A Methodology for Improving Tear Film Lipid Layer Classification

Remeseiro, Beatriz; Bolón-Canedo, Verónica; Peteiro-Barral, Diego; Alonso‐Betanzos, Amparo; Guijarro-Berdiñas, Bertha; Mosquera, A.; Penedo, Manuel G.; Sánchez-Maroño, Noelia

doi:10.1109/jbhi.2013.2294732

Cited by 40 publications

(42 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This ROI was selected from in front of the lower part of the iris because of a better contrast. 22,23 Images from the Tearscope were observed and graded into the categories described by Guillon. 24 Interference images from Doane's interferometer were also graded into the categories described by Thai and Remeseiro.…”

Section: Methodsmentioning

confidence: 99%

Subjective and objective interpretation of tear film interferometry images

Sharma

Oliver

2018

Int J Res Med Sci

View full text Add to dashboard Cite

Background: Assessment of the tear film is necessary in routine clinical practice because an unstable tear film can hamper the quality of life by causing vision-related problems and compromising the ocular surface. One of the major concerns related to an unstable tear film is dry eye. Many of dry eye patients suffer from a lack of meibum which forms the lipid layer of the tear film. The lipid layer can be graded and interpreted by using interferometry. However, interpretation and grading of this dynamic layer may be inconsistent in terms of inter- and intra- observations. This study investigated the difficulty of consistent, subjective grading of clinical findings, in general.Methods: The interferometry images of 30 subjects captured from different equipment were analyzed subjectively. The agreement between intra-observer repeatability was also measured.Results: A positive Spearman’s correlation of 0.81 was found among different grading patterns observed using the Tearsope to compare right and left eyes. Similarly, a positive Spearman’s correlation of 0.63 was found among different grading patterns observed under interferometer in right and left eye. Correlations were statistically significant, p<0.001. The agreement between intra-observer repeatability calculated using Cohen’s kappa values were also statistically significant, p<0.001.Conclusions: A correlation between the findings of different equipment could not be made due to the differences in wavelengths of incident light and the image details. However, a new grading pattern has been proposed to describe the thickness of various lipid layer patterns observed under Doane’s interferometer.

show abstract

Section: Methodsmentioning

confidence: 99%

Subjective and objective interpretation of tear film interferometry images

Sharma

Oliver

2018

Int J Res Med Sci

View full text Add to dashboard Cite

show abstract

“…Interferometry was used to assess the TF-LL patterns, each one reflecting a specific thickness of the TF-LL. The patterns were classified according to a grading scale recommended by the instrument manufacturing company, with categories adopted for humans [8,9,13], and adapted for veterinary use, according to the available literature [10][11][12]. A five-interval scale was used as follows ( Fig.3):…”

Section: Tf Examinationmentioning

confidence: 99%

“…The tear film-lipid layer (TF-LL) can be assessed by interferometry with the observation of interference patterns [8], which provide information on LL thickness and fluidity. Thick LLs show clear meshwork patterns with waves and interference fringes, while thinner layers are more homogeneous [9]. In the veterinary field, the precorneal TF has been examined by polarized light biomicroscopy [10][11][12], and the surface lipid morphology in dogs has been listed under 16 different interference colors and 3 principal pattern variants (wave-like, islet and granitiform) [10].…”

mentioning

confidence: 99%

Meibomian Gland Dysfunction (MGD), as Diagnosed by Non-Contact Infrared Meibography, in Dogs with Ocular Surface Disorders (OSD): A Retrospective Study.

Vinas¹,

Maggio²,

D’Anna³

et al. 2019

Preprint

View full text Add to dashboard Cite

Background Meibomian gland dysfunction (MGD) is one of the possible conditions underlying ocular surface disorders (OSD). Prevalence of MGD in dogs affected by OSD has not yet been reported. We aimed to evaluate the prevalence of MGD among OSD canine patients, which had been assessed by non-contact infrared meibography and interferometry, and to identify MGD associated factors that might guide its diagnosis. Medical records of canine patients examined for OSD between 2016 and 2019 were reviewed. The frequency of MGD was evaluated within different categories (skull conformation, gender, eye and STT-1). The putative MGD risk factors and frequency of MGD within grades of interferometry were evaluated in a regression analysis model and reported as odd ratios (ORs).Results One hundred fifty eyes from 81 dogs with OSD were included with median age 75 months (range 3-192) and female representation with 52%. MGD was present in 70% of the examined eyes. MGD risk was higher in males [OR adj = 3.015 (95% CI: 1.395 - 6.514) ( P = 0.005)] and older patients [OR adj = 1.207 (95% CI: 1.081 - 1.348) ( P = 0.001)]. No significant differences were found between left and right eyes ( P = 0.66) or between the two types of skull conformation ( P = 0.477) and MGD presence. MGD was associated to the lowest LL thickness, as assessed by interferometry (grade 0) [OR= 16.00 (95% CI: 2.104 - 121.68) ( P < 0.001)]. STT values were not significantly associated with the presence of MGD ( P > 0.05).Conclusions MGD is a common underlying pathology in OSD. Being male and higher age are risk factors for MGD. An interferometry grade 0 may guide OSD diagnosis towards MGD.

show abstract

“…Jain & Zongker [70] subsequently also tried to determine whether the classification error rate for synthetic aperture radar images could be reduced by applying feature selection to a set of 18 features derived from four different texture models for each pixel. More recently, several filters were applied to the features extracted with five different texture analysis techniques [71], although, in this case, the authors were not so much interested in finding out which texture features to use, but rather in reducing the computational time necessary to extract the features. When the number of features extracted and processed is reduced, the time required is also reduced in consonance, and this can usually be achieved with minimum performance degradation.…”

Section: Image Classificationmentioning

confidence: 99%