publication Nos. 2019 / 0365292, 2018 / 0310917, and 2017 / 0014053, each of which is incorporated by reference in its entirety. Methods, systems, and media as disclosed herein may be used in combination with devices and methods using optical coherence tomography (OCT), such as those disclosed in, for example, commonly assigned U.S. Patent Publication Nos. 2019 / 0200873 and 2017 / 0360302, each of which is incorporated by reference in its entirety.

[0048] The methods, systems, and media as disclosed herein can be used to characterize a number of biological tissues and provide various diagnostic information. The biological tissues can comprise patient organs. A speculum can be placed within a body cavity to characterize the patient tissue. The patient organs or body cavities can comprise, for example, muscles, tendons, ligaments, mouth, tongue, pharynx, esophagus, stomach, intestines, anus, liver, gallbladder, pancreas, nose, larynx, trachea, lungs, kidneys, bladder, urethra, uterus, vagina, ovaries, testes, prostate, heart, arteries, veins, spleen, glands, brain, spinal cord, nerves, etc., to name a few.

[0049] The methods, systems, and media as disclosed herein may be used to classify the tympanic membrane. For example, the membrane may be classified to determine ear conditions such as acute otitis media (AOM), chronic otitis media, otitis media with effusion, and / or chronic suppurative otitis media. Classification of an ear exhibiting AOM may include detection of the presence of effusion and characterization of the type of effusion as one of serous, mucous, purulent, or combinations thereof. In AOM, middle ear effusion (MEE) may be induced by a pathogen, and the MEE may be thin or serous in the case of a viral infection and thicker and purulent in the case of a bacterial infection. Thus, determining various properties of the fluid adjacent to the tympanic membrane may provide information that may be used to characterize the membrane.

[0050] Reference will now be made in detail to various embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the disclosure and the described embodiments. However, embodiments of the disclosure may, optionally, be practiced without these specific details. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the embodiments. In the drawings, like reference numbers designate like or similar steps or components.

[0051] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the claims. When used in the description of the embodiments and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It is also to be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. It is further to be understood that the terms "comprises" and / or "comprising" as used herein specify the presence of the stated features, integers, steps, operations, elements, and / or components, and do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0052] As used herein, the term "if" is interpreted, at the discretion of the context, to mean "when" or "in response to" or "in response to determining" or "in accordance with a determination" or "in response to detecting" a stated condition precedent is true. Similarly, the phrase "if it is determined that [a stated condition precedent is true]" or "if [a stated condition precedent is true]" or "when [a stated condition precedent is true]" is interpreted, at the discretion of the context, to mean "in response to determining" or "in response to determining" or "in accordance with a determination" or "in response to detecting" or "in response to detecting".

[0053] As used herein, and unless otherwise specified, the term "about" or "approximately" refers to an acceptable error for a particular value as determined by one of ordinary skill in the art, which depends in part on how the value is measured or determined. In certain embodiments, the term "about" or "approximately" refers to within 1, 2, 3, or 4 standard deviations. In certain embodiments, the term "about" or "approximately" refers to within 30%, 25%, 20%, 15%, 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, 0.5%, 0.1%, or 0.05% of a given value or range.

[0054] As used herein, the terms "comprises," "comprising," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus comprising a list of elements does not include only those elements, but may include other elements not expressly recited or inherent to such process, method, article, or apparatus.

[0055] As used herein, the terms "subject" and "patient" are used interchangeably. As used herein, the terms "subject" and "subjects" refer to animals (e.g., birds, reptiles, and mammals), including primates (e.g., monkeys, chimpanzees, and humans) and non-primates (e.g., camels, donkeys, zebras, cows, pigs, horses, cats, dogs, rats, and mice). In certain embodiments, the mammal is 0-6 months old, 6-12 months old, 1-5 years old, 5-10 years old, 10-15 years old, 15-20 years old, 20-25 years old, 25-30 years old, 30-35 years old, 35-40 years old, 40-45 years old, 45-50 years old, 50-55 years old, 55-60 years old, 60-65 years old, 65-70 years old, 70-75 years old, 75-80 years old, 80-85 years old, 85-90 years old, 90-95 years old, or 95-100 years old.

[0056] FIG. 1 is a flowchart of an example method 100 for classifying the tympanic membrane, according to some embodiments. At operation 110, the method 100 may include receiving one or more data sets related to the tympanic membrane from an interrogation system. At operation 120, the method 100 may include determining a set of parameters from the one or more data sets. At least one parameter of the set of parameters may be related to a dynamic property of the tympanic membrane. At operation 130, the method 100 may include outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters. The classification may comprise one or more of a state, condition, or mobility metric of the tympanic membrane.

[0057] Although the above operations illustrate a method 100 for classifying tympanic membranes according to some embodiments, one of ordinary skill in the art will recognize many variations based on the teachings described herein. Steps may be completed in any order. Steps may be added or deleted. Some of the steps may include sub-steps. Many of the steps may be repeated multiple times if beneficial to the manufacturing method.

[0058] One or more steps of method 100 may be implemented using one or more of the circuits as described herein, e.g., digital processing devices or processors or logic circuits such as programmable array logic for field programmable gate arrays. The circuits may be programmed to provide one or more steps of method 100, and the programs may comprise, e.g., program instructions stored on a computer readable memory or programmed steps of logic circuits such as programmable array logic or field programmable gate arrays. Embodiments, variations, and examples of digital processing devices operable to perform one or more steps of method 100 are described elsewhere herein, e.g., in section "Digital Processing Devices" and with respect to FIG. 10 described therein.

[0059] 2 is a schematic diagram illustrating an example system 200 for classifying the tympanic membrane, according to some embodiments. The system 200 may comprise one or more datasets 210 received from an interrogation system. The one or more datasets 210 may be stored in a memory of the system 200. The system 200 may comprise a classifier model 220 operable to classify the tympanic membrane based on the one or more datasets 210 and output a classification 230. The classification 230 may comprise one or more of a state, condition, or mobility metric of the tympanic membrane. An output of the classifier model (e.g., a classification) may be generated. The classification may comprise information about TM mobility, which can assist a physician in diagnosing an OM infection or directly diagnosing a patient's disease state.

[0060] For example, a system for classifying the tympanic membrane is disclosed herein. The system includes a computing system including a memory, which may include instructions for classifying the tympanic membrane, the computing system configured to execute instructions for at least receiving one or more data sets related to the tympanic membrane from an interrogation system, determining a set of parameters from the one or more data sets, at least one parameter of the set of parameters being related to a dynamic characteristic of the tympanic membrane, and outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters, the classification comprising a state, condition, or mobility metric of the tympanic membrane. Embodiments, variations, and examples of computing devices of the system for classifying the tympanic membrane are described elsewhere herein, for example, in section "Digital Processing Devices" and with reference to FIG. 10 described therein.

[0061] (Interrogation System) FIG. 3 is a schematic diagram illustrating an example of an interrogation system 300 as disclosed herein. The interrogation system 300 as disclosed herein may collect one or more data sets 210 related to a biological membrane (e.g., tympanic membrane) of a subject 320 as disclosed herein. The one or more data sets related to the tympanic membrane may comprise data related to physical properties of the biological membrane (e.g., tympanic membrane) to be characterized. In some examples, the interrogation system may comprise an imaging system. In some examples, the interrogation system does not comprise an imaging system. The interrogation system may collect ultrasound data, e.g., reflected ultrasound data. The interrogation system may collect optical coherence tomography data. The interrogation system may collect infrared data. The interrogation system may collect optical data. The interrogation system may collect data in response to a pneumatic excitation. The interrogation system may collect a data set related to membrane movement, for example, in response to a pneumatic excitation. The pneumatic excitation may comprise a pressure excitation, such as a puff of air. The interrogation system may measure pressure in the ear canal as a function of applied pressure. In one example, the interrogation system may measure the pressure in the ear canal following air pressure excitation.

[0062] Any example or combination of examples of an interrogation system as disclosed herein may be used, independently or in combination, to provide a data set from which to generate parameters to be included in a classifier model as disclosed herein. For example, any example or combination of examples of an interrogation system as disclosed herein may be used, independently or in combination, as an input to a machine learning algorithm to provide a classification output.

[0063] An interrogation system as disclosed herein may be remote from a classification system as disclosed herein. The interrogation system may collect one or more data sets (e.g., data), and the one or more data sets may be transmitted to a classification system as disclosed herein. In some cases, the one or more data sets may be received from a remote server. However, in some cases, the interrogation system may be local to a classification system as disclosed herein. For example, the classification system may be part of the on-board logic on a processor of the classification system, such as on the logic within an ultrasound tympanoscope.

[0064] In some cases, the interrogation system does not include an imaging system. For example, the interrogation system may include an ultrasound system. The ultrasound system may collect ultrasound data sets, e.g., Doppler ultrasound data. In some cases, the interrogation system collects frequency-dependent data, e.g., frequency-dependent optical coherence tomography data, absorption spectroscopy, depth-dependent ultrasound tomography data, etc. The ultrasound data may include data from a pneumatic ultrasound tympanoscope.

[0065] The interrogation system may comprise ultrasound data. The ultrasound data may comprise amplitude mode (a-mode) ultrasound data. The ultrasound data may comprise two-dimensional (b-mode) ultrasound data. The ultrasound data may comprise m-mode (motion mode) ultrasound data. The ultrasound data may comprise Doppler ultrasound data. The ultrasound data may comprise ultrasound m-mode phase versus time at a fixed gate depth. The ultrasound data may comprise m-mode amplitude versus time at a fixed gate depth.

[0066] In some examples, the interrogation system includes an imaging system. The imaging system may collect one or more data sets comprising one or more images of the tympanic membrane. In some examples, the images may comprise optical images, such as from a camera. In some examples, the images may comprise ultrasound images. The images may comprise optical coherence tomography images. The imaging system may comprise any other imaging modality, such as magnetic resonance imaging (MRI) or computed tomography (CT) or positron emission tomography (PET) or a combination of imaging modalities. Any of the imaging modalities may be optionally used with the pneumatic excitation. For example, an image may be collected in response to the pneumatic excitation. Some images may be collected before and after the pneumatic excitation. The one or more images may comprise ultrasound images. The one or more ultrasound images comprise images from a pneumatic ultrasound tympanoscope. The one or more ultrasound images may be measured in response to the pneumatic excitation. In some examples, the interrogation system may use a CCD camera to measure dynamic visual movement of the tympanic membrane. The optical image dataset responsive to the pneumatic excitation may comprise an optical integrated image intensity versus time waveform. The optical image data set responsive to the pneumatic excitation may comprise a light intensity cone versus time waveform.The optical image data set responsive to the pneumatic excitation may comprise a light position cone versus time waveform.The optical image data set responsive to the pneumatic excitation may comprise an image sharpness index versus time waveform.

[0067] In an example, the interrogation system as disclosed herein may comprise an embodiment, variation, or example of the methods and systems disclosed in U.S. Patent No. 7,771,356 and U.S. Patent Publication Nos. 2019 / 0365292, 2018 / 0310917, and 2017 / 0014053, each of which is incorporated herein by reference in its entirety. Methods and systems for obtaining information about tympanic membrane motion using ultrasound echo signals as disclosed in the incorporated references may be used to generate one or more parameters related to dynamic properties of the tympanic membrane. A system for measuring ultrasound echo signals may induce tympanic membrane motion by applying systematic pressure pulses and then extracting Doppler shift signals from the ultrasound to analyze the displacement of the TM and / or classify the viscosity of ear effusion.

[0068] In an example, an interrogation system as disclosed herein may comprise an embodiment, variation, or example of the methods and systems disclosed in commonly assigned U.S. Patent Publication Nos. 2019 / 0200873 and 2017 / 0360302, each of which is incorporated herein by reference in its entirety. Methods and systems for characterizing a membrane using optical coherence tomography (OCT) as disclosed in U.S. Patent Publication Nos. 2019 / 0200873 and 2017 / 0360302 may be used to generate one or more parameters related to dynamic properties of the tympanic membrane. For example, the dynamic properties of the tympanic membrane may comprise a phase delay or a time delay of a reflected optical signal in response to an applied air pressure excitation. OCT may be used to collect depth-dependent data related to the tympanic membrane. OCT may be used to collect frequency-dependent data, such as wavelength of absorption of the membrane or fluid adjacent to the membrane.

[0069] The interrogation system may collect data in response to pneumatic excitation. The interrogation system may, for example, collect data related to membrane movement in response to pneumatic excitation. The pneumatic excitation may comprise a pressure excitation, such as a puff of air. The pneumatic excitation may change the membrane's response to ultrasonic excitation. For example, the pneumatic excitation may cause the membrane to deflect, which may change the phase of the reflected ultrasonic wave for a membrane not exposed to the pneumatic excitation. The deflection of the membrane may comprise a damped harmonic motion. This motion may be affected by a change in the elasticity of the membrane. A change in membrane elasticity may occur, for example, when water, bacterial growth, or other foreign matter is adjacent to the membrane.

[0070] In some examples, the pneumatic excitation may cause a movement of the surface or membrane during a time interval. This interval may coincide with an acoustic wave delivered to the surface or membrane by an ultrasonic transmitter. The pneumatic excitation may be continuous, the pneumatic excitation may be pulsed, etc. The ultrasound reflected from the surface may be received at a transducer. The transducer may be the same transducer that generated the incident acoustic wave. The displacement of the surface or membrane may be related to a phase change of the received signal when compared to the transmitted signal. The movement of the membrane may affect the phase change of the received ultrasound. The displacement may vary with time. Analysis of the temporal displacement of the surface or membrane as measured by the phase shift of the reflected ultrasound in response to the pneumatic excitation coupled to the surface or membrane may be used to determine the mechanical properties of the surface or membrane.

[0071] Analysis of the temporal information may be used in combination with the measured temporal displacement from templates of other membrane responses to make comparisons. Analysis of the temporal information may be used in combination with other metrics related to the delay in amplitude of the reflected ultrasound to characterize the response of the surface or membrane. The mechanical properties measured may include ductility, elasticity, hardness, etc. Non-contact measurements of the mechanical properties of the surface, or alternatively, the fluid below the surface of the membrane, may be determined.

[0072] In some embodiments, the elasticity of the surface may be measured. The phase and / or amplitude of the reflected ultrasound from the membrane may be analyzed to generate an elasticity metric. The elasticity measurement may characterize a series of measurements in response to an applied excitation. The elasticity metric may be derived from the response of the surface and may provide an indication of one or more of several different phenomena. For example, the elasticity metric may indicate whether the surface adjacent to the membrane has a gaseous boundary or a fluid boundary. For example, the membrane may move less, more slowly, and / or not at all if the membrane has a fluid boundary. In some examples, the elasticity metric may indicate the extent or characteristics of the fluid with respect to characterizing the fluid behind the fluid boundary of the membrane. In some examples, the elasticity metric may be used to measure the properties of an elastic fluid with or without hysteresis in the response. In a fluid with a hysteretic response, the fluid may exhibit an offset, or "memory," in the displacement response, whereby the response behavior in one direction resembles the response behavior in the opposite direction only after traveling a certain displacement distance. For a hysteretic response, it may be necessary to characterize the linear behavior of the response after a particular measured displacement associated with the hysteresis of the system. A fluid elastic metric can be determined from the characteristic response of a surface or membrane to a surface excitation and from reflected ultrasonic characterization.

[0073] In some embodiments, the deflection of the surface can be estimated. For example, an estimate of the deflection of the surface can be derived from measured estimates of velocity, acceleration, or any other metric associated with the deflection over time. For example, a displacement of the surface will result in a shortened path from the transducer to the surface, and the reflected signal returning from the surface to the transducer will return with a phase shift. The phase shift of the reflected ultrasound relative to the excitation therefore gives information about the amount of deflection. With an estimate of the force exerted by the excitation, an estimate of the elasticity of the membrane can be estimated.

[0074] In some examples, the excitation is a step response or impulse response with a rising edge, a falling edge, or an impulse excitation. The impulse excitation initiates an oscillatory deflection of the membrane. The reflected ultrasound can be measured from the time of excitation through the decay period of the membrane's vibration. In some embodiments, an estimation of the position, elasticity, or viscosity can be performed by examination of the ring-down characteristics. For example, the ring-down characteristics can comprise at least one of an exponential decay time, or a ring cycle interval, or a ring cycle frequency. Decomposition of the response into ring-down characteristics, such as:

number

[0075] The damping constant of the oscillator may be related to the energy lost from the membrane to the surrounding environment. In some examples, if the membrane is adjacent to a fluid, the fluid may dampen the vibration of the membrane. The viscosity of the fluid may be related to the damping of the oscillator. The ring cycle frequency may be related to the restoring constant of the elastic membrane. The restoring constant may be related to the elasticity of the membrane. The restoring constant may be related to the viscosity of the fluid adjacent to the membrane. The ring cycle frequency may be higher for fluids with lower viscosity adjacent to the membrane.

[0076] Each excitation event may initiate a new deflection of the membrane. For example, an impulse excitation may pull the membrane in or push the membrane out for a limited period of time. For example, a square wave excitation may pull the membrane in or push the membrane out for a longer period of time. For example, a sinusoidal or other more complex excitation may be applied and the ringdown observed at the transducer may be a cross-correlation of the excitation field and the response field. Pneumatic excitation may be applied at frequencies below 100 kHz, below 1 kHz, below 100 Hz, below 10 Hz, below 1 Hz or below, or within the range given by any two of the preceding values. Pneumatic excitation may be applied at frequencies above 1 Hz, above 10 Hz, above 100 Hz, above 1 kHz, above 100 kHz or above, or within the range given by any two of the preceding values. Pneumatic excitation may be applied within the range of 10 Hz to 100 Hz.

[0077] FIG. 4 illustrates an example data set provided by an interrogation device of the present disclosure. FIG. 4A shows the intensity scale of the m-mode ultrasound signal. FIG. 4A uses ultrasound measurements to characterize the position of the tympanic membrane moving over a defined time range during the time course during which a pressure pulse is applied. The horizontal time scale in FIG. 4A is the same scale as in FIG. 4B, FIG. 4C, and FIG. 4D (FIG. 4D is omitted in its time range compared to FIG. 4B and FIG. 4C). FIG. 4B shows a plot of a horizontal slice of the heat map from FIG. 4A (indicated by the dashed line). FIG. 4A shows the signal amplitude at all depths, while FIG. 4B shows the membrane position change based on the signal phase change over time. FIG. 4B shows the tympanic membrane position versus the measured time. FIG. 4C shows the pressure pulse applied over the same time range in FIG. 4B. FIG. 4D shows the measurement of pressure in the ear canal during a pressure pulse spanning the first half of the time scale in FIG. 4C. The measurements of pressure in the ear canal during the second half of the time scale of FIG. 4C are plotted as a second curve in FIG. 4D and translated in time to overlap the data from the first half. These sensed pressure profiles in FIG. 4D are very close in value and difficult to distinguish from one another, indicating very similar movements of the tympanic membrane when pressure stimuli are repeated. FIG. 4E shows an exemplary optical image recorded by a CCD camera, where the dynamic movement of the white areas, i.e., the cones of light during the pressure pulse (highlighted by the black frame with arrows), indicates the mobility of the TM.

[0078] In summary, Figure 4 shows that when a pressure pulse is applied (Figure 4C), the generated and converted ultrasound signal provides detailed information about the tympanic membrane position (Figures 4A and 4B) and the unique pressure and optical responses from the images (Figures 4D) and (Figure 4E). The information contained within these measurements can be used to generate a set of parameters for training a classifier model, such as a machine learning model, to output a classification.

[0079] (Model parameters) A set of parameters may be derived from one or more data sets. At least one parameter of the set of parameters may be related to a dynamic property of the tympanic membrane. The set of parameters may also include parameters related to static features of the tympanic membrane (e.g., parameters not related to membrane motion). The set of parameters may include parameters related to membrane motion and parameters not related to membrane motion. The set of parameters may include static parameters and dynamic parameters. The set of parameters may include one or more ultrasound features, one or more pressure features, and / or one or more optical features. The set of parameters or a portion of the set of parameters may be used as input to a classifier model, such as a trained machine learning algorithm. The classifier model may output a classification. The classification may include a mobility output. The classification may include a disease state.

[0080] At least one parameter of the set of parameters may be related to a dynamic property of the tympanic membrane. The dynamic property of the tympanic membrane may change over time. For example, the dynamic property of the tympanic membrane may change over time in response to a pneumatic excitation, e.g., a puff of air or a puff of gas. A non-exhaustive list of potential parameters related to the dynamic property is as follows: an indication of membrane movement or membrane mobility, a minimum or maximum displacement of the tympanic membrane, an outlier displacement, a difference or ratio between the minimum and maximum displacements, a slope of the displacement or a slope of the difference or ratio between the minimum and maximum displacements to the pressure of the pneumatic excitation, a response of the measured pressure to the applied pressure, a visual movement of the tympanic membrane in response to the pneumatic excitation, one or more statistical components generated from singular value decomposition, principal component analysis, and K-means clustering, and an ultrasound pulse echo amplitude or ultrasound echo phase or a derivative or a moving average thereof. The dynamic parameters may be derived from normalized or averaged dynamic properties of the tympanic membrane in response to the dynamic excitation, such as a normalization or average of the parameters listed above.

[0081] Table 1 shows an exemplary list of parameters related to the dynamic properties of the tympanic membrane. The displacement of the tympanic membrane in response to an applied pneumatic excitation can be measured by the method of Figure 4. Alternatively, the displacement of the tympanic membrane in response to a pneumatic excitation can be measured by optical coherence tomography.

[0082] The parameters related to the dynamic properties of the tympanic membrane may comprise the minimum or maximum displacement of the tympanic membrane. The absolute distance the tympanic membrane moves may vary from patient to patient. Typically, larger displacements may be associated with higher mobility, while smaller displacements may be associated with lower mobility. These displacements may exhibit various classifications. For example, lower mobility may be associated with viscous fluid behind the membrane.

[0083] The parameters related to the dynamic properties of the tympanic membrane may comprise an outlier displacement. In some cases, the measured displacement may be large or small enough to move outside the measurement range of the ultrasound device. This "outlier" displacement may be treated as a categorical "yes" / "no" feature. A large outlier may typically indicate a highly mobile tympanic membrane. A small outlier may typically indicate a highly immobile tympanic membrane.

[0084] A parameter related to the dynamic properties of the tympanic membrane may comprise the difference or ratio between the minimum and maximum displacement. The difference between the minimum and maximum displacement may be related to the static retraction or distension of the tympanic membrane. The static retraction or distension may be measured by measuring the relationship between the distance traveled by the membrane during + and - pressure pulses. A distended membrane (towards the interrogation device) may be indicated by a higher mobility during positive pressure cycles and a lower mobility during negative pressure cycles. A retracted membrane may be indicated by a lower mobility during positive pressure cycles and a higher mobility during negative pressure cycles. Such measurements may be used to extract static membrane parameters, e.g., a distended membrane or a retracted membrane.

[0085] Parameters related to the dynamic properties of the tympanic membrane may comprise the variation of the displacement characteristics, including normalizing any of the displacement characteristics by the amplitude of the measured pressure pulse, as the movement will depend at least in part on the applied pressure stimulus, and any other linear or nonlinear transformation using the displacement and the measured pressure. This metric may be bidirectional, e.g., characterization of the tympanic membrane mobility may be more complete by knowing the metric under positive and negative pressure stimuli.

[0086] Parameters related to the dynamic properties of the tympanic membrane may include the slope of the displacement, the slope of the difference between the minimum and maximum displacement versus the pressure of the pneumatic excitation, or the slope of the ratio between the minimum and maximum displacement versus the pressure of the pneumatic excitation. The speed at which the membrane returns to its original position after it has been displaced by a pressure pulse may indicate membrane mobility. The speed may be measured by the derivative of the displacement measurement as described herein. Second or higher derivatives may also be utilized. The damping coefficient of the ring-down feature as described herein may also be related to the restoring force of the membrane.

[0087] The parameter related to the dynamic properties of the tympanic membrane may comprise a response of measured pressure versus applied pressure. A measured pressure below the applied pressure may indicate membrane mobility. It may also indicate a break in the instrument seal, which would allow errors in the procedure to be identified. The parameter related to the dynamic properties of the tympanic membrane may comprise a slope of the pressure response. The slope of the pressure may directly correlate to the mobility of the tympanic membrane.

[0088] The tympanic membrane movement may be visually observed via a CCD camera during pressure stimulation. These visual movements and displacements (rate, magnitude) may be converted into input parameters or features to train a classifier. For example, a user (e.g., a medical professional) may indicate an observation of membrane mobility. The user may indicate that the membrane is qualitatively mobile or immobile. The user may indicate mobility on a 0-4 point scale. For example, after pressure is applied in the ear canal, a normal tympanic membrane may move vigorously and be classified as 4+, while a partially or completely compromised tympanic membrane may be classified as 3+, 2+, 1+, or no mobility to represent the reduced degree of mobility on a rank scale. For example, a medical provider may qualitatively indicate the vigor of the tympanic membrane, e.g., how the membrane bounces back quickly in response to the applied excitation. Pressure assessment may be performed by a clinician during pneumatic otoscopy.

[0089] Other parameters can be derived from other types of ultrasound signals that may result from different pre-processing of the ultrasound signal, including, but not limited to, bandpass filters at different frequencies, different sequences of pulses and measurements, transformation of the (I,Q) Doppler signal, and data processing steps. Various non-limiting examples include: a bandpass filter can be applied in the "fast time domain" of the ultrasound signal capture to remove noise artifacts that are outside the passband of the ultrasound transducer, thereby improving the signal-to-noise ratio; the phase of the (I,Q) Doppler signal at the depth of the tympanic membrane can be integrated to derive a membrane displacement versus time parameter; the displacement versus time parameter can be analyzed to reveal the response during changes in ear canal stimulus pressure, and evaluated in terms of slope (activeness or inactivity) and position at the start and end of the pressure stimulus change (indicating the total change in position); the Hilbert transform can be used to assess the amplitude of the ultrasound reflection from the tympanic membrane, and the optical under pressure stimulus as a further indication of the mobility of the tympanic membrane. It can be evaluated at the time of stimulus pressure change to detect changes in cone orientation; membrane movement directionality can be evaluated as a prerequisite for a priori understanding, for example, that under a decrease in pressure in the ear canal, the tympanic membrane may move toward the ultrasound transducer and vice versa under an increase in pressure in the ear canal; the signal-to-noise ratio of reflections from the tympanic membrane can be evaluated to qualify the resulting mobility measurement as valid or not; and multiple adjacent gates covering a wider depth range than the tympanic membrane signal can be identified and monitored in parallel to best track the tympanic membrane signal during a time frame positioned in front of the transducer and evaluate mobility from the best of the gate depths with a qualified tympanic membrane signal.

[0090] Parameters related to the dynamic properties of the tympanic membrane may be derived from the Doppler ultrasound data. For example, the tympanic membrane Doppler signal as a function of time may include ultrasound echo amplitude and ultrasound echo phase. These parameters may depend on a particular pulse-echo gate depth. The parameters related to the ultrasound echo amplitude may comprise data related to the tympanic membrane acoustic cross section and the direction of the tympanic membrane surface normal. The parameters related to the ultrasound echo phase may comprise the time derivative of the ultrasound echo phase. The ultrasound echo phase may be proportional to the membrane velocity.

[0091] The parameters related to the dynamic properties of the tympanic membrane may comprise features derived from principal component analysis, K-means, or similar cluster algorithms on any of the above datasets. For example, any linear or nonlinear transformation of static data (e.g., principal component analysis, K-means clustering) may be used to extract one or more parameters to reduce collinearity of features or to cluster similar groups of samples based on their respective features. PCA and K-means may be used on various datasets to create the parameters. PCA and K-means may be attempted when collinearity between features creates redundant information in the dataset. PCA and K-means may provide parameters that may reduce variance in predictions or improve accuracy. In the case of PCA, principal components may be used as training features, either alone or in combination with non-transformed features. Any number of principal components may be selected (e.g., only the top 2-3 or N number of components until an explained variance such as 95% is met). For K-means clustering, features such as "clusters" may be derived and categorical training features consisting of the clusters into which each sample is classified may be used as parameters in the training model.

[0092] Additional features over time are generated including moving averages and exponential moving averages over different window lengths. [Table 1]

[0093] The set of parameters may include parameters related to static properties of the membrane. For example, an interrogation system generating optical measurements may allow for extraction of parameters based on visual features. Visual indicators of otitis media may include hue, transparency, and / or bulging. Other static parameters may include color, opacity, translucency, and position. Visual features from optical measurements may be indicated by a user and represented by parameters, for example, visible or invisible bulging.

[0094] In another example, a parameter related to the static position characteristics of the tympanic membrane may be derived from an optical coherence tomography method. For example, the parameter may be related to the depth characteristics of the tympanic membrane. For example, if the ear contains a low viscosity infectious exudate, an initial peak in the tympanic membrane optical coherence tomography signal may generate an axial region of extended reflection with reduced amplitude from the optical attenuation of the reflected signal. For example, if the ear contains a bacterial infection, a bacterial membrane may be present on the opposite surface of the tympanic membrane, which may generate a larger axial range of reflection, followed by a high scattering coefficient and a corresponding increase in attenuation. For example, the parameter may be related to the wavelength of the absorption feature of the tympanic membrane. If the tympanic membrane is found to absorb at wavelengths corresponding to bacterial exudate, viral exudate, or no exudate, the wavelength of absorption or the identity of the absorption feature may constitute a parameter of the parameter set.

[0095] (classification) The methods and systems of the present disclosure may output a classification (e.g., output) of the tympanic membrane. The classification may be based on a classifier model as disclosed herein, where the classifier may be a machine learning algorithm. The set of parameters or a portion of the set of parameters may be used as input to a classifier model, such as a trained machine learning algorithm. The classifier model may output a classification. The classification may include a mobility output. The classification may include a disease state. The classification may comprise one or more of a state, condition, or mobility metric of the tympanic membrane. An output of the classifier model may be generated. The output may comprise information about TM mobility, which may assist a physician in diagnosing an otitis media infection or directly diagnosing a patient's disease state.

[0096] 5A and 5B are schematic diagrams of a set of parameters derived from pressure, ultrasound, and optical measurements that may constitute inputs 510 to a classifier model 520 for outputting a classification, according to some embodiments. As shown in FIG. 5A, the amplitude of reflected ultrasound, the slope of the ultrasound response after pressure excitation, the measured pressure versus the applied pressure, the slope of the pressure response, and the optical activity may constitute inputs to the classifier. The classifier may output a mobility metric 530. As shown in FIG. 5A, the mobility metric 530 may comprise a binary classification, such as mobile or non-mobile. As shown in FIG. 5B, the amplitude of reflected ultrasound, the slope of the ultrasound response after pressure excitation, the measured pressure versus the applied pressure, the slope of the pressure response, and the optical activity may constitute inputs to the classifier. The classifier may output a state or condition 531, such as acute otitis media, otitis media with effusion, and no effusion, as shown.

[0097] In some cases, the classification may comprise a mobility metric of the tympanic membrane. The mobility metric may comprise one or more of a binary classification, a rank output, a categorical rank output, or a continuous numerical value. The classification may include various forms of tympanic membrane information, which may be provided to a health care provider (e.g., a physician). The classification may be based on one or more of ultrasound, pressure, and optical measurement features.

[0098] In some examples, the classification comprises a categorical readout. The categorical readout may comprise an indication of the degree of membrane mobility as at least one of highly mobile, moderately mobile, semi-mobile, or non-mobile. For example, the categorical classification may comprise highly mobile, moderately mobile, semi-mobile, or non-mobile. In some examples, the categorical readout comprises a binary classification. The binary classification may comprise mobile or non-mobile. The binary classification may comprise normal mobility or abnormal mobility. In one example, the simple classification may be a binary classifier, and a binary choice based on the maximum probability of either "mobile" or "non-mobile" is displayed to the physician.

[0099] In some examples, the classification comprises a category rank classification. The category rank classification may comprise a 0-4+ scale as described herein above. In another example, the category rank output may comprise a 0-10 point scale, e.g., 10.0 being highly mobile, 5.0 being slightly mobile, and 0 being non-mobile (or vice versa). The user may indicate that the membrane is qualitatively mobile or immobile. The user may indicate mobility on a 0-4 point scale. For example, after pressure is applied within the ear canal, a normal tympanic membrane may be actively mobile and classified as 4+, while a partially or completely compromised tympanic membrane may be classified as 3+, 2+, 1+, or no mobility to represent the reduced degree of mobility on a rank scale. Pressure assessment may be performed by a clinician during pneumatic otoscopy.

[0100] In some embodiments, the classification comprises a continuous numeric output comprising one or more of the measured membrane displacement, the rate of membrane movement, or the rate of membrane recovery. In some examples, the rank readout comprises a numeric scale related to the degree of membrane mobility. The continuous metric may comprise a numeric scale with decimated increments. In some examples of continuous numeric outputs, a continuous variable regression algorithm may be used to generate a continuous prediction, e.g., the percent likelihood of a particular state or condition. In some examples, the state or condition may be normalized to a value or scale. For example, the continuous metric may be normalized such that the maximum value is 10 and the minimum value is zero, with 10 associated with a highly mobile membrane and 0 associated with an immobile membrane, or vice versa.

[0101] In some examples, multi-class classification may be used. For example, a binary classification (e.g., mobile or immobile) may be determined, and various degrees of immobility may be classified (e.g., 0 vs. 1+ vs. 2+ vs. 3+ on a four-point scale) where immobility is most likely. In another example, the classification may comprise an expected probability of each possible category. The expected probability may comprise the maximum probability that is chosen as the classification displayed to the health care provider. For example, the probability may be calculated using strategies such as a softmax function or one-vs-all classification.

[0102] The classification may indicate the state or condition of the tympanic membrane to a user (e.g., a health care provider). In some cases, the classification may comprise the state or condition of the tympanic membrane. The state or condition of the tympanic membrane may comprise one or more of acute otitis media, acute otitis media with effusion, middle ear effusion, bacterial infection, viral infection, no effusion, and unknown classifications. Classification of an ear indicating acute otitis media may include detection of the presence of effusion and characterization of the type of effusion as serous, mucous, purulent, or one of a combination thereof. In acute otitis media, the middle ear effusion may be induced by a pathogen, and in viral infection, it may be thin or serous, and in bacterial infection, it may be thicker and purulent.

[0103] (Classifier model) 6 is a schematic diagram illustrating an example classifier model 600 comprising a machine learning algorithm, according to some embodiments. The classifier model may be generated based on a set of parameters 610. The set of parameters 610 may include ultrasound and / or pressure and / or optical features. For example, the set of parameters may constitute an input to the classifier model, and the classification may be an output of the classifier model. The classifier model may output a classification 630. The classification may include a mobility output. The classification may include a disease state.

[0104] The classifier model may comprise a machine learning algorithm 620. For example, the machine learning algorithm may comprise one or more of linear regression, logistic regression, classification and regression tree algorithms, support vector machines (SVMs), naive Bayes, K-nearest neighbors, random forest algorithms, boosting algorithms such as XGBoost and LightGBM, neural networks, convolutional neural networks, and recurrent neural networks. The machine learning algorithm may be a supervised learning algorithm, an unsupervised learning algorithm, or a semi-supervised learning algorithm.

[0105] Machine learning algorithms may be used to make predictions using a set of parameters. One class of machine learning algorithms, artificial neural networks (ANNs), may form part of the classifier model. For example, feed-forward neural networks (such as convolutional neural networks or CNNs) and recurrent neural networks (RNNs) may be used. A neural network binary classifier may be trained by comparing the predictions made by its underlying machine learning model with the ground truth. An error function calculates the discrepancy between the predicted value and the ground truth, and this error is iteratively back-propagated through the neural network over multiple cycles or epochs to modify a set of weights that affect the value of the predicted output. Training ends when the predicted value meets a convergence condition, such as obtaining a small magnitude of the calculated error. Multiple layers of neural networks may be employed to create a deep neural network. Using a deep neural network may increase the predictive power of the neural network algorithm. In some cases, machine learning algorithms using neural networks may further include Adam optimization (e.g., adaptive learning rate), regularization, etc. The number of layers, the number of nodes in a layer, the stride length in a convolutional neural network, padding, filters, etc. may be adjustable parameters in a neural network.

[0106] Additional machine learning algorithms and statistical models may be used to obtain insights from the parameters disclosed herein. Additional machine learning methods that may be used are logistic regression, classification and regression tree algorithms, support vector machines (SVM), naive Bayes, K-nearest neighbors, and random forest algorithms. These algorithms may be used for many different tasks, including data classification, clustering, density estimation, or dimensionality reduction. Machine learning algorithms may be used for active learning, supervised learning, unsupervised learning, or semi-supervised learning tasks. In the present disclosure, various statistical, machine learning, or deep learning algorithms may be used to generate an output based on a set of parameters.

[0107] Machine learning algorithms may use a supervised learning approach. In supervised learning, an algorithm may generate a function or model from training data. The training data may be labeled. The training data may include metadata associated with it. Each training example of the training data may be a pair consisting of at least an input object and a desired output value. Supervised learning algorithms may require a user to determine one or more control parameters. These parameters may be adjusted by optimizing performance on a portion of the training data, for example, a validation set. After parameter adjustment and learning, the performance of the resulting function / model may be measured on a test set, which may be separate from the training set. Regression methods may be used in supervised learning approaches.

[0108] In some embodiments, the supervised machine learning algorithms may include, but are not limited to, neural networks, support vector machines, nearest neighbor interpolators, decision trees, boosted decision stumps, boosted versions of such algorithms, derived versions of such algorithms, or combinations thereof. In some embodiments, the machine learning algorithms may include one or more of a Bayesian model, a decision graph, an inductive logic programming, a Gaussian process regression, a genetic programming, a kernel estimator, a minimum message length, a multilinear subspace learning, a naive Bayes classifier, a maximum entropy classifier, a conditional random field, a minimum complexity machine, a random forest, an ensemble of classifiers, and a multi-criteria classification algorithm.

[0109] Machine learning algorithms may use semi-supervised learning approaches that can combine both labeled and unlabeled data to generate appropriate functions or classifiers.

[0110] In some embodiments, the machine learning algorithm may use an unsupervised learning approach. In unsupervised learning, the algorithm may generate functions / models to describe hidden structures (i.e., classifications or categories that cannot be indicated, observed, or calculated) from unlabeled data. Because the examples given to the learner are not labeled, there is no assessment of the accuracy of the structures output by the associated algorithm. Approaches to unsupervised learning include clustering, anomaly detection, and neural networks.

[0111] Machine learning algorithms may use reinforcement learning approaches, in which algorithms can learn policies for how to act given observations of the world: every action may have some effect in the environment, and the environment may provide feedback that guides the learning algorithm.

[0112] Although the machine learning algorithms discussed above can generate much more complex relationships between input features and output indicators, some simple examples of relationship algorithms are shown by the illustrative decision trees of FIG.

[0113] 7A, 7B, and 7C are schematic diagrams illustrating example decision trees that may be generated by a machine learning algorithm, according to some embodiments. In some embodiments, the machine learning algorithm builds a decision tree based on a set of parameters. The decision tree may comprise some threshold of parameters in the set of parameters. The threshold may determine the branch of the tree into which the dataset should be classified. The decision tree may comprise the greatest information gain at the top nodes of the tree. The decision tree may be built, in part, by searching for the most informative nodes (e.g., parameters) for a given dataset.

[0114] In some examples, the decision tree is pruned. Pruning may include reducing the number of parameters in the parameter set to a fraction of the most relevant parameters. The fraction of parameters may be a minimum number of parameters to classify the data set within a specified sensitivity or specificity, for example, 90% sensitivity or specificity. In some embodiments, the decision tree comprises J48, C4.5, or ID3. In some embodiments, the decision tree comprises ADABoost or DecisionStump.

[0115] As shown in Figure 7A, the first exemplary tree comprises a single parameter (e.g., absolute value of displacement). If the parameter is above a threshold of 0.2 mm, the membrane is classified as mobile. If the parameter is below the threshold of 0.2 mm, the membrane is classified as non-mobile.

[0116] As shown in FIG. 7B, the second exemplary tree comprises three parameters (e.g., absolute value of displacement with fixed pressure stimulation amplitude, absolute value of displacement with fixed time measurement window duration, and outlier displacement). The classification of the model is related to the mobility of the tympanic membrane. If the absolute value of the displacement with fixed pressure stimulation amplitude is above a threshold of 0.1 and the outlier displacement is above 0.5, the membrane is highly mobile. If the absolute value of the displacement with fixed pressure stimulation amplitude is above a threshold of 0.1 and the outlier displacement is below 0.5, the membrane is semi-mobile. If the absolute value of the displacement with fixed pressure stimulation amplitude is below a threshold of 0.1 and the absolute value of the displacement with fixed time measurement window duration is above 1.5, the membrane is almost immobile. If the absolute value of the displacement with fixed pressure stimulation amplitude is below a threshold of 0.1 and the absolute value of the displacement with fixed time measurement window duration is below 1.5, the membrane is completely immobile.

[0117] As shown in FIG. 7C, the third exemplary tree comprises three parameters (e.g., absolute value of displacement with fixed pressure stimulus amplitude, absolute value of displacement with fixed time measurement window duration, and outlier displacement). The classification of the model is related to the state or condition of the tympanic membrane. If the absolute value of the displacement with fixed pressure stimulus amplitude is above a threshold of 0.1 and the outlier displacement is above 0.5, the membrane is characterized as not having any effusion. If the absolute value of the displacement with fixed pressure stimulus amplitude is above a threshold of 0.1 and the outlier displacement is below 0.5, the membrane is characterized as having otitis media with effusion. If the absolute value of the displacement with fixed pressure stimulus amplitude is below a threshold of 0.1 and the absolute value of the displacement with fixed time window duration is above 1.5, the membrane is characterized as having otitis media with effusion. If the absolute value of the displacement with a fixed pressure stimulus amplitude falls below a threshold of 0.1 and the absolute value of the displacement with a fixed time window duration falls below 1.5, the membrane is characterized as having acute otitis media.

[0118] (Training a Computer-Implemented Classifier) FIG. 8 illustrates a flowchart of an exemplary method of training a computer-implemented classifier, according to some embodiments. At operation 810, the method 800 may include receiving a set of parameters based on one or more datasets associated with one or more tympanic membranes and one or more classified datasets associated with one or more tympanic membranes. The classification may comprise a state, condition, or mobility metric of the tympanic membrane, and the set of parameters comprises at least one parameter related to a dynamic characteristic of the one or more tympanic membranes. At operation 820, the method 800 may include storing the set of parameters and the one or more classified datasets in a database. At operation 830, the method 800 may include constructing a classifier model based on the set of parameters and the one or more classified datasets. The classifier model may be derived from the set of parameters, and the classifier model outputs a classification based on a dataset of the one or more classified datasets. At operation 840, the method 800 may include using the classifier model to provide a classification of the unclassified dataset.

[0119] In some cases, the method further includes updating the database based on the second one or more classified data sets. In some cases, the method further includes updating the classifier model based on the second one or more classified data sets. The method may further include using the classifier model to provide a classification of the unclassified data set by the method of classifying the tympanic membrane described herein. The classifier may be continuously updated. For example, the method may use online gradient descent or stochastic gradient descent methods to continuously update the model parameters for improved detection. In some cases, the classifier may be updated with the entire data set. The classifier may be updated in several instalments with recent samples.

[0120] Although the above acts illustrate a method 800 for training a computer-implemented classifier according to some embodiments, one of ordinary skill in the art will recognize many variations based on the teachings described herein. Steps may be completed in any order. Steps may be added or deleted. Some of the steps may include sub-steps. Many of the steps may be repeated multiple times if beneficial to the manufacturing process.

[0121] One or more steps of method 800 may be implemented using one or more of the circuits as described herein, e.g., digital processing devices or processors or logic circuits such as programmable array logic for field programmable gate arrays. A circuit may be programmed to provide one or more steps of method 800, and a program may comprise, for example, program instructions stored on a computer readable memory or programmed steps of logic circuits such as programmable array logic or field programmable gate arrays. Embodiments, variations, and examples of digital processing devices operable to perform one or more steps of method 800 are described elsewhere herein, e.g., in section "Digital Processing Devices" and with respect to FIG. 10 described therein.

[0122] FIG. 9 is a schematic diagram illustrating an example of a system 900 for training a computer-implemented classifier, according to some embodiments. The system of FIG. 9 may comprise an example of a system operable to perform the method 800 as shown in FIG. 8. The system 900 may comprise one or more classified datasets (e.g., historical data). The one or more classified datasets may comprise one or more datasets related to one or more tympanic membranes. The classified datasets may comprise data with a known classification. For example, the classified datasets may comprise a dataset from a biopsied tympanic membrane. The classified datasets may have an associated set of parameters including at least one parameter based on a dynamic property of the tympanic membrane. The classified datasets may have an associated classification. In some cases, a set of parameters (e.g., features) may be extracted from one or more classified datasets.

[0123] In the lower loop of the system 900, all or a portion of one or more classified datasets may be used as training data. A set of parameters for the classified datasets may be used as part of an inductive learning algorithm to train a classifier model to find a classification. For example, all or a portion of the classified dataset may be used as a training dataset for the inductive learning algorithm.

[0124] As shown in the upper loop of the system 900, a portion of the classified data set may be reserved as a validation data set (e.g., test data). The validation data may have, for example, a set of established classifications by experiment. The validation data set may be used to test the predictions of the classifier model. The validation data set may be used to validate a model trained based on the training data set. The training data may be used as part of an engineering design testing loop. For example, a model or set of models (e.g., an "ensemble") may be determined using a validation procedure, where the model is trained based on a portion of the available data (e.g., training data) and then used to predict the remaining portion of the available data (e.g., test data). Accuracy metrics (e.g., root mean square error for numerical data, precision and recall for categorical data, area under the receiver operating curve (AUC), etc.) may be defined to classify the degree to which a given model is accurate with respect to the provided data set. In some examples, over- or under-sampling (e.g., synthetic minority oversampling technique (SMOTE)) may be used during model training as part of model building.

[0125] The risk of a machine learning model being unable to generalize to data that the model has never encountered before can be mitigated by using as many representative training datasets as possible. In addition, utilizing techniques such as hyperparameter tuning along with cross-validation to maximize model generalization can improve model generalization. Briefly, hyperparameter tuning can modify aspects of a given model to prevent either bias (i.e., underfitting) or variance (i.e., overfitting) on ​​the training and validation data. In addition, cross-validation can be used to train successively on different portions of the data (e.g., K-fold and leave-one-out), and the accuracy predictions generated from each portion can be averaged to create an overall accuracy metric for a given model.

[0126] The trained classifier model may then be subjected to testing based on an unclassified data set. For example, the unclassified data set may comprise one or more tympanic membranes that are not in the set of classified data. The classifier model may be used to generate a classification based on a set of parameters extracted from the unclassified data set. The classifier model may be used to classify the membrane as, for example, one of highly mobile, slightly mobile, low mobile, or non-mobile.

[0127] Systems for training a computer-implemented classifier may comprise embodiments, variations, and examples of digital processing devices described elsewhere in this specification, for example, in section "Digital Processing Devices" and with respect to FIG. 10 described therein. (Digital Processing Device)

[0128] In some embodiments, the devices, systems, and methods of use thereof described herein include a digital processing device or the use thereof. For example, the digital processing device may be used to control various aspects of the devices and methods disclosed herein. For example, the digital processing device may be used to perform a method of classifying a tympanic membrane. The digital processing device may comprise a computing system, e.g., the computing system may comprise a memory, the memory may comprise instructions for classifying a tympanic membrane. The digital processing device may be configured to perform one or more steps of the method of classifying a tympanic membrane. The digital processing device may be configured to perform one or more steps of method 100 or method 800 as disclosed herein. The digital processing device may be configured to control an interrogation system, such as any example, variation, or example of an interrogation system as disclosed herein. The digital processing device may receive and / or read one or more data sets from the interrogation system. The digital processing device may comprise a database management system for one or more data sets. The digital processing device may perform one or more steps for training a computer-implemented classifier as disclosed herein.

[0129] In further embodiments, the digital processing device includes one or more hardware central processing units (CPUs), general purpose graphics processing units (GPGPUs), or field programmable gate arrays (FPGAs) that perform the functions of the device. In still further embodiments, the digital processing device further comprises an operating system configured to execute the executable instructions. In some embodiments, the digital processing device may be optionally connected to a computer network. In further embodiments, the digital processing device is optionally connected to the Internet such that it accesses the World Wide Web. In still further embodiments, the digital processing device is optionally connected to a cloud computing infrastructure. In other embodiments, the digital processing device is optionally connected to an intranet. In other embodiments, the digital processing device is optionally connected to a data storage device.

[0130] According to the description herein, suitable digital processing devices include, by way of non-limiting examples, server computers, desktop computers, laptop computers, notebook computers, sub-notebook computers, netbook computers, netpad computers, set-top computers, media streaming devices, handheld computers, Internet appliances, mobile smart phones, tablet computers, personal digital assistants, video game consoles, and vehicles. Those skilled in the art will recognize that many smart phones are suitable for use in the systems described herein. Those skilled in the art will also recognize that optional televisions, video players, and digital music players with optional computer network connections are suitable for use in the systems described herein. Suitable tablet computers include those with booklet, slate, and convertible configurations known to those skilled in the art.

[0131] In some embodiments, the digital processing device includes an operating system configured to execute the executable instructions. An operating system is software, including programs and data, that manages the device's hardware and provides services for the execution of applications.

[0132] In some embodiments, the device includes a storage and / or memory device. A storage and / or memory device is one or more physical devices used to temporarily or permanently store data or programs. In some embodiments, the device is a volatile memory and requires power to maintain the stored information. In some embodiments, the device is a non-volatile memory and retains the stored information when the digital processing device is not powered. In further embodiments, the non-volatile memory comprises a flash memory. In some embodiments, the non-volatile memory comprises a dynamic random access memory (DRAM). In some embodiments, the non-volatile memory comprises a ferroelectric random access memory (FRAM®). In some embodiments, the non-volatile memory comprises a phase change random access memory (PRAM). In other embodiments, the device is a storage device, including, by way of non-limiting examples, CD-ROMs, DVDs, flash memory devices, magnetic disk drives, magnetic tape drives, optical disk drives, and cloud computing-based storage devices. In further embodiments, the storage and / or memory device is a combination of devices such as those disclosed herein.

[0133] In some embodiments, the digital processing device includes a display to transmit visual information to a user. In some embodiments, the display is a cathode ray tube (CRT). In some embodiments, the display is a liquid crystal display (LCD). In further embodiments, the display is a thin film transistor liquid crystal display (TFT-LCD). In some embodiments, the display is an organic light emitting diode (OLED) display. In various further embodiments, there is a passive matrix OLED (PMOLED) or active matrix OLED (AMOLED) display on an OLED display. In some embodiments, the display is a plasma display. In other embodiments, the display is a video projector. In still further embodiments, the display is a combination of devices such as those disclosed herein.

[0134] In some embodiments, the digital processing device includes an input device to receive information from a user. In some embodiments, the input device is a keyboard. In some embodiments, the input device is a pointing device, including, by way of non-limiting examples, a mouse, a trackball, a trackpad, a joystick, a game controller, or a stylus. In some embodiments, the input device is a touch screen or a multi-touch screen. In other embodiments, the input device is a microphone for capturing voice or other sound input. In other embodiments, the input device is a video camera or other sensor for capturing motion or visual input. In further embodiments, the input device is a Kinect, Leap Motion, or the like. In still further embodiments, the input device is a combination of devices such as those disclosed herein.

[0135] Referring to FIG. 10, in a particular embodiment, an exemplary digital processing device 1001 is programmed or otherwise configured to control or implement the systems and methods for classifying tympanic membranes and systems and methods for training computer-implemented classifiers as described herein. The device 1001 may coordinate various aspects of the systems and methods for classifying tympanic membranes and systems and methods for training computer-implemented classifiers of the present disclosure, such as, for example, performing processing steps. In this embodiment, the digital processing device 1001 includes a central processing unit (CPU, also referred to herein as "processor" and "computer processor") 1005, which may be a single-core or multi-core processor or multiple processors for parallel processing. The digital processing device 1001 also includes memory or memory locations 1010 (e.g., random access memory, read-only memory, flash memory), an electronic storage unit 1015 (e.g., hard disk), a communication interface 1020 (e.g., network adapter) for communicating with one or more other systems, and peripheral devices 1025, such as cache, other memory, data storage, and / or electronic display adapters. The memory 1010, storage unit 1015, interface 1020, and peripheral devices 1025 communicate with the CPU 1005 through a communication bus (solid lines) such as a motherboard. The storage unit 1015 may be a data storage unit (or data repository) for storing data. The digital processing device 1001 may be operatively coupled to a computer network ("network") 1030 using the communication interface 1020. The network 1030 may be the Internet, an intranet and / or an extranet, or an intranet and / or an extranet that communicates with the Internet. The network 1030 in some cases is a telecommunications and / or data network. The network 1030 may include one or more computer servers, which may enable distributed computing such as cloud computing.Network 1030 may, in some cases, implement a peer-to-peer network with device 1001 that may allow devices coupled to device 1001 to behave as clients or servers.

[0136] Continuing to refer to FIG. 10, the CPU 1005 can execute a sequence of machine-readable instructions that may be embodied in a program or software. The instructions may be stored in a memory location, such as memory 1010. The instructions may be directed to the CPU 1005, which may then program or otherwise configure the CPU 1005 to implement the methods of the present disclosure. Examples of operations performed by the CPU 1005 include fetch, decode, execute, and writeback. The CPU 1005 may be part of a circuit, such as an integrated circuit. One or more other components of the device 1001 may be included in the circuit. In some cases, the circuit is an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

[0137] Continuing with reference to FIG. 10, the storage unit 1015 can store files such as drivers, libraries, and saved programs. The storage unit 1015 can store user data, such as user preferences and user programs. The digital processing device 1001 in some cases can include one or more additional data storage units that are external, such as located on a remote server that communicates through an intranet or the Internet. The digital processing device 1001 can communicate with one or more remote computer systems through a network 1030. For example, the device 1001 can communicate with a user's remote computer system.

[0138] Examples of remote computer systems include a personal computer (e.g., a portable PC), a slate or tablet PC (e.g., Apple® iPad®, Samsung® Galaxy Tab), a telephone, a smartphone (e.g., Apple® iPhone®, Android-enabled devices, Blackberry®), or a personal digital assistant.

[0139] Methods as described herein may be implemented using machine (e.g., computer processor) executable code stored on electronic storage locations of the digital processing device 1001, such as, for example, on the memory 1010 or electronic storage unit 1015. The machine executable or machine readable code may be provided in the form of software. In use, the code may be executed by the processor 1005. In some cases, the code may be read from the storage unit 1015 and stored on the memory 1010 for quick access by the processor 1005. In some circumstances, the electronic storage unit 1015 may be omitted and machine executable instructions are stored on the memory 1010.

[0140] The digital processing device 1001 may include or communicate with an electronic display 1035 having a user interface (UI) 1040. Examples of UIs include, but are not limited to, graphical user interfaces (GUIs) and web-based user interfaces. In some cases, the electronic display 1035 may be connected to the computer system 1001 via a network, such as via network 1030.

[0141] In some embodiments, the platform, system, medium, and method disclosed herein include one or more non-transitory computer-readable storage media encoded with a program including instructions executable by an operating system of a digital processing device, which is optionally networked. In further embodiments, the computer-readable storage medium is a tangible component of the digital processing device. In yet further embodiments, the computer-readable storage medium is optionally removable from the digital processing device. In some embodiments, the computer-readable storage medium includes, by way of non-limiting examples, CD-ROMs, DVDs, flash memory devices, solid-state memories, magnetic disk drives, magnetic tape drives, optical disk drives, cloud computing systems and services, and the like. In some cases, the programs and instructions are encoded on the medium permanently, substantially permanently, semi-permanently, or non-transitoryly.

[0142] In some embodiments, the platform, system, medium, and method disclosed herein include at least one computer program or the use thereof. A computer program includes a sequence of instructions executable in a CPU of a digital processing device that is written to perform a specified task. Computer-readable instructions can be implemented as program modules, such as functions, objects, application programming interfaces (APIs), data structures, etc., that perform a particular task or implement a particular abstract data type. In light of the disclosure provided herein, those skilled in the art will recognize that computer programs can be written in various versions of various languages.

[0143] The functionality of the computer readable instructions may be combined or distributed as desired in various environments. In some embodiments, the computer program comprises a single sequence of instructions. In some embodiments, the computer program comprises multiple sequences of instructions. In some embodiments, the computer program is provided from one location. In other embodiments, the computer program is provided from multiple locations. In various embodiments, the computer program includes one or more software modules. In various embodiments, the computer program includes, in part or in whole, one or more web applications, one or more mobile applications, one or more standalone applications, one or more web browser plug-ins, extensions, add-ins, or add-ons, or a combination thereof.

[0144] In some embodiments, the platforms, systems, media, and methods disclosed herein include software, server, and / or database modules, or uses thereof. In view of the disclosure provided herein, the software modules are created by techniques known to those skilled in the art using machines, software, and languages ​​known in the art. The software modules disclosed herein are implemented in numerous ways. In various embodiments, the software modules comprise files, sections of code, programming objects, programming structures, or combinations thereof. In further various embodiments, the software modules comprise multiple files, multiple sections of code, multiple programming objects, multiple programming structures, or combinations thereof. In various embodiments, one or more software modules comprise, by way of non-limiting examples, web applications, mobile applications, and standalone applications. In some embodiments, the software modules are in one computer program or application. In other embodiments, the software modules are in two or more computer programs or applications. In some embodiments, the software modules are hosted on one machine. In other embodiments, the software modules are hosted on two or more machines. In further embodiments, the software modules are hosted on a cloud computing platform. In some embodiments, the software modules are hosted on one or more machines in one location. In other embodiments, the software modules are hosted on one or more machines in two or more locations.

[0145] In some embodiments, the platforms, systems, media, and methods disclosed herein include one or more databases or the use thereof. In view of the disclosure provided herein, one skilled in the art will recognize that many databases are suitable for storing and retrieving data sets from an interrogation system, storing classified data sets, determining parameters from one or more data sets, storing parameters associated with classified data sets, and the like. In various embodiments, suitable databases include, as non-limiting examples, relational databases, non-relational databases, object-oriented databases, object databases, entity-relationship model databases, associative databases, and XML databases. Further non-limiting examples include SQL, PostgreSQL, MySQL, Oracle, DB2, and Sybase. In some embodiments, the database is Internet-based. In further embodiments, the database is web-based. In yet further embodiments, the database is cloud computing-based. In other embodiments, the database is based on one or more local computer storage devices.

[0146] While preferred embodiments of the present disclosure have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions will now occur to those skilled in the art without departing from the scope of the present disclosure. It is understood that various alternatives to the embodiments of the present disclosure described herein may be employed in practicing the invention of the present disclosure. The following claims define the scope of the present invention, and it is intended that methods and structures within the scope of these claims and their equivalents be covered thereby.

Claims

[Claim 1] The invention described in this specification.

Images