Analysis of interobserver variability for endomicroscopy of the gastrointestinal tract

Endoscopic confocal laser-microscopy for the intraoperative nerve recognition: is it feasible?

Surgeons lose most of their tactile tissue information during minimal invasive surgery and need an additional tool of intraoperative tissue recognition. Confocal laser microscopy (CLM) is a well-established method of tissue investigation. The objective of this study was to analyze the feasibility and diagnostic accuracy of CLM nervous tissue recognition. Images taken with an endoscopic CLM system of sympathetic ganglions, nerve fibers and pleural tissue were characterized in terms of specific signal-patterns ex-vivo. No fluorescent dye was used. Diagnostic accuracy of tissue classification was evaluated by newly trained observers (sensitivity, specificity, PPV, NPV and interobserver variability). Although CLM images showed low CLM image contrast, assessment of nerve tissue was feasible without any fluorescent dye. Sensitivity and specificity ranged between 0.73 and 0.9 and 0.55-1.0, respectively. PPVs were 0.71-1.0 and the NPV range was between 0.58 and 0.86. The overall interobserver variability was 0.36. The eCLM enables to evaluate nervous tissue and to distinguish between nerve fibers, ganglions and pleural tissue based on backscattered light. However, the low image contrast and the heterogeneity in correct tissue diagnosis and a fair interobserver variability indicate the limit of CLM imaging without any fluorescent dye.

Confocal Laser Microscopy for in vivo Intraoperative Application: Diagnostic Accuracy of Investigator and Machine Learning Strategies

BACKGROUND

Confocal laser microscopy (CLM) is one of the optical techniques that are promising methods of intraoperative in vivo real-time tissue examination based on tissue fluorescence. However, surgeons might struggle interpreting CLM images intraoperatively due to different tissue characteristics of different tissue pathologies in clinical reality. Deep learning techniques enable fast and consistent image analysis and might support intraoperative image interpretation. The objective of this study was to analyze the diagnostic accuracy of newly trained observers in the evaluation of normal colon and peritoneal tissue and colon cancer and metastasis, respectively, and to compare it with that of convolutional neural networks (CNNs).

METHODS

Two hundred representative CLM images of the normal and malignant colon and peritoneal tissue were evaluated by newly trained observers (surgeons and pathologists) and CNNs (VGG-16 and Densenet121), respectively, based on tissue dignity. The primary endpoint was the correct detection of the normal and cancer/metastasis tissue measured by sensitivity and specificity of both groups. Additionally, positive predictive values (PPVs) and negative predictive values (NPVs) were calculated for the newly trained observer group. The interobserver variability of dignity evaluation was calculated using kappa statistic. The F1-score and area under the curve (AUC) were used to evaluate the performance of image recognition of the CNNs' training scenarios.

RESULTS

Sensitivity and specificity ranged between 0.55 and 1.0 (pathologists: 0.66-0.97; surgeons: 0.55-1.0) and between 0.65 and 0.96 (pathologists: 0.68-0.93; surgeons: 0.65-0.96), respectively. PPVs were 0.75 and 0.90 in the pathologists' group and 0.73-0.96 in the surgeons' group, respectively. NPVs were 0.73 and 0.96 for pathologists' and between 0.66 and 1.00 for surgeons' tissue analysis. The overall interobserver variability was 0.54. Depending on the training scenario, cancer/metastasis tissue was classified with an AUC of 0.77-0.88 by VGG-16 and 0.85-0.89 by Densenet121. Transfer learning improved performance over training from scratch.

CONCLUSIONS

Newly trained investigators are able to learn CLM images features and interpretation rapidly, regardless of their clinical experience. Heterogeneity in tissue diagnosis and a moderate interobserver variability reflect the clinical reality more realistic. CNNs provide comparable diagnostic results as clinical observers and could improve surgeons' intraoperative tissue assessment.

Pathologists should probably forget about kappa. Percent agreement, diagnostic specificity and related metrics provide more clinically applicable measures of interobserver variability

Kappa statistics have been widely used in the pathology literature to compare interobserver diagnostic variability (IOV) among different pathologists but there has been limited discussion about the clinical significance of kappa scores. Five representative and recent pathology papers were queried using clinically relevant specific questions to learn how IOV was evaluated and how the clinical applicability of results was interpreted. The papers supported our anecdotal impression that pathologists usually assess IOV using Cohen's or Fleiss' kappa statistics and interpret the results using some variation of the scale proposed by Landis and Koch. The papers did not cite or propose specific guidelines to comment on the clinical applicability of results. The solutions proposed to decrease IOV included the development of better diagnostic criteria and additional educational efforts, but the possibility that the entities themselves represented a continuum of morphologic findings rather than distinct diagnostic categories was not considered in any of the studies. A dataset from a previous study of IOV reported by Thunnissen et al. was recalculated to estimate percent agreement among 19 international lung pathologists for the diagnosis of 74 challenging lung neuroendocrine neoplasms. Kappa scores and diagnostic sensitivity, specificity, positive and negative predictive values were calculated using the majority consensus diagnosis for each case as the gold reference diagnosis for that case. Diagnostic specificity estimates among multiple pathologists were > 90%, although kappa scores were considerably more variable. We explain why kappa scores are of limited clinical applicability in pathology and propose the use of positive and negative percent agreement and diagnostic specificity against a gold reference diagnosis to evaluate IOV among two and multiple raters, respectively.

Prediction of clinical outcomes in Crohn's disease by using confocal laser endomicroscopy: results from a prospective multicenter study

BACKGROUND AND AIMS

Assessment of prognostic factors in patients with Crohn's disease (CD) is of pivotal importance for early intervention and "treat-to-target" strategies. Confocal laser endomicroscopy (CLE) enables on-demand in vivo characterization of mucosal inflammatory and architectural changes during endoscopy. We prospectively assessed the value of CLE for prediction of clinical outcome parameters in CD.

METHODS

Consecutive patients with CD undergoing colonoscopy were included in a multicenter study. Confocal imaging focused on 2 highly reproducible histologic hallmarks of active colonic inflammation: focal cryptitis and crypt architectural abnormality. We evaluated whether CLE, CD endoscopic index of severity (CDEIS), serum C-reactive protein (CRP), and CD activity index (CDAI) were associated with the risk of medical treatment escalation, transmural adverse events, and CD-related hospitalization or surgery during a 4-year follow-up.

RESULTS

Among 49 patients (53% men, median age, 39 years), baseline CRP was ≥5 mg/L in 47%, CDEIS ≥3 in 75%, and CDAI >150 in 51%. Focal cryptitis and crypt architectural abnormality were observed in 63% (CLE⁺ group). CLE⁺ patients showed an increased incidence of medical treatment escalation (P < .001; relative risk [RR] = 3.27) and transmural lesions (P = .025; RR = 1.70), whereas patients with CRP ≥5 mg/L had increased CD-related hospitalization and surgery (P = .020, RR = 2.71) at 1-year follow-up. No further association with prognostic clinical outcomes was found over the 1-year follow-up as well as for CDEIS and CDAI at any time.

CONCLUSIONS

CLE reveals CD-related features of mucosal inflammation and allows for early prediction of relevant clinical outcomes. Further studies should now address whether this promising prognostic tool could refine the timing of treatment strategies in patients with CD.

Diagnostic yield of endomicroscopy for dysplasia in primary sclerosing cholangitis associated inflammatory bowel disease: a feasibility study

BACKGROUND AND STUDY AIMS

Primary sclerosing cholangitis associated inflammatory bowel disease (PSC-IBD) is characterized by a high risk of colorectal dysplasia. Surveillance colonoscopies with random biopsies have doubtful power for dysplasia detection. Our aim was to prospectively investigate the feasibility and efficacy of pCLE in surveillance colonoscopies in patients with PSC-IBD.

PATIENTS AND METHODS

Sixty-nine patients with PSC-IBD underwent colonoscopy in 2 steps. On the way from rectum to cecum, the mucosa was inspected with high definition endoscopy (HDE) and random biopsies were taken according to the standard routine. On the way from cecum to rectum, fluorescein-enhanced pCLE and chromoendoscopy were performed. Regions where random biopsies had been taken, as well as visible lesions, were examined with pCLE and targeted biopsies were taken of lesions suspicious for dysplasia. Two investigators, blinded to histology and endoscopy results, analyzed all pCLE videos off-line.

RESULTS

Nineteen biopsies obtained in 13 patients (17 targeted biopsies, 2 random biopsies) revealed the presence of low-grade dysplasia. Thirteen lesions with dysplasia were endoscopically visible but by using pCLE-targeted biopsies, additional endoscopically invisible dysplasias in 4 biopsies obtained from 3 patients were detected. The sensitivity, specificity, and accuracy of pCLE in predicting dysplasia were respectively 89 % (95 % CI: 65 - 98), 96 % (95 % CI: 94 - 97), and 96 % (95 % CI: 94 - 97). pCLE showed a good performance for differentiating neoplastic from non-neoplastic mucosa with negative predictive value of 99 %.

CONCLUSIONS

pCLE in PSC-IBD surveillance is feasible and may be a good complement to HDE. Future research should aim at elucidating whether real-time pCLE is applicable in PSC-IBD surveillance.

Laparoscopic Confocal Laser Microscopy Without Fluorescent Injection

Introduction. Histological analysis of surgical specimen is the gold standard for cancer classification. In particular, frozen histological diagnosis of vague peritoneal spots or uncertain excision of tumors plays a crucial role for proceeding with or without change of the operation procedure. Confocal laser microscopy (CLM) enables in vivo and real-time high-resolution tissue analysis. To evaluate a novel technique of CLM without any fluorescent dye, this pilot ex vivo study demonstrates a CLM camera device for minimal invasive surgical approach. Methods. In 5 cases, a laparoscopic CLM camera was used for examining colon and rectum specimen. Images of nonmalignant and malignant intestinal mucosa were characterized in terms of specific signal-patterns. No fluorescent dye was used. Correlations to findings in conventional histology were systematically recorded and described. Results. Using this CLM camera device, it is possible to analyze colon specimen mucosa. Nonmalignant and malignant intestinal mucosa show specific signal patterns. Nonmalignant mucosa is defined by honeycomb structure. There is deregulated structure in colon and rectum carcinoma mucosa. The inside lumen is irregular. The radial border appears swollen with reduced contrast. Discussion. This pilot study shows that the assessment of colon mucosa with a prototype of CLM camera for minimally invasive surgical approach without any fluorescent dye is feasible. It is possible to differentiate between benign and malignant mucosa in colon specimen by easy to evaluate and reproducible parameters. These first steps of this pioneering achievement to establish CLM in minimal invasive surgical procedures show a great potential for a more reliable intraoperative evaluation of suspect foci.

The learning curve, interobserver, and intraobserver agreement of endoscopic confocal laser endomicroscopy in the assessment of mucosal barrier defects

BACKGROUND AND AIMS

Confocal laser endomicroscopy can dynamically assess intestinal mucosal barrier defects and increased intestinal permeability (IP). These are functional features that do not have corresponding appearance on histopathology. As such, previous pathology training may not be beneficial in learning these dynamic features. This study aims to evaluate the diagnostic accuracy, learning curve, inter- and intraobserver agreement for identifying features of increased IP in experienced and inexperienced analysts and pathologists.

METHODS

A total of 180 endoscopic confocal laser endomicroscopy (Pentax EC-3870FK; Pentax, Tokyo, Japan) images of the terminal ileum, subdivided into 6 sets of 30 were evaluated by 6 experienced analysts, 13 inexperienced analysts, and 2 pathologists, after a 30-minute teaching session. Cell-junction enhancement, fluorescein leak, and cell dropout were used to represent increased IP and were either present or absent in each image. For each image, the diagnostic accuracy, confidence, and quality were assessed.

RESULTS

Diagnostic accuracy was significantly higher for experienced analysts compared with inexperienced analysts from the first set (96.7% vs 83.1%, P < .001) to the third set (95% vs 89.7, P = .127). No differences in accuracy were noted between inexperienced analysts and pathologists. Confidence (odds ratio, 8.71; 95% confidence interval, 5.58-13.57) and good image quality (odds ratio, 1.58; 95% confidence interval, 1.22-2.03) were associated with improved interpretation. Interobserver agreement κ values were high and improved with experience (experienced analysts, 0.83; inexperienced analysts, 0.73; and pathologists, 0.62). Intraobserver agreement was >0.86 for experienced observers.

CONCLUSION

Features representative of increased IP can be rapidly learned with high inter- and intraobserver agreement. Confidence and image quality were significant predictors of accurate interpretation. Previous pathology training did not have an effect on learning.

Characterization of lesions in the stomach: will confocal laser endomicroscopy replace the pathologist?

Confocal laser endomicroscopy (CLE) permits microscopic visualization of the mucosa during endoscopy at an approximately 1000fold magnification, permitting endoscopists to obtain microscopic analysis during gastroscopy. This can result in optimized diagnosis of diffuse alterations such as gastric atrophy and intestinal metaplasia and may limit the sampling error of untargeted biopsies. It also allows risk stratification prior to endoscopic therapy of neoplastic lesions of the stomach. In these areas, CLE represents a valuable adjunct for targeted histopathology. In addition, CLE allows on-site in vivo imaging, and by this insight into physiologic and pathophysiologic as well as molecular events of the stomach without major artifacts.