Validity of tissue homogeneity in confocal laser endomicroscopy on the diagnosis of laryngeal and hypopharyngeal squamous cell carcinoma

Matti Sievert, Marc Aubreville, Antoniu‐Oreste Gostian, Konstantinos Mantsopoulos, Michael Koch, Sarina K. Mueller, Markus Eckstein, Robin Rupp, Florian Stelzle, Nicolai Oetter, Andreas Maier, Heinrich Iro, Miguel GoncalvesEuropean 2022 · DOI

TL;DR

This study investigates the use of confocal laser endomicroscopy (CLE) to objectively assess tissue homogeneity for diagnosing laryngeal and hypopharyngeal squamous cell carcinoma (SCC). They found that malignant tissue exhibits significantly greater inhomogeneity compared to healthy tissue, suggesting this as a valuable diagnostic marker. While promising, the authors emphasize that homogeneity should be considered alongside other criteria, acknowledging the role of examiner experience.

What Problem Does This Paper Address?

Diagnosing squamous cell carcinoma (SCC) in the larynx and hypopharynx traditionally relies on histopathological examination of tissue biopsies, which is invasive. While non-invasive optical imaging techniques like confocal laser endomicroscopy (CLE) offer cellular-level visualization, their diagnostic accuracy can be limited by examiner dependency and the need for robust interpretation criteria. Specifically, while tissue inhomogeneity is recognized as a potential indicator of malignancy in CLE, there's a lack of objective, quantitative analysis of this feature. This paper aims to bridge that gap by objectively measuring tissue homogeneity in CLE images and comparing it to histopathological findings, seeking to establish its validity as a diagnostic tool.

Methods

The researchers employed a prospective pilot study design, including five patients undergoing planned total laryngectomy for SCC. They utilized a probe-based CLE system (GastroFlex probe and Cellvizio system) to capture images of both tumor and healthy mucosal tissue. To quantify tissue homogeneity, they defined four regions of interest (ROIs) within each CLE frame and analyzed the signal intensity variations. A "kinetic graph" of the mean histogram value was computed for each ROI, and the standard deviation (SD) of these values across the four ROIs was used as a measure of inhomogeneity. This objective measurement was then compared to the gold standard of hematoxylin and eosin (H&E) staining for histopathological classification. Additionally, seven independent raters were asked to subjectively assess tissue homogeneity, blinded to the histology, to evaluate examiner-dependent analysis.

Datasets

The study utilized CLE image sequences captured from five SCC patients. A total of 12,065 images were acquired, from which 60 sequences (3600 images) were selected for analysis based on acceptable fluorescein flooding and image quality. These sequences were then matched with corresponding histopathological diagnoses. No new benchmark datasets were introduced; rather, existing patient data was used for analysis.

Key Insights and Ideas

  • Objective Inhomogeneity as a Biomarker: The study successfully demonstrated that malignant SCC tissue exhibits a statistically significant and quantifiable increase in inhomogeneity compared to healthy epithelium, as measured by the standard deviation of signal intensity across ROIs.
  • Bridging Subjective and Objective: By comparing objective measurements of inhomogeneity with subjective evaluations by experienced and inexperienced raters, the study provides evidence that objective analysis can align with, and potentially standardize, visual assessments.
  • Motion Artifacts are a Challenge: The research highlights that motion artifacts can significantly impact the assessment of tissue homogeneity, potentially leading to false positives for inhomogeneity. This underscores the need for robust artifact detection or mitigation strategies.
  • CLE as a Complementary Tool: While CLE shows promise for identifying malignant areas, the authors emphasize that it is not a standalone substitute for histopathology. Its strength lies in providing real-time, in-vivo information that can guide surgical decisions and potentially reduce the need for unnecessary biopsies.

Key Results

  • Quantitative Difference: Benign mucosa had a mean signal intensity SD of 39.1 ± 1.03 across ROIs, while SCC had a significantly higher mean SD of 101.5 ± 2.6 (P < 0.001).
  • Classification Accuracy: Using tissue homogeneity as the sole criterion, the study achieved an overall sensitivity of 81.8% and a specificity of 86.2% in differentiating between benign and malignant tissue.
  • Rater Agreement: Inhomogeneous tissue architecture was observed in an average of 14.8% of healthy mucosa CLE sequences and 81.8% of SCC sequences, aligning with the objective measurements. The interrater reliability was classified as moderate (κ = 0.53).

Limitations and Future Work

The authors acknowledge several limitations:

  • Motion Artifacts: Tissue homogeneity assessment is susceptible to motion artifacts, which can mimic inhomogeneity. While manual frame selection can help, the authors suggest that AI-based solutions are needed for robust automatic artifact elimination.
  • Probe Artifacts: Other artifacts like blood, saliva, air bubbles, or debris on the probe can also affect image quality and simulate inhomogeneity.
  • Limited Depth of Examination: CLE's fixed depth of examination (around 60 µm) prevents the evaluation of the basal membrane, a crucial determinant in histopathological differentiation between carcinoma in situ and invasive carcinoma.
  • Examiner Dependency: Despite objective measurements, the examiner-dependent evaluation still shows variability, emphasizing that homogeneity is a sub-criterion that needs to be considered within a broader context.

Future work should focus on developing AI-driven approaches to automatically detect and mitigate motion artifacts and other image contaminants. Further research is also needed to explore the diagnostic value of combining tissue homogeneity with other CLE-derived features, such as capillary aberrations, and to investigate its utility in differentiating carcinoma in situ from invasive carcinoma. The potential for combining CLE with other optical techniques like narrow-band imaging (NBI) for enhanced diagnostic accuracy also warrants further study.

Summary updated

Abstract

SCC shows a significant tissue inhomogeneity in comparison to the healthy epithelium. The results support this feature's importance in identifying malignant mucosa areas during CLE examination. However, the examiner-dependent evaluation emphasizes that homogeneity is a sub-criterion that must be considered in a broad context.
BibTeX
@inproceedings{sievert2022validity,
  author = {Matti Sievert and Marc Aubreville and Antoniu‐Oreste Gostian and Konstantinos Mantsopoulos and Michael Koch and Sarina K. Mueller and Markus Eckstein and Robin Rupp and Florian Stelzle and Nicolai Oetter and Andreas Maier and Heinrich Iro and Miguel Goncalves},
  title = {{Validity of tissue homogeneity in confocal laser endomicroscopy on the diagnosis of laryngeal and hypopharyngeal squamous cell carcinoma}},
  booktitle = {European},
  year = {2022},
  doi = {10.1007/s00405-022-07304-y},
}
Figure