Use of ImageJ software for histomorphometric evaluation of normal and severely affected canine ear canals

Abstract

Morphological studies comparing normal and diseased ear canals use primarily subjective scoring. The aim of this study was to compare normal and severely affected ears in dogs with objective measurements using ImageJ software. Ear canals were harvested from cadavers with normal ears and from dogs that underwent total ear canal ablation for unresolved otitis. Histopathology samples from ear canals were evaluated by semi-quantitative scoring and also by using ImageJ-software for histomorphometric measurements. The normal ears were compared to the severely affected ears using the 2 methods. The 2 methods were significantly (P < 0.0001) correlated for epidermal hyperplasia, ceruminous gland dilation, and hyperplasia and tissue inflammation, which were significantly greater in the severely affected ears (P < 0.0001). This study demonstrated that there is a very high correlation between the 2 methods for the most markedly affected components of otitis externa and that ImageJ software can be efficiently used to measure and evaluate ear canal histomorphometry.

Introduction

Infections of the external ear (otitis externa, OE), which includes the pinna, the external auditory meatus, and the ear canal, are very common in dogs, estimated to occur among 10% to 20% of canine patients seen by veterinarians (1). Primary causes and predisposing factors create a suitable environment for the proliferation of secondary pathogens that exacerbate the disease.

The most common primary causes of OE in dogs are allergies such as atopic dermatitis and food hypersensitivity. Fifty-five percent of dogs diagnosed with atopic dermatitis may show signs of otitis externa and, in 3% to 5% of dogs, OE may be the only clinical sign (2). Up to 80% of dogs diagnosed with food hypersensitivity have otitis externa, which may be the only sign in 25% of them (3). Among the predisposing factors, conformation abnormalities of the external ear canal are very common (1,4,5). It has been suggested that differences in the shape of the external ear canal and pinnae among breeds may contribute to the susceptibility of certain breeds to OE (6). Of the breeds that are genetically affected by conformational abnormalities, German shepherds and cocker spaniels are prone to seborrhea, German shepherds have high moisture levels in their ear canals, cocker spaniels have pendulous pinnae, shar peis have hypoplastic and stenotic ear canals, and poodles have a high density of hair in their ear canals (4–11).

Perpetuating factors may delay or even prevent healing of the disease (4,5). The most important perpetuating factor is the thickening of the epidermis and dermis. This leads to numerous skin folds within the ear canal that cause both a favorable environment for the proliferation of yeasts and bacteria and prevent epithelial migration, which constitutes the normal cleaning functioning of the ear. The proliferation causes increased narrowing of the ear canal, which leads to disease and prevents healing. The proliferating and chronic processes of OE lead to fibrotic changes of the dermis and subcutaneous tissues that, if not treated appropriately, will lead to irreversible changes and to a situation of “end-stage” ear, which necessitates surgical intervention. In some cases, the cartilage may even undergo calcification and ossification (5). Total ear-canal ablation (TECA) is recommended in cases of end-stage ears. This procedure is often combined with lateral bulla osteotomy (TECAS/LBO) because of middle ear involvement. Postoperative complications include soft tissue complications, such as wound complications, pinna necrosis, and chronic fistulation, which may require more surgery. Neurologic complications include exacerbation of preoperative facial nerve deficits, peripheral vestibular syndrome, and hearing loss. In many cases, these complications are temporary (12–16).

There are also cases in which tissue response may prevent healing and it is not clear whether this is a predisposing or perpetuating factor. Severe dilation of ceruminous glands was found in severe OE in cocker spaniels, which is believed to be a major causation of end-stage otitis in this breed (7,17). It is not clear, however, whether this is a cause or effect of OE as an increased glandular structure was found in longhaired breeds, such as spaniels (18), and also in other breeds, such as Labrador retrievers, that are prone to OE (19). Histopathological evaluation of more normal and diseased ear canals, either in specific breeds or in the general dog population, is required to resolve this dilemma.

Up until now, histopathological features of normal and diseased ear canals have been evaluated semi-quantitatively and subjectively (7,17–20) or by a stereological method (21). In subjective scoring, investigators evaluate pathological changes and rank them according to their severity (7). In the stereological method, the ratio between the areas that the examined component occupied and the examined tissue was calculated (21). In subjective scoring, fine differences between cases cannot be detected and the cases are therefore assigned the same grade. It is possible that quantitative measurements and more accurate measurements that can be achieved with special computerized programs will provide better diagnostic tools to evaluate even subtle changes in the process of the disease and may help to differentiate between cause and effect in different cases. Furthermore, this kind of measurement can also help to assess the benefits of treatments for OE.

ImageJ is a readily available, free software (http://imagej.nih.gov/ij/) that enables accurate measurements of computerized images following predetermined calibration. This program has been used for histomorphometric measurements (22–24). Since it was developed in 2004, it has been widely used in many different fields of scientific research, including molecular and single cell evaluations (http://www.ncbi.nlm.nih.gov/pubmed). However, data about its use in veterinary dermatohistopathology or in evaluating ear canals have not yet been presented.

The objective of this study was to describe for the first time the use of ImageJ software for evaluating ear canals in dogs and to compare histomorphometric measurements of various parameters of normal and end-stage ear canals. These parameters were also evaluated subjectively and semi-quantitatively in accordance with methods described in the published literature (7,17–19). To that end, this scoring is referred to as the “gold standard” or “state of the art” method and the 2 methods were compared. We evaluated the histological features of both normal and severely affected canine ear canals in breeds that generally do not have anatomical predisposition to otitis in order to prevent bias caused by known changes in these breeds, e.g., cocker spaniels with chronic severe otitis externa have distinct differences in pathologic characteristics of the horizontal ear canal compared with other breeds (7).

Materials and methods

Study animals

Dogs with no history or clinical signs of otitis externa — normal cadavers

Ninety-six ear canals were harvested from 48 fresh cadavers of mature dogs soon after death and evaluated cytologically and histopathologically at the Department of Pathology, The Kimron Veterinary Institute, Veterinary Services, Ministry of Agriculture, Israel. Most of the dogs were strays and mixed breeds and none resembled breeds with an anatomical predisposition to OE, i.e., German shepherds, cocker spaniels and other spaniels, shar peis, and poodles, in order to prevent bias caused by specific anatomical changes and tissue response in these breeds (1,4,6,7). Dogs with apparent skin lesions or ear disease were excluded. Dogs in which cytology of ear canals revealed the presence of possible infection or inflammation, i.e., any amount of inflammatory cells or rods or high amount of cocci or yeasts, were also excluded (5,25). Other exclusion criteria included a known history of skin and/or ear disease and glucocorticoid therapy before death, as outlined in the necropsy request form. It was thought unlikely that the stray dogs received any previous treatment.

Dogs with end-stage otitis externa

Sixteen dogs that had undergone total ear-canal ablation (TECA) for unresolved OE at the Veterinary Teaching Hospital, The Koret School of Veterinary Medicine, The Hebrew University of Jerusalem, Israel and that met the breed-exclusion criteria were included: 2 Labrador retrievers, 1 collie, 2 Staffordshire bull terriers, 1 miniature pinscher, and 10 mixed-breed dogs. These dogs were examined by a board-certified dermatologist and surgeons and TECA was recommended due to lack of response to prolonged medical treatments and chronic irreversible changes to the ear canal. In the case of bilateral ear-canal ablation, both ear canals were evaluated.

Histological evaluation and histomorphometric measurements

Samples were obtained for histology from the distal horizontal canal and stored in 10% (v/v) neutral buffered formalin until examination. The tissues were embedded in wax, sectioned, and stained with hematoxylin and eosin. One biopsy cut from each ear, approximately 1 cm length of circular or semi-circular specimen, was randomly selected and evaluated.

In the subjective scoring, the entire specimen was examined and the evaluation took into consideration the dermis, epidermis, and appendages in the full specimen.

The histological findings were graded subjectively on a scale of 0 to 3 (0 — not present; 1 — low grade; 2 — moderate grade; 3 — high grade) for the following variables: hyperkeratosis, ceruminous gland hyperplasia, ceruminous gland atrophy, sebaceous gland hyperplasia, sebaceous gland atrophy, density of hair, dermal fibrosis, inflammation, and mineralization. For the variables epidermal hyperplasia and ceruminous gland dilation, an extended grading was used (0 to 4 instead of 0 to 3; 4 — very high grade).

To obtain histomorphometric measurements, each examined specimen had 3 different fields with a 10× objective randomly selected and photographed with a camera (Edmund Optics, Barrington, New Jersey, USA) coupled to a microscope (Eclipse 80i; Nikon Instruments, Japan). In each photomicrograph, the following parameters were measured with ImageJ software: thickness of the epidermis and cornified layer, area of the ceruminous gland lumen and sebaceous glands, and thickness of the ceruminous gland wall. Each of these parameters was measured 3 times for each image and the average of the 9 measurements of each sample was used for the statistical analysis. This was done in order to have more measured data to compare to the subjective evaluation of an entire sample. Before measuring, a known scale bar of the 10× objective was calibrated as described in a previous study (22). The number of hair follicles in each photomicrograph was counted in the entire area of the photomicrograph and the average counting of the 3 photomicrographs of each sample was calculated and compared to the subjective evaluation of hair follicle density that was done on an entire specimen.

Three other histological parameters, fibrosis, inflammation, and mineralization, were evaluated subjectively in a semi-quantitative manner and scored on a scale of 0 to 3 (0 — not present; 1 — low grade; 2 — moderate grade; 3 — high grade). This was done once for an entire computer image in the 3 photomicrographs of each ear sample and the average scoring of the 3 samples was taken for the statistical analysis. The subjective scoring of the last 3 parameters was compared between their evaluation in a whole sample and 3 randomly selected much smaller areas of that sample. The author who examined the histopathological samples was blinded to their source.

Ethical approval from the participating institutions was not required for this study because in one part of the study cadavers were used and, in the other part, the authors examined tissue that had been removed for medical reasons and submitted for histopathology evaluation as part of the surgical procedure.

Statistical analysis

Measurements carried out on both ears of 1 dog were correlated. Therefore, each variable was analyzed using a generalized estimating equation (GEE) regression model with an identity link function and an exchangeable covariance structure for the working matrix, i.e., correlation between measurements of 2 ears in the same dog but not between dogs. The Wald Chi-square statistic was used to compute statistical significance and 95% confidence intervals of the subjective scorings of various histopathological parameters between normal and diseased ear canals. The same procedure was carried out for each histomorphometric parameter measured by the ImageJ program after averaging 3 measurements for each photomicrograph, except for the variables “number of hair follicles” and scoring of inflammation and fibrosis, which were measured once for each of 3 photomicrographs and then averaged and analyzed. Marginal means were calculated from the model and are presented in the relevant tables. A P-value of 0.05 was considered an indicator of statistical significance.

The results of the 2 evaluation methods, i.e., subjective scoring versus ImageJ, were compared for each variable by using the Spearman nonparametric correlation coefficient. All tests were done using SPSS Statistics 19.0 (http://www.helsinki.fi/~komulain/Tilastokirjat/IBM-SPSS-Statistics-User's-Guide.pdf).

In order to examine intra-observer reproducibility of the ImageJ method, a coefficient of variation (CV) was calculated from the measurement results. This was done by calculating the standard deviation for every ear, divided by the average of each of the measurements [9 measurements for parameters: epidermis, cornified layer, ceruminous gland area (of the lumen), ceruminous gland wall, and sebaceous gland area; and 3 measurements for the parameters: number of hair follicles, fibrosis, and inflammation]. The average of all CVs of each parameter was then calculated by using Microsoft Office Excel software (Microsoft, Redmond, Washington, USA).

Results

Among the ear canals harvested from fresh cadavers, 70 ears had no signs of infection or inflammation and were categorized as normal (NOR). Among these, 53 were also suitable for histomorphometric analysis. Twenty-four ear canals categorized with severe otitis (SOE) were harvested from 16 dogs that underwent TECA; 20 of those ear canals were also examined using the ImageJ program. Results of the histopathology evaluation, subjective scoring, and the histomorphometric measurements are shown in Tables I and II, respectively.

Table I.

Histopathological features examined in dogs with normal ears and with severe otitis externa in a subjective scoring (0 — not present; 1 — low grade; 2 — moderate grade; 3 — high grade; 4 — very high grade)

Group	Normal ears (NOR) n = 70	Severe otitis externa (SOE) n = 24	P-value
Histopathology parameter	Mean rank (CI95)a	Mean rank (CI95)a
Epidermal hyperplasia	0.91 (0.60 to 1.22)	2.90 (2.48 to 3.31)	< 0.0001
Hyperkeratosis	0.55 (0.28 to 0.82)	1.85 (1.28 to 2.42)	< 0.0001
Ceruminous gland dilation	1.51 (1.11 to 1.91)	2.68 (2.03 to 3.32)	0.003
Ceruminous gland atrophy	1.10 (0.77 to 1.42)	1.38 (0.53 to 2.22)	0.545
Ceruminous gland hyperplasia	0.46 (0.16 to 0.76)	1.85 (1.30 to 2.40)	< 0.0001
Sebaceous gland hyperplasia	0.44 (0.18 to 0.71)	1.64 (0.86 to 2.42)	0.004
Sebaceous gland atrophy	1.05 (0.72 to 1.37)	1.63 (0.83 to 2.43)	0.185
Density of hair follicles	1.84 (1.64 to 2.04)	1.69 (1.31 to 2.07)	0.507
Fibrosis	0.31 (0.14 to 0.47)	0.74 (0.24 to 1.24)	0.107
Inflammation	0.67 (0.40 to 0. 49)	2.44 (1.99 to 2.88)	< 0.0001
Mineralization	0	0

^aCI₉₅ minimum-maximum 95% confidence interval.

Table II.

Histomorphometric parameters in normal ears and ears with severe otitis externa given as mean (standard deviation)

Histomorphometric parameter	Normal ears (NOR) n = 53 Mean rank (CI95)a	Severe otitis externa (SOE) n = 20 Mean rank (CI95)a	P-value	Mean CVb
Epidermis (μ)	8.27 (6.89 to 9.65)	32.88 (24.53 to 41.22)	< 0.0001	0.39
Cornified layer (μ)	20.80 (15.25 to 26.35)	22.93 (13.89 to 31.97)	0.694	0.73
Ceruminous gland area (of the lumen) (μ2)	245.96 (114.59 to 378.33)	1856.76 (869.53 to 2843.99)	0.002	0.55
Ceruminous gland wall (μ)	3.93 (3.34 to 4.53)	6.32 (4.93 to 7.71)	0.002	0.47
Sebaceous gland area (μ2)	1421.14 (1029.49 to 1812.79)	1014.58 (776.97 to 1252.19)	0.082	0.32
Number of hair follicles	1.87 (1.48 to 2.26)	2.82 (2.02 to 3.63)	0.036	0.63
Fibrosis (scoring 0 to 3)c	0.52 (0.31 to 0.74)	1.0 (0.54 to 1.46)	0.068	0.57
Inflammation (scoring 0 to 3)c	0.42 (0.27 to 0.57)	2.04 (1.56 to 2.53)	< 0.0001	0.52
Mineralizationc	0	0

^aCI₉₅ minimum-maximum 95% confidence interval.

^bCoefficient of variation — mean of 9 measurements in each ear of all dogs for parameters: epidermis (μ), cornified layer (μ), ceruminous gland area (of the lumen) (μ²), ceruminous gland wall (μ), sebaceous gland area (μ²), and of 3 measurements in each ear for the parameters: number of hair follicles, fibrosis, and inflammation.

^cFibrosis, mineralization, and inflammation scoring: 0 — not present; 1 — low grade; 2 — moderate grade; 3 — high grade.

Subjective scoring (Table I)

Very high statistically significant differences were found between NOR and SOE ears, for which the SOE group received higher scores in the following parameters: epidermal hyperplasia, hyperkeratosis, ceruminous gland hyperplasia, and dermal inflammation (P < 0.0001). Ceruminous gland dilation and sebaceous gland hyperplasia also scored higher in the SOE than the NOR group (P = 0.003 and 0.004, respectively). All the other examined parameters were similar between groups and no statistically significant differences were detected.

Histomorphometric measurements and scoring of images (Table II)

Epidermal hyperplasia was observed in the SOE group (Figure 1), while the thickness of the cornified layer did not differ significantly between the groups. In the SOE, the ceruminous glands were significantly larger than in NOR ears (P = 0.002) (Figure 1) and were also more hyperplastic, as defined by thicker gland walls (P = 0.002). There was only a marginal difference between the groups in the size of the sebaceous glands (P = 0.082). Dogs with SOE had significantly more hair in their ear canal than the other dogs (P = 0.036). The average score for fibrosis was higher in the SOE group than in the other group, but it reached only marginal statistical significance (P = 0.068). The score for ear-canal inflammation increased with increasing severity of the disease (P < 0.0001).

Figure 1.

Photomicrograph of ear canal of a dog that underwent total ablation of the ear canal, showing severe dilation of ceruminous gland and epidermal hyperplasia. Hematoxylin and eosin, bar = 100 μm.

Results of calculating the CV show that, for all the examined parameters, the intra-assay deviation was between 32% and 73%. Measurements of the epidermis width and the area of sebaceous glands were more consistent than measurements of the other parameters and the cornified layer was the least consistent measurement.

Comparison of subjective scoring and histomorphometric measurements (Table III)

Table III.

Correlation between subjective scoring and histomorphometric measurements of histopathology parameters of ear canals

Histopathology parameter	Histomorphometric measurement	Correlation coefficient	P-value
Epidermal hyperplasia	Epidermis (μ)	0.705	< 0.0001
Hyperkeratosis	Cornified layer (μ)	0.228	0.060
Ceruminous gland dilation	Ceruminous gland area (of the lumen) (μ2)	0.435	< 0.0001
Ceruminous gland atrophy	Ceruminous gland area (of the lumen) (μ2)	−0.261	0.076
Ceruminous gland hyperplasia	Ceruminous gland wall (μ)	0.404	0.012
Sebaceous gland hyperplasia	Sebaceous gland area (μ2)	−0.068	0.699
Sebaceous gland atrophy	Sebaceous gland area (μ2)	−0.330	0.018
Density of hair follicles	Number of hair follicles	0.252	0.035
Fibrosis	Fibrosis	0.103	0.395
Inflammation	Inflammation	0.671	< 0.0001

There was a statistically significant high correlation between the 2 methods for epidermal hyperplasia, ceruminous gland dilation/area, and dermal inflammation (P < 0.0001). Counting hair follicles was correlated to the estimation of the hair follicle density (P = 0.035), although there was a statistically significant difference between severely inflamed ear canals and normal ears only when counting the hair, as previously discussed.

The 2 methods were also positively correlated for ceruminous gland hyperplasia (P = 0.012) and negatively correlated for sebaceous gland atrophy (P = 0.018).

Discussion

In the present study, histopathologic features of normal and severely affected ear canals were compared using 2 methods: i) the current standard using semi-quantitative subjective scores, and ii) measurement by computerized program. The 2 methods were significantly correlated (P < 0.0001) for the most pronounced differences between NOR and SOE ears, which were epidermal hyperplasia, ceruminous gland dilation, and dermal inflammation.

Epidermal hyperplasia has been previously found to correlate with OE (7,17,20). It has been proposed as one of the more common disease-perpetuating factors preventing resolution (26) and is considered a relevant histological criterion for OE severity (1,5,19).

Ceruminous gland dilation and hyperplasia were prominent features of SOE, but the sebaceous glands were significantly more hyperplastic when evaluated subjectively and only marginal significance was seen in the histomorphometric measurements. Ceruminous gland dilation has been reported previously in diseased ear canals (17,19,26) and correlated with disease duration (19). It has been speculated that increased glandular structures, which have been found in longhaired dogs, may be the cause of the higher incidence of OE in these dogs (18). Furthermore, ceruminous gland ectasia has been found to be more severe in end-stage ear canals of cocker spaniels than in other breeds, which it has been suggested contributes to that breed’s susceptibility to progress to end-stage OE (7). The cocker spaniel breed or its crosses were not included in our study, which suggests that ceruminous gland dilation is a general feature of severe ear disease. Interestingly, the objective measurements were correlated with subjective scoring of sebaceous gland atrophy, although neither method showed a statistically significant difference between SOE and NOR, even though the sebaceous glands were smaller and more atrophic in the OE. In recent studies, sebaceous gland atrophy was no longer considered a feature of OE (19,21). Because this parameter had marginal statistical significance in the objective measurements, it would be interesting to support or refute this observation by carrying out objective measurements on larger groups of normal and severely affected ear canals.

Another morphologic issue that has shown contradictory findings as a predisposing factor for OE is the amount of hair in the ear canal (4,6,9,21). Statistically significant greater numbers of hair follicles were counted in images taken from SOE than in images of NOR canals (3 different images from each specimen obtained from every canal). When full specimens (1 from every canal) were evaluated subjectively for hair density, however, the scoring was similar between the groups. Although the accurate measurements that can be achieved with the ImageJ program were not carried out for this parameter, this difference between the subjective evaluation and counting can support the argument that it is better to use quantitative and objective tools than subjective, semi-quantitative evaluations. To this end, the results obtained by counting the hair support the assumption that high amounts of hair play a role in the pathogenesis of OE, as excessive growth of hair in the ear canal may prevent the normal excretion of cerumen and debris from the ear (1,4,6,7).

Fibrosis was not a major feature of end-stage ears in this study, although it scored higher in SOE. However, there was only marginal statistical significance when evaluating the images and no significance in the subjective scoring. Fibrosis has been reported as one of the major complications of chronic OE that leads to end-stage ears (7). Moreover, mineralization was not detected in any of the SOE slides examined. The ablated ear canals in this study were evaluated clinically as “end-stage ears” because the otitis did not respond to medical treatment and chronic changes were observed clinically. Severe changes were observed on histopathology of the ablated canals, which were epidermal hyperplasia, ceruminous gland dilation and hyperplasia, and severe inflammation. These led to the end-stage condition and, in addition to the failure of medical treatment, justified removing the ear canals, although fibrosis or mineralization did not occur. Another explanation for the unexpected results regarding fibrosis might be the subjective scoring of this variable from 3 snapshots of each biopsy specimen and the fact that more accurate measurements, such as those obtained with the ImageJ program, could not be used. Indeed, the standard deviation values were very high for the fibrosis and inflammation results, which were the 2 subjectively scored variables.

Histopathological evaluation is limited by the “snapshot” nature of the procedure. Serial cuts are preferable, but are very labor-intensive and may not be feasible for clinical use and single cuts are also used in research (27). Using a more accurate measurement method can improve histopathological evaluation. This study shows that, for the more distinct histopathological changes, there was a high correlation with the standard method and it was not affected by the fact that only 3 areas of the whole slide had been examined. In a recent study, a mathematical method was suggested for measuring the glandular tissue along the entire ear canal, which the authors claimed was more accurate than morphological examination (21). Naturally, the more precise measurements that can be achieved by using computerized programs are even better for evaluating pathological changes.

Successful use of the ImageJ program for evaluating pathological changes in the ear canal can be extended to other skin or tissue pathologies in order to improve diagnoses and to assess different treatment modalities, better understand pathogenesis processes, and more. In this study, the intra-assay variation (represented by CV) for the measured parameters was from 32% to 55% and higher for counted and evaluated parameters. The measured CV might seem a little high because different chosen parameters are not evenly distributed in the tissue and depicting 3 randomly elected fields of a specimen and randomly choosing 3 tissue components in each photo results in a big diversity. When looking at the differences between OE and healthy ears, however, it is apparent that the results for ears severely affected by OE were 4 to 9 times higher than in normal ears for the most significant measurements, i.e., epidermal thickness and ceruminous gland area. For these measurements, this relatively high CV should not affect validity. The difference is less significant for other significant measurements, such as ceruminous gland wall and number of hair follicles, for which more repetitions are needed for each measurement. Inter-observer reproducibility is another important validation criterion that was not practicable for this study. For now, we did the first step for its validation by primarily comparing it to the standard method.

In conclusion, this study demonstrates that there is a very high correlation between semi-quantitative subjective scoring and histomorphometric measurement for the most markedly affected components of severe otitis externa with the ImageJ software. This software can be efficiently used to measure and evaluate ear canal histomorphometry and can be considered a useful tool in research to reduce reliance on subjective measurements of disease severity.