WO2019092723A1

WO2019092723A1 - System and method for determining pathological status of a laryngopharyngeal area in a patient

Info

Publication number: WO2019092723A1
Application number: PCT/IL2018/051214
Authority: WO
Inventors: Zeev Vladimir VOLKOVICH; Tal MARSHAK; Elena KLEIMAN; Katerina KORENBLAT; Forsan JAHSHAN; Ohad RONEN
Original assignee: Ofek Eshkolot Research And Development Ltd.; Health Corporation Of Galilee Medical Center
Priority date: 2017-11-12
Filing date: 2018-11-11
Publication date: 2019-05-16

Abstract

A system for determining a pathological condition in the laryngopharyngeal area of a patient, the system comprising: an image acquisition module configured to acquire images of a laryngopharyngeal area of a patient, connected to: a processor configured to process images acquired by the image acquisition module, analyze the images and determine the pathological condition in the laryngopharyngeal area of a patient, wherein the system is configured to: capture at least one image of the laryngopharyngeal area of a patient; select at least one image from the captured images; and determine the pathological condition of the laryngopharyngeal area by analyzing the selected at least one image. Additional embodiments of the system and methods performed by the system are described herein.

Description

SYSTEM AND METHOD FOR DETERMINING PATHOLOGICAL STATUS OF A LARYNGOPHARYNGEAL AREA IN A PATIENT

CROSS-REFERENCE TO RELATED APPLICATION

This application claims priority to United States Provisional Patent Application No. 62/584,840, filed November 12, 2017, the entire contents of which is incorporated herein by reference in its entirety. FIELD

The present subject matter relates to systems and methods for image analysis of pathological conditions. More particularly, the present subject matter relates to systems and methods for determining the pathological status of a laryngopharyngeal area in a patient based on image analysis.

BACKGROUND

The laryngopharyngeal area may suffer from several pathological conditions including inflammation from different causes, for example laryngopharyngeal reflux.

Laryngopharyngeal reflux, designated hereinafter "LPR", is the regurgitation of gastric contents onto the mucosal linings of the pharynx, larynx, and upper aero-digestive tract, resulting in a spectrum of nonspecific symptoms. The presence of acid and pepsin in this sensitive region causes a variety of physiological responses, such as laryngeal edema and erythema, mucosal hypertrophy, granuloma, and subglottic stenosis. These physical signs are often considered during the determination of the status of LPR in a patient along with common symptoms such as throat clearing, persistent cough, globus sensation, and changes in voice quality. Due to the non-specificity of these symptoms, determination of the status of LPR is currently fairly subjective and often inaccurate, leading to significant numbers of false positive results and false negative results (see, Fritz MA, Persky MJ, et al, The Accuracy of the Laryngopharyngeal Reflux Diagnosis: Utility of the Stroboscopic Exam Otolaryngol. Head Neck Surg. 2016 Jun 14. pii: 0194599816655143). In order to properly identify symptomatic patients, Koufman et al published the reflux symptom Index, designated hereinafter "RSI", questionnaire. In the RSI questionnaire, patients were asked to answer different questions regarding their symptoms, a score exceeding 9 indicates symptomatic patients[l l, 12]. Unfortunately, in practice, practitioners often do not perform the entire RSI questionnaire, and the answers that the patients give when scoring their own symptoms are highly subjective and not necessarily accurate.

The Reflux Finding Score, designated hereinafter "RFS", is another well-accepted alternative method for the measurement of laryngeal signs which is similarly subjective. Patients are evaluated according to the laryngeal signs found in an office-performed direct or fiberoptic laryngoscopy. A score exceeding 7 indicates the presence of LPR [12, 14]. RFS scoring, if performed well, is a clinically validated tool to determine the status of LPR, but it has its own drawbacks, including:

the physician that performs the laryngoscopy has to be a highly experienced otolaryngologist in order tocorrectly determine the status of LPR;

it may be impacted by the limited time allotted for physical examination of patients in many clinical settings;

it does not archive results for future objective reference or follow the progress of treatment; and,

low inter-rater reliability and reproducibility, even among otolaryngologists when evaluating the same laryngeal signs during laryngoscopic exam, and even if performed by the same person at different times, as this evaluation is subjective and performer-dependent [6].

Currently considered the "gold-standard" method for determining the status of LPR is to perform a 24-hour ambulatory pH monitoring by placement of two probes in the aero- digestive tract [7,8]. Some studies have reported low sensitivity for identifying the presence of acid reflux in the proximal esophagus [9] . Possible issues which may account for low overall efficacy of this method include: incorrect placement of the proximal probe; displacement of the probe during the 24-hour measurement; the irregularity of the regurgitation of gastric contents during the test, discomfort of the prolonged and invasive nature of the test; the unavailability of the test at all institutions; and, its high cost [6,10]. These apparently all combine to limit its widespread clinical adoption. Upon the determination of the status of LPR, the LPR patient is typically treated with medications to reduce gastric acid levels and allow the inflamed tissue to heal. Treatment is long term, often continuing for months or even years. In the absence of an objective and convenient method of determining the status of LPR, physicians often err on the side of misguided caution, prescribing medication, even if the results are inconclusive.

A result of prescribing medication for a patient determined incorrectly as having LPR is long-term intake of medication that causes inconvenience and unnecessary expenditures as well as being harmful for the patients' health. Recent studies show that reducing the acidity level in the stomach can interfere with normal digestive function, causing vulnerability to infections [22-27]. Long-term intake of antacid medications from the proton pump inhibitor (PPI) group leads to malabsorption of key minerals in the body, namely calcium and magnesium, possibly causing osteoporosis and cardiac abnormalities needlessly [14-21]. There is a real need for an affordable, objective system that permits tracking and comparison of patients' response to LPR treatment, even for cases where the status of LPR is correctly determined and medication is clinically justifiable. There is also a need for a tool that provides a definite indication to stop medication treatment in cases where the patient is cured. References:

[1] Belafsky, P.C., et al., Symptoms and findings of laryngopharyngeal reflux. Ear Nose Throat J, 2002. 81(9 Suppl 2): p. 10-3.

[2] DeVault, K.R., Overview of therapy for the extraesophageal manifestations of

gastroesophageal reflux disease. Am J Gastroenterol, 2000. 95(8 Suppl): p. S39-44.

[3] Postma, G.N., L.F. Johnson, and J. A. Koufman, Treatment of laryngopharyngeal reflux.

Ear Nose Throat J, 2002. 81(9 Suppl 2): p. 24-6. [4] Koufman, J.A., M.R. Amin, and M. Panetti, Prevalence of reflux in 113 consecutive patients with laryngeal and voice disorders. Otolaryngol Head Neck Surg, 2000. 123(4): p. 385-8.

[5] Khan, A.M., et al., Laryngopharyngeal reflux: A literature review. Surgeon, 2006. 4(4): p. [6] Branski, R.C., N. Bhattacharyya, and J. Shapiro, The reliability of the assessment of endoscopic laryngeal findings associated with laryngopharyngeal reflux disease.

Laryngoscope, 2002. 112(6): p. 1019-24.

[7] Rees, CJ. and P.C. Belafsky, Laryngopharyngeal reflux: Current concepts in

pathophysiology, diagnosis, and treatment. Int J Speech Lang Pathol, 2008. 10(4): p. 245- 53.

[8] Postma, G.N., Ambulatory pH monitoring methodology. Ann Otol Rhinol Laryngol Suppl, 2000. 184: p. 10-4.

[9] Vaezi, M.F., P.L. Schroeder, and J.E. Richter, Reproducibility of proximal probe pH parameters in 24-hour ambulatory esophageal pH monitoring. Am J Gastroenterol, 1997.

92(5): p. 825-9.

[10] Ulualp, S.O. and R.J. Toohill, Laryngopharyngeal reflux: state of the art diagnosis and treatment. Otolaryngol Clin North Am, 2000. 33(4): p. 785-802.

[11] Belafsky, P.C, G.N. Postma, and J.A. Koufman, Validity and reliability of the reflux symptom index (RSI). J Voice, 2002. 16(2): p. 274-7.

[12] Habermann, W., et al., Reflux symptom index and reflux finding score in

otolaryngologic practice. J Voice, 2012. 26(3): p. el23-7.

[13] Cohen, J.T., Z. Gil, and D.M. Fliss, [The reflux symptom index— a clinical tool for the diagnosis of laryngopharyngeal reflux]. Harefuah, 2005. 144(12): p. 826-9, 912. [14] Belafsky, P.C, G.N. Postma, and J.A. Koufman, The validity and reliability of the reflux finding score (RFS). Laryngoscope, 2001. 111(8): p. 1313-7.

[15] Khalili, H, Huang E.S., Jacobson B.C., et al., Use of proton pump inhibitors and risk of hip fracture in relation to dietary and lifestyle factors: a prospective cohort study. BMJ. 2012; 344:e372.

[16] Yang Y.X., Lewis J.D., Epstein S., Metz D.C., Long-term proton pump inhibitor therapy and risk of hip fracture. JAMA. 2006; 296:2947-2953.

[17] Targownik L.E., Lix L.M., Metge C.J., et al., Use of proton pump inhibitors and risk of osteoporosis-related fractures. CMAJ. 2008; 179:319-326.

[18] Targownik L.E., Leslie W.E., Davison K.S., et al., The relationship between proton pump inhibitor use and longitudinal change in bone mineral density: a population-based study from the Canadian Multicentre Osteoporosis Study (CaMos). Am J Gastroenterol. 2012; 107: 1361-1369.

[19] Proton pump inhibitors: risk of bone fractures. Health Canada. April 4, 2013.

http://healthycanadians.gc.ca/recall-alert-rappel-avis/hc-sc/2013/26523a-eng.php.

Accessed August 27, 2013.

[20] Danziger J., William J.H., Scott D.J., et al., Proton-pump inhibitor use is associated with low serum magnesium concentrations. Kidney Int. 2013; 83:692-699.

[21] Hess M.W., Hoenderop J.G., Bindels R.J., Drenth J.P., Systematic review:

hypomagnesaemia induced by proton pump inhibition. Aliment Pharmacol Ther. 2012; 36:405-413. [22] Dial S., Delaney J.A., Barkun A.N., Suissa S., Use of gastric acid-suppressive agents and the risk of community-acquired Clostridium difficile-associated disease. JAMA. 2005 294:2989-2995.

[23] Linsky A., Gupta K., Lawler E.V., et al., Proton pump inhibitors and risk for recurrent Clostridium difficile infection. Arch Intern Med. 2010; 170:772-778.

[24] Gulmez S.E., Holm A., Frederiksen H., et al., Use of proton pump inhibitors and the risk of community-acquired pneumonia: a population-based case-control study. Arch Intern Med. 2007; 167:950-955. [25] Sarkar M., Hennessy S., Yang Y.X., Proton-pump inhibitor use and the risk for community-acquired pneumonia. Ann Intern Med. 2008; 149:391-398. [26] Laheij R.J., Sturkenboom M.C., Hassing R.J., et al. Risk of community- acquired pneumonia and use of gastric acid-suppressive drugs. JAMA. 2004; 292: 1955-1960.

[27] de Jager CP., Wever P.C., Gemen E.F., et al., Proton pump inhibitor therapy

predisposes to community-acquired Streptococcus pneumoniae pneumonia. Aliment Pharmacol Ther. 2012; 36:941-949.

[28] Witt D.R., Chen H., Mielens J.D., McAvoy K.E., Zhang F., Hoffman M.R., Jiang J.J., Detection of chronic laryngitis due to laryngopharyngeal reflux using color and texture analysis of laryngoscopic images. J Voice. 2014 Jan; 28(1):98-105.

[29] Marcelja, S. Mathematical description of the responses of simple cortical cells.

Journal of the Optical Society of America. 1980; 70 (11): 1297-1300.

[30] J. G. Daugman. Uncertainty relation for resolution in space, spatial frequency, and orientation optimized by two-dimensional visual cortical filters. Journal of the Optical

Society of America, 1985 Jul; 2(7): 1160-1169.

[31] Haghighat, M.; Zonouz, S.; Abdel-Mottaleb, M. (2013). "Identification Using

Encrypted Biometrics". Computer Analysis of Images and Patterns. Lecture Notes in Computer Science 8048. p. 440. doi: 10.1007/978-3-642-40246-3_55.

SUMMARY

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this subject matter belongs. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present subject matter, suitable methods and materials are described below. In case of conflict, the patent specification, including definitions, will control. In addition, the materials, methods, and examples are illustrative only and not intended to be limiting.

According to one aspect of the present subject matter, there is provided a system for determining a pathological condition in the laryngopharyngeal area of a patient, the system comprising:

an image acquisition module configured to acquire images of a laryngopharyngeal area of a patient, connected to:

a processor configured to process images acquired by the image acquisition module, analyze the images and determine the pathological condition in the laryngopharyngeal area of a patient,

wherein the system is configured to:

capture at least one image of the laryngopharyngeal area of a patient;

select at least one image from the captured images; and

determine the pathological condition of the laryngopharyngeal area by analyzing the selected at least one image. According to one embodiment, the system further comprises a memory 40 configured to store data and images.

According to another embodiment, the system is additionally configured to store data and images for objective follow-up of patients and follow-up of patients during treatment of the pathological condition.

According to yet another embodiment, the at least one image is at least one stills image.

According to still another embodiment, the at least one image is at least one video image.

According to a further embodiment, the at least one image is a Red-Green-Blue (RGB) color image, or Hue-Saturation- Value (HSV) color image, or Hue-Saturation-Lightness (HSV) color image. According to yet a further embodiment, the selecting at least one image from the captured images comprises selecting an image with open vocal folds.

According to still a further embodiment, the selecting at least one image from the captured images comprises selecting an image with closed vocal folds.

According to an additional embodiment, the selecting at least one image from the captured images further comprises selecting the most focused images. According to yet an additional embodiment, the pathological condition is

Laryngopharyngeal reflux (LPR).

According to still an additional embodiment, the selecting an image for diagnosis from the captured images comprises selecting at least one image from the at least one image that are suitable for applying eight algorithms, one for each Reflux Finding Score (RFS) sign.

According to one embodiment, the analyzing the selected at least one image comprises: evaluating RFS for each RFS sign on the selected at least one image; and

calculating a total RFS score by summing the scores of each RFS sign.

According to another embodiment, the evaluating RFS for each RFS sign on the selected at least one image comprises:

segmenting of the selected at least one image according to different anatomical areas of the larynx specific for each of the eight RFS signs;

extracting color data and texture data from respective segmented parts of the at least one image for obtaining extracted data; and

determining an RFS according to the extracted data.

According to yet another embodiment, the determining an RFS according to the extracted data comprises determining conditions of RFS signs and giving scores accordingly, as follows:

if Subglottic Edema is present giving a score of 2 and if Subglottic Edema is absent giving a score of 0; if Ventricular Obliteration is absent giving a score of 0, if Ventricular Obliteration is partial giving a score of 2 and if Ventricular Obliteration is complete giving a score of 4; if Erythema or Hyperemia is absent giving a score of 0, if there are arytenoids only giving a score of 2 and if Erythema or Hyperemia is diffuse giving a score of 4;

if Vocal Fold Edema is absent giving a score of 0, if Vocal Fold Edema is mild giving a score of 1 , if Vocal Fold Edema is moderate giving a score of 2, if Vocal Fold Edema is severe giving a score of 3 and if Vocal Fold Edema is polypoid giving a score of 4; if Diffuse Laryngeal Edema is absent giving a score of 0, if Diffuse Laryngeal Edema is mild giving a score of 1, if Diffuse Laryngeal Edema is moderate giving a score of 2, if Diffuse Laryngeal Edema is severe giving a score of 3 and if Diffuse Laryngeal Edema is obstructing giving a score of 4;

if Posterior Commissure Hypertrophy is absent giving a score of 0, if Posterior Commissure Hypertrophy is mild giving a score of 1 , if Posterior Commissure

Hypertrophy is moderate giving a score of 2, if Posterior Commissure Hypertrophy is severe giving a score of 3 and if Posterior Commissure Hypertrophy is obstructing giving a score of 4;

if Granuloma or Granulation is present giving a score of 2 and if Granuloma or Granulation is absent giving a score of 0; and

if Thick Endolaryngeal Mucus is present giving a score of 2 and if Thick

Endolaryngeal Mucus is absent giving a score of 0.

According to still another embodiment, it is determined that the patient has LPR when a sum of the RFS scores is bigger than 7. BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments are herein described, by way of example only, with reference to the accompanying drawings. With specific reference now to the drawings in detail, it is stressed that the particulars shown are by way of example and for purposes of illustrative discussion of the preferred embodiments, and are presented in the cause of providing what is believed to be the most useful and readily understood description of the principles and conceptual aspects of the embodiments. In this regard, no attempt is made to show structural details in more detail than is necessary for a fundamental understanding, the description taken with the drawings making apparent to those skilled in the art how several forms may be embodied in practice. In the drawings:

- Fig. 1 schematically illustrates, according to an exemplary embodiment, a block diagram of a system 1 for determining a pathological status of a laryngopharyngeal area in a patient. - Fig. 2 illustrates, according to an exemplary embodiment, a laryngoscope image of a healthy normal laryngopharynx of the kind which could be used in an exemplary embodiment of a collection of baseline images.

- Fig. 3A schematically illustrates, according to an exemplary embodiment, a flow diagram of an algorithm for open vocal fold image selection that selects images suitable for further analysis from a set of patient's laryngoscopic images.

- Fig. 3B schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for selecting an image with closed vocal folds, following the open vocal fold image selection illustrated in Fig. 3A, suitable for further analysis.

- Fig. 3C schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for selecting the most focused images with open vocal folds, suitable for further analysis.

- Fig. 4A illustrates, according to an exemplary embodiment, choosing a macro-block in a laryngoscope image for the block matching algorithm.

- Fig. 4B illustrates, according to an exemplary embodiment, sliding a macro block found as in Fig. 4A in a compared laryngoscope image, in search for the closest matching block, so as to check similarity of images.

- Fig. 5 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for marking of areas of interest L and R.

- Figs. 6A-B illustrate, according to an exemplary embodiment, annotated reference laryngoscope images to be used for marking of areas of interest A, B, L and R.

- Fig. 7 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for identifying the presence of subglotic edema in images selected from a patient's laryngoscopic video, following the algorithm for open vocal fold image selection shown in Fig. 3A.

- Fig. 8 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for illustration for an algorithm for determining the presence of subglotic edema. - Fig. 9 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for identifying the presence and level of ventricular obliteration in images selected from a patient's laryngoscopic video, following the algorithm for open vocal fold image selection shown in Fig. 3A.

- Fig. 10 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for analyzing erythema/hyperemia.

- Fig. 11 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for identifying the presence and level of vocal fold edema in an image selected from a patient's laryngoscopic video, following the algorithm for open vocal fold image selection shown in Fig. 3 A.

- Fig. 12 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for identifying the presence and level of diffuse laryngeal edema in an image selected from a patient's laryngoscopic video, following the open/closed vocal folds pair selection process described in Fig. 3B.

- Fig. 13 schematically illustrates according to an exemplary embodiment, a flow diagram of a method for identifying the presence and level of posterior commissure hypertrophy in an image selected from a patient's laryngoscopic video, following the general selection process of images described in Fig. 3A.

- Fig. 14 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for illustration for an algorithm for determining the presence and level of posterior commissure hypertrophy.

- Fig. 15A schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for identifying the presence of initial stages of granuloma in an image selected from a patient's laryngoscopic video, based on the algorithm for open vocal fold image selection shown in Fig. 3A.

- Fig. 15B schematically illustrates another embodiment of a flow diagram of a method for identifying the presence of advanced stages of granuloma in an image selected from a patient's laryngoscopic video, based on the algorithm for open vocal fold image selection shown in Fig. 3A.

- Fig. 16 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for illustration for an algorithm for identifying the presence of granuloma. - Fig. 17 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for illustration for an algorithm for identifying the presence of thick endolaryngeal mucus.

- Fig. 18 schematically illustrates, according to an exemplary embodiment, a flow diagram showing the process for identifying the presence of thick endolaryngeal mucus in a set of images suitable for analysis.

DESCRIPTION OF THE PREFERRED EMBODIMENTS Before explaining at least one embodiment in detail, it is to be understood that the subject matter is not limited in its application to the details of construction and the arrangement of the components set forth in the following description or illustrated in the drawings. The subject matter is capable of other embodiments or of being practiced or carried out in various ways. Also, it is to be understood that the phraseology and terminology employed herein is for the purpose of description and should not be regarded as limiting. In discussion of the various figures described herein below, like numbers refer to like parts. The drawings are generally not to scale.

For clarity, non-essential elements were omitted from some of the drawings.

One aim of the present subject matter is to provide a system and method for determining the pathological condition in the laryngopharyngeal area of a patient. It should be noted that the term "pathological condition in the laryngopharyngeal area" as disclosed herein is defined as any type of inflammation in the laryngopharyngeal area, for example LPR.

Another aim of the present subject matter is to provide a system and method for introducing repeatable objectivity into the determination and use of RFS.

Yet another aim of the present subject matter is to provide a system and method for reducing the number of false positive results and unnecessary or inappropriate administration of medication to patients incorrectly determined as suffering from a pathological condition in the laryngopharyngeal area, including an inflammation, for example LPR. This issue is of crucial importance, since administration of such medications, to patients suffering from a pathological condition in the laryngopharyngeal area, especially to patients who are mistakenly determined as suffering from LPR, may have some harmful undesired consequences as some of the medications may have harmful influences and side effects on the patient. Yet another aim of the present subject matter is to provide a system and method for improving the accuracy and appropriateness of correct medication treatment of a pathological condition in the laryngopharyngeal area, including an inflammation, for example LPR.

Still another aim of the present subject matter is to provide a system and method for developing, preferably by highly ranked world class specialists in laryngopharyngeal pathology, particularly LPR specialists, an objective, and statistically significant, image database. This image database is aimed to be easily accessible to practitioners by using the system and method of the present subject matter. Thus, a patient's laryngoscope images can be compared, thus allowing determination of the status of a pathological condition in the laryngopharyngeal area, including an inflammation, for example LPR in, and high-level treatment of, patients, for example in underdeveloped countries, that otherwise may not be accessible to.

A further aim of the present subject matter is to provide a system and method for developing an objective, statistically significant, and easily accessible image database against which a patient's laryngoscope images can be compared by a health provider with less experience and training than that of a narrow profiled specialist.

Yet a further aim of the present subject matter is to provide a system and method for developing an objective and easily accessible image database permitting a patient's pre- treatment laryngoscope images to be compared with the patient' s own intra- and post-treatment laryngoscope images.

Still a further aim of the present subject matter is to provide a system and method for developing an objective database containing data related to pathological conditions in the laryngopharyngeal area of a patient, including status of an inflammation, for example LPR and scores for each RFS sign, since the object is to track changes in the pathological condition during treatment, including changes in a condition of an inflammation, for example LPR, and in each RFS sign during treatment, that enables comparison of the degree of severity of the pathological condition, including an inflammation condition, for example LPR, based on the accumulated data, including numerical data of RFS signs, for example during treatment of a patient, for example in order to track the progress of the treatment. The present subject matter provides a system and method for determining the pathological condition in the laryngopharyngeal area of a patient. More particularly, the present subject matter provides a system and method for computerized image acquisition, database creation and objective determination the status of laryngopharyngeal reflux for pre-, intra- and post-treatment follow-up, for patients presenting potential LPR.

Fig. 1 schematically illustrates, according to an exemplary embodiment, a block diagram of a system 1 for determining a pathological condition in the laryngopharyngeal area of a patient. For the sake of simplicity only, the system 1 for determining the pathological condition in the laryngopharyngeal area of a patient is occasionally designated hereinafter "system 1 ". According to one embodiment, the pathological condition is at least one inflammation episode in the laryngopharyngeal area. According to another embodiment, the inflammation episode in the laryngopharyngeal area is laryngopharyngeal reflux (LPR).

According to one embodiment, the system 1 comprises an image acquisition module 10 configured to acquire images of a laryngopharyngeal area of a patient, connected, with a communication line 60, to a processor 20 configured to process images acquired by the image acquisition module 10, analyze the images and determine the pathological condition in the laryngopharyngeal area of a patient, as described herein. According to one embodiment, the image acquisition module 10 comprises a laryngoscope 102, a light source 104 and an image acquiring device 106, designated in Fig. 1 as camera 106. Any type of laryngoscope 102, light source 104 and image acquiring device 106 is under the scope of the present subject matter. According to one embodiment, the image acquiring device 106 is configured to acquire stills images of the laryngopharyngeal area of a patient. According to another embodiment, the image acquiring device 106 is configured to acquire video images of the laryngopharyngeal area of a patient. According to an additional embodiment, any type of processor 20 is under the scope of the present subject matter. According to a further embodiment, data from the image acquisition module 10, for example data related to images acquired by the image acquisition module 10, are transferred to the processor 20 through the communication line 60. According to yet a further embodiment, data may be transferred from the processor 20 to the image acquisition module 10, for example data related to the operation of the image acquisition module 10. Thus, according to one embodiment, communication line 60 is bidirectional, as illustrated in Fig. 1, and configured to transfer data from the image acquisition module 10 to the processor 20 and vice versa. According to another embodiment, communication line 60 is unidirectional (not shown) and configured to transfer data from the image acquisition module 10 to the processor 20. The processor 20 comprises at least one software that is configured to process images acquired by the image acquisition module 10, analyze the images and diagnose the pathological status of a laryngopharyngeal area of a patient, for example an inflammation in the laryngopharyngeal area, such as LPR and the like. This at least one software is designated hereinafter "diagnostic software".

According to one embodiment, the system 1 further comprises a learning database module 30, configured to store and operate a learning database and participate in the diagnosis process as described herein. Any learning database module 30 known in the art is under the scope of the present subject matter. The learning database module 30 is connected with a communication line 70 to the processor 20. According to one embodiment, communication line 70 is bidirectional and configured to transfer data from the processor 10 to the learning database module 30 and vice versa. According to one embodiment, the system 1 further comprises a memory 40 configured to store data accumulated during the analysis and diagnosis processes, as described herein, as well as images acquired by the image acquisition module 10. Any memory known in the art is under the scope of the present subject matter, for example a flash disk, a hard drive, a cloud- type drive, and the like. The memory 40 is connected with a communication line 80 to the processor 20. According to one embodiment, communication line 80 is bidirectional and configured to transfer data from the processor 10 to the memory 40 and vice versa.

According to one embodiment, the system 1 further comprises a display 50 configured to display images acquired by the image acquisition module 10 and results of analysis and diagnosis as described herein. The display 50 is connected with communication line 90 to the processor 20. Any type of display 50 known in the art is under the scope of the present subject matter, for example but not limited to, a monitor, a mobile device screen and the like. Furthermore, the display 50 may be part of any component of the system 1, for example, part of the image acquisition module 10, the processor 20, and the like. In addition, the display 50 may be physically positioned either in the vicinity of components of the system 1 , or remotely separated from the other component of the system 1. According to one embodiment, communication line 90 is bidirectional, as illustrated in Fig. 1. Thus, communication line 90 is configured to transfer data from the processor 20 to the display 50, for example data related to results of diagnosis that are to be displayed. In addition, communication line 90 is configured to transfer data from the display 50 to the processor 20, for example data related to the operation of the processor 20, for example in a case when the display 50 is a touch screen. According to another embodiment, communication line 90 is unidirectional (not shown) and configured to transfer data from the processor 20 to the display 50.

According to one embodiment, communication lines 60, 70, 80 and 90 are any communication lines known in the art. According to another embodiment, communication line 60 and/or 70 and/or 80 and/or 90 is wired. According to yet another embodiment, communication line 60 and/or 70 and/or 80 and/or 90 is wireless. According to still another embodiment, any combination of wired and wireless communication lines in a given configuration of the system 1 is under the scope of the present subject matter.

According to one embodiment, all the components of the system 1 , namely the image acquisition module 10, the processor 20, the learning database module 30, the memory 40, the display 50, and the communication lines 60, 70, 80 and 90, are physically separated one from the other. According to another embodiment, all the components of the system 1 are physically together. According to yet another embodiment, any combination of the components of the system 1 that is physically together is under the scope of the present subject matter. For example, a display 50 may be physically connected to a processor 20 while the other components of the system 1 are physically separated; or a processor 20 may be physically connected to an image acquisition module 10 while the other components of the system 1 are physically separated; and the like.

Additional embodiments of the system 1 include the following:

According to one embodiment, the image acquiring device 106 is configured to acquire stills images of a larynx of a patient. According to another embodiment, the image acquiring device 106 is configured to acquire video images of a larynx of a patient. According to one embodiment, the learning database module 30 is configured to store and operate a validated reference database comprising a collection of various types of information, for example color, texture and the like, derived from healthy and deviant laryngoscopic images against which a processor 20 is configured to compare subsequent laryngoscopic images from actual patients.

According to further embodiment, the processor 20 is configured to receive a set of stills images. According to a preferred embodiment, the processor 20 is configured to receive a video image.

According to one embodiment, the processor 20 is configured to receive the set of stills images and/or the video image from the memory 40, for example for diagnosing previously saved still images and/or video images. According to an additional embodiment, the processor 20 is configured to automatically select suitable images from either a set of stills images or a video image; perform unsupervised, or supervised, segmentation of at least one laryngoscopic image according to different anatomical areas of the larynx, for example different anatomical areas of the larynx specific for each of eight reflux finding score (RFS) signs. Furthermore, for some RFS signs, relevant data, for example color, texture, shape and the like from respective segmented parts of the at least one laryngoscopic image have to be compared with references, so as to yield comparison data; and a score for the considered sign is determined. For other RFS signs, a score is determined according to methods described hereinafter. Finally, RFS is determined by summing up separated signs' scores. In other words, the obtained data allow the evaluation of an RFS sign - whether it is present or not, or what is the condition of the RFS sign, as described hereinafter in relation to Table 1.

It should be noted that, as described above, the selection of suitable images from either a set of stills images or a video image is, according to one embodiment, fully automatic, without the involvement of a human. According to another embodiment, the selection of suitable images from either a set of stills images or a video image is made by a human, for example a professional in the field of otorhinolaryngology, and then the rest of the analysis is conducted by the system 1 according to the algorithms described herein. It should be noted that the system 1 described above is configured to determine a pathological condition in the laryngopharyngeal area of a patient according to the embodiments of the method described hereinafter. A method for determining a pathological condition in the laryngopharyngeal area of a patient comprises in general the following steps:

capturing at least one image of the laryngopharyngeal area of a patient;

selecting at least one image from the captured images; and

determining the pathological condition of the laryngopharyngeal area by analyzing the selected at least one image.

According to one embodiment, the at least one image is at least one stills image. According to another embodiment, the at least one image is at least one video image. According to yet another embodiment, the at least one image is a combination of at least one stills image and at least one video image. In other words, the at least one image may be at least one stills image, or at least one video image, or any combination thereof.

According to one embodiment, the pathological condition of the laryngopharyngeal area is inflammation. According to another embodiment, the inflammation is laryngopharyngeal reflux (LPR). According to this embodiment, the analyzing the selected at least one image comprises:

Evaluating Reflux Finding Score (RFS) signs on the selected at least one image, and calculating an RFS score. Thus, the method for determining the presence of laryngopharyngeal reflux (LPR) in a patient comprises the following steps:

capturing at least one image of the laryngopharyngeal area of a patient;

selecting at least one image for diagnosis from the captured at least one image;

evaluating RFS signs on the selected at least one image, and

calculating an RFS score.

Here is a detailed description of these steps. According to one embodiment, the selecting an image for diagnosis from the captured images comprises selecting at least one image from the at least one image that are suitable for applying eight algorithms, one for each RFS sign. According to one embodiment, the evaluating RFS signs on the selected at least one image comprises:

unsupervised segmenting of the selected at least one image according to different anatomical areas of the larynx specific for each of the eight RFS signs; extracting color data and texture data from respective segmented parts of the at least one image;

comparing extracted color data and texture data with a reference, so as to yield comparison data; and

determining an RFS according to the comparison data. According to one embodiment, the at least one image is at least one stills image, or at least one video image, or any combination thereof.

According to another embodiment, the selecting images from the at least one image is performed by an operator. According to a preferred embodiment, the selecting images from the at least one image is automatic.

It should be noted that as mentioned above, for some RFS signs, relevant data, for example color, texture, shape and the like from respective segmented parts of the at least one laryngoscopic image have to be compared with references, so as to yield comparison data; and a score for the considered sign is determined. For other RFS signs, a score is determined according to methods described hereinafter. Finally, RFS is determined by summing up separated signs' scores.

1. Capturing an image of the laryngopharyngeal area of a patient

An image, preferably a color video image, of a patient's laryngopharyngeal area is acquired. According to a preferred embodiment, the resolution of the image is not lower than 400x400 resolution. The image is acquired by the system, preferably operated by a medically- trained person, such as an ear-nose-and throat (ENT) physician. The image has patient and other relevant indicia tagging attached thereto, and the image, for example in the form of a digital file, is stored in the memory 40. Then, the image is transmitted to the processor 20 for analysis. According to one embodiment, the image acquisition module 10 is configured to produce a Red-Green-Blue (RGB) image file, or any other compatible format of an image file. According to other embodiments, the image may be acquired in various types of light, for example, infra-red (IR) light, ultraviolet (UV) light, multispectral light, and visible light. Accordingly, the light source 104 and the image acquiring device 106 are configured to acquire images in the aforementioned types of light. According to one embodiment, during the capturing of an image, namely a stills image, or a video image, or any combination thereof - an additional calibration process may be performed, for example calibration of illumination level. According to an additional embodiment, the system 1 is configured to provide an image in any color format known in the art, for example RGB, Hue-Saturation- Value (HSV), Hue- Saturation-Lightness (HSL) and the like.

2. Selecting and retrieving an image from the captured images

The video acquired in the step "capturing an image of the laryngopharyngeal area of a patient" is received either directly from the image acquisition module 10 or from the memory 40, and is used as input for the diagnostic software contained in the processor 20. Each selected image, preferably an RGB color image, from the video input is then coarsely represented using a number of bins. A clustering algorithm, for example K-means, is used to cluster the coarse image data. The purpose of clustering is for recognizing a "triangle" in the image, which is actually the area of the triangular laryngopharyngeal passageway as seen in Fig. 2.

Fig. 2 illustrates, according to an exemplary embodiment, a laryngoscope image of a healthy normal laryngopharynx 200 of the kind which could be used in an exemplary embodiment of a collection of baseline images.

Selection of suitable images for analysis is described hereinafter in algorithm 2.1. The selection of at least one image from the captured images comprises selection of an image with open vocal folds 202. For most of the RFS signs, an algorithm for open vocal fold image selection, described hereinafter, configured to choose images with open vocal folds 202, is used. Some of the signs require a matching pair of images with open/closed vocal folds 202, selection of which is described hereinafter in algorithm 2.2.

Figs. 3A-C illustrate some embodiments of a method for selecting an image. Algorithm 2.1: Open vocal fold image selection

Fig. 3A schematically illustrates, according to an exemplary embodiment, a flow diagram of an algorithm for open vocal fold image selection that selects images suitable for further analysis from a set of laryngoscopic images. These are images with open vocal folds. In other words, the algorithm for open vocal fold image selection is configured to select good quality images having a sufficiently identifiable "triangle". According to one embodiment, the algorithm is configured to select images from a set of laryngoscopic images. According to a preferred embodiment, the algorithm is configured to select images from a laryngoscopic video.

According to one embodiment, images where a "triangle" is recognized are selected by: 1. Extracting images from a video 310, for example by splitting the video into separate frames 320;

2. Calibrating illumination level (in the whole image excluding black/overexposed areas) 330;

3. Recognizing a candidate for "triangle", as follows:

For each frame:

3.1 cutting an image to bins 340;

3.2 clustering bins from the image (by color) 350;

3.3 choosing as a candidate for "triangle" one of K (up to 3) darkest clusters based on cluster configuration and size 360;

3.4 Filling-in "islands" within a chosen cluster 370. The remaining area of the image has to be filled in another color.

4. Checking correspondence of the candidate from step 3 to triangle form up to the threshold T (varying in the interval from 0.4 to 1) 380. If triangle form is not identified, the image is not suitable for analysis.

From the images where a triangle was successfully recognized, mostly suitable images are automatically selected while using the following criteria:

"triangle" form is identified;

image is in focus (as described in detail hereinafter);

"triangle" is not close to the borders of the image (all analyzed areas are present); no essential overexposed parts in the areas of interest. Algorithm 2.2: Selection of opened/closed vocal folds pair

Fig. 3B schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for selecting an image with closed vocal folds 400, following the open vocal fold image selection illustrated in Fig. 3A, suitable for further analysis.

According to one embodiment, the selection of open/closed vocal folds pair comprises:

1. Choosing suitable images:

1.1 splitting video to separate frames;

1.2 calibrating illumination level;

1.3 recognizing a "triangle" up to threshold T (as in the algorithm for open vocal fold image selection described above).

1.4 selecting automatically N most suitable images which satisfy the following two criteria: "triangle" is identified, and image is in focus.

2. For each selected image, checking a series of consequently following images to find an image with closed vocal folds:

2.1 while "triangle" is identified, skipping the image 420;

2.2 saving symmetry axis (SL) of "triangle" in the last image with triangle 430;

2.3 for each of the following M images, checking if vocal folds are closed:

(a) cutting an image to bins 440 and clustering bins from the image (by color) 450;

(b) approximating the darkest cluster by rectangle with symmetry axes SL of width s 460.

(c) if the darkest cluster is approximated up to the threshold T, the candidate for closed vocal folds image is found and saved 470;

(d) choosing from obtained candidates the best quality candidate with respect to percentage of cluster falling into the rectangle 480.

The output of this algorithm is a pair of images: one with opened vocal folds, another with closed vocal folds 490.

Algorithm 2.3: Analysis and enhancement of focus Images obtained as an input for this step are images with a "triangle" selected by the algorithm for open vocal fold image selection, as illustrated in Fig. 4A, that are sequenced in order of appearance in video. Fig. 3C schematically illustrates, according to an exemplary embodiment, a flow diagram of a method for selecting the most focused images with open vocal folds, suitable for further analysis.

Selection of most focused images

1. Classifying images to similarity groups:

For a pair of consequent images with detected "triangle":

1.1 checking if heights of the "triangles" in compared images are within the same range. The range is defined as a predefined threshold;

If the heights don't fall in the same range, the second image starts a new similarity group;

otherwise, checking similarity using histogram comparison or block matching method, as described hereinafter;

1.2 If the images are similar up to a predefined threshold, they both enter the same group;

1.3 If the images are not similar, the second image (the compared one) starts a new group;

1.4 Continuing to compare the last image from the pair to the next image, until the end of the image set.

2. Applying focus function measures:

For each similarity group:

2.1 Measuring the focus of each image, using one of focus measurement functions, for example, based either on image statistics, or on depths of peaks and valleys, or on image differentiation;

2.2 Returning the most focused image of the group.

3. If no focused image, up to predefined threshold, was found, using a local approach:

3.1 Selecting at least one area of interest, as described hereinafter, and then applying the focus measures on each area of interest separately;

3.2 Checking if the obtained images stand up to predefined focus quality demands.

If yes, finishing; Else, sharpening the image, for example using Unsharp masking or Spatial filters.

Checking image similarity by histogram comparison

Either RGB histograms or HS histograms of the images are considered. The histograms are normalized, and then compared by a distance function such as correlation, chi-square, intersection, Bhattacharyya distance, and the like.

Checking image similarity by block matching

1. Choosing a macro block:

1.1 Calculating the height h and the base length of the triangle in the compared image.

1.2 In relation to the "upper left" vertex of the triangle with a pixel factor (a safety factor of several pixels to ensure that the vertex is covered): within a parameter P 110, taking a rectangular macro block of size that is in relation with the triangle measures, for example 1.5 x height x 1.5 x base (see Fig.4A).

2. Block matching:

In relation to the "upper left" vertex of the triangle in the image where the matching block is searched for, with a pixel safety factor, searching up to a predefined search parameter Q for the block that is the closest to the macro block (see Fig. 4B), using distance functions, for example Mean Absolute Difference or Mean Squared Error).

Fig. 4A illustrates, according to an exemplary embodiment, choosing a macro-block in a laryngoscope image for the block matching algorithm.

Fig. 4B illustrates, according to an exemplary embodiment, sliding a macro block found as in Fig. 4A in a compared laryngoscope image, in search for the closest matching block, so as to check similarity of images.

3. Evaluating RFS signs

For each sign, areas specific for the considered sign are analyzed using different features, for example texture, color, form and size evaluation, as described in detail hereinafter.

4. Calculating RFS scoring

Summing up the grades for each sign obtained in the previous step, according to Table 1 and concluding about LPR presence. A score above 7 indicates that LPR is detected. Table 1 : Reflux Finding Score (RFS)

Methods used for signs' scoring Texture recognition - In a selected area of interest, the texture of the area on interest may be analyzed using wavelet decomposition. Multivariate Laplacian distribution may be used for fitting the wavelet coefficients histogram. A set of filters has to be used for recognition of the area specifics.

Classification of images - To calculate a score of some RFS signs a supervised-learning classification algorithm, for example KNN, is used. In the learning stage of the classification algorithm a dataset is built for each analyzed sign, for each possible grade of this sign. Images that comprise the dataset are manually evaluated by narrow-profiled ENT experts. On testing stage of the classification algorithm the tested image is compared against existing datasets as described hereinafter.

General description of RFS scoring method

Relevant area detection

For analysis of different RFS signs the following areas of interest are used:

1. Vocal fold areas L and R (areas to the left/ to the right of the "triangle", see Fig. 5).

Fig. 5 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for marking of areas of interest L and R.

2. Arytenoids area A (areas to the left/ to the right of the upper edges of "triangle", see Figs. 6A-B). Figs. 6A-B illustrate, according to an exemplary embodiment, annotated reference laryngoscope images to be used for marking of areas of interest A, B, L and R.

3. Posterior Commissure area B (area above the "triangle", see Fig. 14). It should be noted that in a case when an area of interest contains overexposed parts, they have to be removed from consideration. For example overexposed bins cannot be used in the sign detection algorithms).

RFS Sign 1: Subglotic Edema Fig. 7 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method 500 for identifying the presence of subglotic edema in images selected from a patient's laryngoscopic video, following the algorithm for open vocal fold image selection shown in Fig. 3A.

Identification of presence of Subglotic Edema

For the chosen image 510:

1. Detecting ventricular folds 520, as described hereinafter in the algorithm for RFS Sign 2. If none is identifiable, defining and using a standard width of vocal fold

(by default half of L area width);

2. Cutting vocal folds 530 (areas between the "triangle" and ventricular folds), not including vocal fold border area;

3. Dividing each vocal fold into left and right parts 540 (as shown in Fig. 8);

4. Building a histogram of each part and comparing them 550, for example by using Kolmogorov-Smirnov distance;

5. If right and left parts are similar up to threshold 7j, there is no Subglatic Edema and the score is 0; otherwise it is considered that there is Subglatic Edema 570 and the score is 2.

Fig. 8 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for illustration for an algorithm for determining the presence of subglotic edema.

RFS Sign 2: Ventricular Obliteration

Fig. 9 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method 600 for identifying the presence and level of ventricular obliteration in images selected from a patient's laryngoscopic video, following the algorithm for open vocal fold image selection shown in Fig. 3A.

Identification of presence and level of Ventricular Obliteration

For the chosen image:

1. Cutting from the chosen image 610 areas L and R 620 (see Fig. 5) excluding a border of the "triangle";

2. Searching for a ventricular fold (shown by the red arrow in Fig. 5) in the obtained area: 2.1 Clustering the area into three clusters by color 630;

2.2 Analyzing an area U which presents a union of darker clusters amounting to less than a half of the considered area 640:

If U can be approximated by a rectangle of width W (defined to be in the interval [0, 0.2] of the area's width) up to some threshold T2 (T2 is a percentage of considered clusters falling into the rectangle; defined to be in the interval [60,100]), the score is 0;

Else, decrease (or reduce) threshold T2 in N% of T2 660;

If U can be approximated by a rectangle up to the decreased threshold, the score is 2 670;

Otherwise the score is 4 680.

RFS Sign 3: Erythema/Hyperemia

Images suitable for analysis of RFS Sign 3 are selected according to the opened/closed vocal folds pair selection algorithm, described in Fig. 3B. The learning database of the system 1 contains bins with corresponding textures for comparison.

Fig. 10 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method 700 for analyzing erythema/hyperemia.

Identification of presence and level of Erythema/Hyperemia

1. Cutting from the opened/closed vocal folds pair of images areas 710 A,B,L,R 720 (see Figs. 6A-B).

2. Using supervised-learning classification algorithm, comparing texture of arytenoids (area A) bins with textures of bins with/without Hyperemia from the learning database. If all bins of area A are classified as "without Hyperemia", and

Erythema is not detected by an algorithm described hereinafter, the score is 0 750.

3. If there exist bins in area A classified as "with Hyperemia", or Erythema is detected, continuing to check Erythema/Hyperemia in areas B, L and R;

if in one of these areas there exist bins classified as "with Hyperemia", or Erythema is detected, the score is 4 760.

Otherwise, the score is 2 770.

An algorithm for detecting Erythema in area X in the image im Transforming an image im from RGB to HSV format;

Calculating standard deviation of hue values in the image im that evaluate its homogeneity;

If homogeneity of im is less than threshold 7¾ (namely, the image is

homogeneous):

If average hue value of im is less than threshold 7V, the score is 1 ; Else, the score is 0;

If homogeneity of im is greater than threshold 7¾ (namely, the image is not homogeneous) check similarity between the whole image im and the area X: remove from im irrelevant areas ("triangle", black/overexposed areas);

if Kolmogorov-Smirnov distance between histograms of hue values of im and X is greater than threshold D (varying in the interval [0, 0.5]), the score is 1; Else, the score is 0.

RFS Sign 4: Vocal Fold Edema

Fig. 11 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method 800 for identifying the presence and level of vocal fold edema in an image selected from a patient's laryngoscopic video, following the algorithm for open vocal fold image selection shown in Fig. 3A. Bins with corresponding textures for comparison are saved in the learning database.

Identification of presence and level of Vocal Fold Edema

For the chosen image:

1. Choosing ventricular folds (as for Sign 2); if none is identifiable, defining and using for the next step the standard width of vocal folds;

2. Cutting vocal folds 810 (areas between the "triangle" and ventricular folds rectangle);

3. Cutting vocal folds into bins 820 in nonempty way, namely if width of vocal fold is about the size of bin, finding the best correspondence of chosen bin to the form of the vocal fold);

4. Using supervised-learning algorithm for comparing texture of vocal fold bins with textures of vocal fold bins from the learning database divided into five subsets with different scores of the sign 830 and grading Vocal Fold Edema 840. RFS Sign 5: Diffuse Laryngeal Edema

Fig. 12 schematically illustrates, according to an exemplary embodiment, a flow diagram of a method 900 for identifying the presence and level of diffuse laryngeal edema in an image selected from a patient's laryngoscopic video, following the open/closed vocal folds pair selection process described in Fig. 3B 910.

Identification of presence and level of Diffuse Laryngeal Edema

1. Cutting from the pair of images (see Figs. 6A-B) areas A, B, and LUR 920.

2. Using supervised-learning classification algorithm, comparing texture of the obtained areas with textures of corresponding areas in images from the learning database divided into five subsets with a different score of the sign 930.

3. If a number of areas with a non-zero score is higher than 1, return a maximal score; otherwise return 0 940. RFS Sign 6: Posterior Commissure Hypertrophy

Fig. 13 schematically illustrates according to an exemplary embodiment, a flow diagram of a method 100 for identifying the presence and level of posterior commissure hypertrophy in an image selected from a patient' s laryngoscopic video, following the algorithm for open vocal fold image selection shown in Fig. 3A.

Identification of presence and level of Posterior Commissure Hypertrophy

For the chosen image:

1. Cutting, in other words choosing, from the image 110 a relevant area B 120 (located inside the rectangle shown in Fig. 14);

2. Using supervised-learning classification algorithm, comparing texture of the obtained area with textures of corresponding areas in images from the learning database 130.

If no essential area with "pachyderma" texture was detected 140, returning 0; Otherwise 150 finding upper boundary of the "triangle" and defining the sign grade up to curvature of the obtained boundary 160. Fig. 14 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for illustration for an algorithm for determining the presence and level of posterior commissure hypertrophy. RFS Sign 7: Granuloma

Fig. 15A schematically illustrates, according to an exemplary embodiment, a flow diagram of a method 1000 for identifying the presence of initial stages of granuloma in an image selected from a patient's laryngoscopic video, based on the algorithm for open vocal fold image selection shown in Fig. 3A.

Following hereinafter are algorithms for identifying the presence of both advanced and initial stages of Granuloma.

It should be noted that this approach is not suitable for extremely large granulomas amounting to more than 50% of "triangle".

It should be also noted that the input for the algorithm analyzing Sign 7 for presence of advanced stages of Granuloma are images selected based on the algorithm for open vocal fold image selection described in Fig. 3A, the "triangle" and vertices of the "triangle".

Identification of presence of advanced stages of Granuloma

1. Applying filter for edge detection on the image 1010;

2. Calculating the height of the "triangle" using vertices 1020;

3. Determining the area of interest as a rectangle that is lifted to 1/3 from the bottom vertex of the "triangle", having height that equals to the height of the

"triangle" and width that equals twice the length of base of the "triangle" 1030;

4. Applying clustering to rectangle chosen in Step 3 in the filtered image 1040, and trying to approximate each cluster by an ellipse, up to a threshold Γ'γ, e.g. in the following way:

4.1. Removing cluster outliers;

4.2. For each cluster:

4.2.1. For each point inside the cluster:

Finding the farthest point (according the Euclidean distance) that belongs to the same cluster. Obtaining main axis of the ellipse as a maximum distance, getting the center of the axis as ellipse center 1050; 4.2.2. Using the center and the main axis of the ellipse obtained in Step 4.2.1 to find the perpendicular axis of the ellipse 1060;

4.2.3 Finding the boundary of the cluster and filling up this boundary 1060.

4.2.4 Comparing the shape obtained in Step 4.2.3 to the ellipse found in Step 4.2.2. up to a threshold Τ_Ί 1070.

If the shape does not concise with the ellipse, dismissing the cluster.

Else using supervised-learning classification algorithm, comparing texture of bins cut from the ellipse 1080 with bins with/without granuloma texture from the learning database;

Checking if the texture of bins cut from the ellipse is uniform 1090;

5. If in one of the ellipses found in Step 4 there exist N bins with granuloma texture and the texture is uniform up to a threshold Τ' Ί, score is 2 1092;

otherwise, checking for presence of initial stages of Granuloma as described hereinafter

1094;

if an initial stage of Granuloma is found, the score is 2;

otherwise, the score is 0.

Fig. 15B schematically illustrates another embodiment of a flow diagram of a method 1100 for identifying the presence of advanced stages of granuloma in an image selected from a patient's laryngoscopic video, based on the algorithm for open vocal fold image selection shown in Fig. 3A.

Fig. 16 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for illustration for an algorithm for identifying the presence of granuloma.

The input for the algorithm analyzing Sign 7 for presence of initial stages of Granuloma are images selected based on the algorithm for open vocal fold image selection described in Fig. 3A, the "triangle" and vertices of the "triangle" 1110.

Identification of presence of initial stages of Granuloma

1. Calculating the height of the "triangle" using vertices 1120;

2. Cutting from the image a band of vocal fold width to the left/to the right of the "triangle", starting from the 1/3 bottom of "triangle" up to the base of the "triangle" 1130;

3. Using supervised-learning classification algorithm, comparing texture of selected area bins with texture of bins with/without granuloma from the learning database

1140;

4. If all bins of selected area are classified as "without granuloma texture", an initial stage of granuloma is not detected 1150;

Otherwise, initial stage of granuloma is detected 1160.

RFS Sign 8: Endolaryngeal Mucus

Images suitable for analysis of Sign 8 are selected based on the algorithm for open vocal fold image selection described in Fig. 3A, focus of the "triangle" area and illumination level.

Fig. 17 illustrates, according to an exemplary embodiment, an annotated reference laryngoscope image to be used for illustration for an algorithm for identifying the presence of thick endolaryngeal mucus.

Fig. 18 schematically illustrates, according to an exemplary embodiment, a flow diagram showing the process 1200 for identifying the presence of thick endolaryngeal mucus in a set of images suitable for analysis.

Identification of presence of Endolaryngeal Mucus

For each ofN chosen images 1210:

1. Cutting a "triangle" area from the image 1220;

2. Clustering pixels of the chosen area into two clusters, by color 1230;

3. Checking the form of the lighter cluster 1240:

3.1 approximating the form of the lighter cluster a by rectangular form;

3.2 checking correspondence of the form of the lighter cluster to the

approximating rectangle up to a certain threshold;

3.3 If it is well approximated 1250, Sign 8 is determined, and the score is 2;

Otherwise, the score is 0 1260.

It is appreciated that certain features of the subject matter, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the subject matter, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable sub combination. Although the subject matter has been described in conjunction with specific embodiments thereof, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art. Accordingly, it is intended to embrace all such alternatives, modifications and variations that fall within the spirit and broad scope of the appended claims.

Claims

1. A system for determining a pathological condition in the laryngopharyngeal area of a patient, the system comprising:

wherein the system is configured to:

capture at least one image of the laryngopharyngeal area of a patient;

select at least one image from the captured images; and

determine the pathological condition of the laryngopharyngeal area by analyzing the selected at least one image.

2. The system of claim 1 , further comprising a memory 40 configured to store data and images.

3. The system of claim 2, configured to store data and images for objective follow-up of patients and follow-up of patients during treatment of the pathological condition.

4. The system of claim 1 , wherein the at least one image is at least one stills image.

5. The system of claim 1, wherein the at least one image is at least one video image.

6. The system of claim 1, wherein the at least one image is a Red-Green-Blue (RGB) color image, or Hue-Saturation- Value (HSV) color image, or Hue-Saturation-Lightness (HSV) color image.

7. The system of claim 1, wherein the selecting at least one image from the captured images comprises selecting an image with open vocal folds.

8. The system of claim 1, wherein the selecting at least one image from the captured images comprises selecting an image with closed vocal folds.

9. The system of claim 1, wherein the selecting at least one image from the captured images further comprises selecting the most focused images.

10. The system of claim 1, wherein the pathological condition is Laryngopharyngeal reflux (LPR).

11. The system of claim 10, wherein the selecting an image for diagnosis from the captured images comprises selecting at least one image from the at least one image that are suitable for applying eight algorithms, one for each Reflux Finding Score (RFS) sign.

12. The system of claim 10, wherein the analyzing the selected at least one image comprises: evaluating RFS for each RFS sign on the selected at least one image; and

calculating a total RFS score by summing the scores of each RFS sign.

13. The system of claim 12, wherein the evaluating RFS for each RFS sign on the selected at least one image comprises:

determining an RFS according to the extracted data.

14. The system of claim 13, wherein the determining an RFS according to the extracted data comprises determining conditions of RFS signs and giving scores accordingly, as follows: if Subglottic Edema is present giving a score of 2 and if Subglottic Edema is absent giving a score of 0;

if Ventricular Obliteration is absent giving a score of 0, if Ventricular Obliteration is partial giving a score of 2 and if Ventricular Obliteration is complete giving a score of 4; if Erythema or Hyperemia is absent giving a score of 0, if there are arytenoids only giving a score of 2 and if Erythema or Hyperemia is diffuse giving a score of 4;

if Thick Endolaryngeal Mucus is present giving a score of 2 and if Thick

Endolaryngeal Mucus is absent giving a score of 0.

15. The system of any one of claims 10 and 14, wherein it is determined that the patient has LPR when a sum of the RFS scores is bigger than 7.