Discussion
Residual cholesteatoma is the first cause of failure in cholesteatoma
surgery, leading sometimes the surgeon to adopt expensive and invasive
strategies such as the “second look” tympanoplasty. Among strategies
to reduce the risk of residual cholesteatoma, EES has shown an
improvement in the visualization of hidden areas of the middle ear and
has consequently been accepted as a useful tool in cholesteatoma
surgery. DIE endoscopic visualization may be a further advance in the
intraoperative detection of residual cholesteatoma.
DIE systems such as SPIES have been developed to better identify and
distinguish pathological from healthy tissue and thereby remove the
disease while respecting normal mucosa. In contrast to the commonly used
white light, DIE makes avascular structures (e.g. cholesteatoma) appear
whiter than vascularized tissues: This allows the surgeon to define its
extension precisely and to remove the disease completely.
In other surgical specialties such as ureteroscopy, endoscopic
thyroidectomy and laparoscopy, SPIES was already used for early tumour
detection by means of highlighting hyper-vascular
structures13,14. Improved visualisation of differently
vascularized tissue and thus clearer distinction between pathologies and
normal anatomy are generally cited as the main
advantages8,9.
To the best of authors’ knowledge, the retrospective case series by
Lucidi et al. (2020) was the first paper to investigate the usefulness
of the SPIES image enhancing system in ear surgery8.
The challenge in cholesteatoma surgery is to completely remove the
squamous epithelial, while sparing normal middle ear and mastoid mucosa.
According to the previously mentioned study, spectra A and B filters in
EES are deemed suitable to recognize cholesteatoma
remnants8.
In our study, the comparison of the three DIE techniques (clara, spectra
A, spectra B) makes clear that each of those modalities brings along its
own advantages. When considering the surgeons’ point of view on the
suitability of the three image-improving modalities, a general attitude
in favour of the use of clara followed by spectra B and lastly spectra A
was found. This is expected, as Clara is the “state-of-the-art”
technique used to perform EES. The spectra A and B techniques have
recently been introduced to depict residual cholesteatoma after complete
removal of the disease.
We investigated and quantified separately the added value of clara,
spectra A and spectra B for the first time. In all three imaging
techniques, diseased images are detected more successfully than
disease-free images. This difference becomes very apparent with Spectra
B. With the highest sensitivity but the lowest specificity, a lot of
tissue is misinterpreted as cholesteatoma but only few cholesteatomas
are missed. In our opinion, this increase in sensitivity justifies its
use at the end of every cholesteatoma surgery. Using the spectra B
technique most cholesteatoma-remnants are detected and results regarding
residual cholesteatoma may be improved. Moreover, a high intra-rater
reliability was observed and was highest for spectra B. In our opinion,
this further indicates an additional value regarding the performance of
spectra B. Clara in contrast has the highest accuracy overall and its
general image properties with high illumination makes it most suitable
to perform the surgery itself. By having the smallest number of false
positives, Clara is considered the most conservative restrained method.
Spectra A lies between Clara and Spectra B in terms of
sensitivity-specificity-difference, as well as in terms of detection,
missing, identification and misidentification of cholesteatoma.
Therefore, we don’t see any indication in the use of the Spectra A
technique in cholesteatoma surgery.
Interestingly, some special situations require special considerations
while using digital image enhancement in EES (figure 4). For example,
the tendon of the tensor tympani muscle appears bright at its insertion
in the malleus and may be mistaken for squamous epithelium. Similarly,
upon thorough review of the image set represented in Figure 4 D-F and
related histopathology, it appears that bone fragments produced during
drilling may also be mistaken for cholesteatoma. In summary, Clara may
be considered the most appropriate image modality for use during
surgery. However, as the removal of small amounts of inflamed mucosa has
few consequences for the patient, whereas cholesteatoma remnants can
result in serious complications and require a second operation, we
recommend the use of Spectra B as a final overview at the end of
cholesteatoma surgery, in order to avoid missing any pathological
finding. Indeed, the combined use of Clara and Spectra B, as proposed,
permits to increase the overall accuracy of the endoscopic visualization
up to 75.4% in our study.
The present study includes responses from ear surgeons with different
surgical background (microscopic and endoscopic), different experience
and from different countries. This is a strength of the study, as it
explores the accuracy of the DIE technique in ear surgeons with variable
experience in dealing with endoscopic images of the middle ear,
eventually providing data which can be considered closer to real-life.
In addition, this can also explain the difference of the observed
sensitivity and specificity compared with previous
studies8 and permits to hypothesize a learning
curve-related outcome.
Since not all surgeons who were invited to fill out the questionnaires
responded, we eventually obtained an uneven distribution of responses
across the questionnaires. As a result, not all images were presented
with each technique the same number of times overall. This is certainly
a limitation of the current study, as some responses therefore
influenced the overall result heavier than others. Nonetheless, due to
the high number of responses received, the limited availability of ear
surgeons overall and also limited time resources specifically, we
decided to stop data collection with a slightly unbalanced distribution.
Additionally, the images in the questionnaire were always presented in
the same order, which could have influenced the intra-rater reliability
measure. However, as the focus of the current study was on the results
of the first time point, this limitation can be neglected.
Finally, another limitation can be the static nature of pictures, which
is different from the dynamic visualization, which occurs during EES
along with the manipulation of the suspicious tissue. Indeed, accuracy
in the detection of cholesteatoma could be affected by the dynamic
movements of the endoscope, changes in the point of view, light
reflexion, direct suction and information about previous surgical steps.