Confocal Laser Endomicroscopy For The Diagnosis Of Colorectal Cancer In ViVO

Backgrounds and aimsColorectal cancer (CRC) is one of the highest-incidence malignant tumors in the world.Colonoscopy together with biopsy remains the gold standard for the diagnosis of CRC. However, there are several common disadvantages of biopsy including bleeding, sampling error and specimen crush, etc. Risks of biopsy-induced epithelial misplacement in the muscularis propria and metastasis are also of concern.Furthermore, local fibrosis after biopsy could influence the non-lifting sign of submucosal CRCand rectal carcinoids, which is accepted to be a contraindication of endoscopic mucosal resection. In some endoscopy units, patients who take antiplatelet or antithrombotic drugs are asked to stop the anticoagulant medication one week before colonoscopy in order to reduce the risk of acute bleeding events, which will undoubtedly increase the risk of disease progression. Thus we should find a noninvasive and real-time diagnostic method.Confocal laser endomicroscopy (CLE) is a new endoscopic technique, and itcombines the miniature laser confocal microscope with the traditional electronic endoscope. This technique is capable of both ordinary endoscopy in white light mode and confocal laser scanning on gastrointestinal mucosa. After intravenous injection of fluorescent developer, CLE can scan gastrointestinal mucosa layer by layer, and the real-time images obtainedcan be magnified into1000times of high resolution images, which make it possible to display histological structure of mucous membrane on thecellular level.Several studies have shown that CLE can diagnose bronchial lung cancer, superficial gastric cancer, and flat-type CRC in vivo with high sensitivity, specificity and accuracy. Another study have shown that squamous cell carcinoma of the head and neck can be accurately diagnosed with CLE during surgery.These studies have demonstrated CLE has high value on real-time diagnosis of mucosal carcmogenesis.Previous reports on CRC have mainly focused on the examination of circumscript lesions in the colon. They did not report whether CLE could enable diagnosis of different forms of CRC with both high sensitivity and specificity, or how the lesion sites, macroscopic and microscopic types could affect the feasibility of diagnosis have not been commonly reported yet. And CLE is a new endoscopic technique which can make real-time diagnosis of suspicious lesions in cell level rather than the organ level, which requires endoscopic doctors to have a basic understanding of pathological image of gastrointestinal mucosa and high resolution magnification confocal endomicroscopic images of CRC. The time for cognitive ability training and the training effect is also a problem to be solved urgently.Hence the aims of this study were:(1) Through a prospective study, to evaluate the sensitivity, specificity and accuracy of CLE for diagnosis of CRC, to assess the intra-observer agreement and to find out whether the characteristics and positions of the lesion affect the feasibility of CLE diagnosis;(2) according to the standards we formulate, to evaluate the learning curve, the diagnostic accuracy, and the intra-observer agreement of inexperienced observers for the diagnosing of colorectal cancer.MethodsPart one:The value of confocal laser endomicroscopy for the diagnosis of CRC in vivo.This part was divided into two steps. In the first step, CLE was used to examine histologically confirmed colonic mucosa, noncancerous and cancerous colonic lesions. CLE images were compared with horizontal sectioned histopathological pictures obtained from the same colonic lesions. The diagnosis of CRC made using CLE was based on the literatures and the two endoscopists’ own experience. The literatures on the diagnosis of colonic neoplasia using CLE were reviewed, and each endoscopist had performed hundreds of CLE procedures prior to this study. Other diagnoses made during CLE, such as inflammation or adenoma, were based on our previous studies..In the second step, consecutive patients were prospectively recruited from February1to July31,2012in Qilu Hospital. Those with suspected malignant lesions which were detected during CLE were included in the study. CLE procedures were not different from conventional colonoscopy, except for the specific stain agent (5mL10%fluorescein sodium), injected i.v. prior to the observation of the lesions. Each lesion was investigated at four quadrants and at least four Z-stack images (from superficial to deep layer) were collected at each quadrant. The distance between the surgical resection margin of CRC and the edge of the lesion was at least5cm, indicating that the mucosa less than5cm around the cancer was an easily diffused region; therefore, in addition to the cancerous lesions, inflammation or normal adjacent mucosae less than5cm from the lesion was also observed and tissue specimens were taken using biopsy. All the CLE procedures were performed by two endoscopists. In vivo real-time CLE diagnosis was made for endoscopic suspicious lesions by the performing endoscopist. The endoscopist reassessed the CLE pictures collected by himself after half a year interval. Images collected by one endoscopist were assessed by the other endoscopist blinded to the patients’ clinical history and endoscopic information. All the CLE diagnoses which were compared with the conventional histological diagnosis were made according to the afterwards still images analysis in order to eliminate the influence of the white light images on the judgment.In addition, a post hoc assessment of inter-observer and intra-observer agreements for the CLE findings were performed.The ratio of number of images with cancerous characteristics against the whole number of the images from each lesion was calculated (ratio of cancer images/whole images,RCW) to indicate the feasibility of diagnosis using in vivo virtual histology.The RCW of different lesions were calculated to find out whether the characteristics and positions of the lesion affect the feasibility of CLE diagnosis.Part two:The learning curve of confocal laser endomicroscopy for the diagnosis of CRCAn experienced CLE endoscopist selected120pairs of high qualityimages from thedata base. We divided the selected images into3groups according to the randomnumbers. There are5endoscopists in this study, including2experienced observersand3inexperienced endoscopists who knew nothing about CLE.After forty minutes’training, including our previous studies and experiences for diagnosis of CRC and20images (11benign and9malignant) with known histology results, all the observers were asked to evaluate the three sets of selected images independently. Their results were corrected and discussed after each set. The diagnostic results, confidence and quality of each image were recorded. After a session of study, every observer took a twenty minutes’rest. The accuracy of CLE diagnosis for CRC of each observer and the intra-observer agreement were assessed.ResultsPart oneSeventy-one patients with suspected malignant lesions were included in the study. A total of74lesion and92adjacent mucosa sites were observed. The sensitivity, specificity, and accuracy were97.10%,98.97%, and98.19%(endoscopist A), and98.55%,96.94%, and97.59%(endoscopist B), respectively. The intra-observer agreement was graded as excellent (endoscopist A,kappa=0.953;endoscopist B, kappa=0.934).The inter-observer agreement between the two endoscopists was alsoexcellent (kappa=0.950). The accuracy in diagnosing poorly differentiated CRC using CLE was97.01%for endoscopist A and95.59%for endoscopist B.RCWs were significantly higher in the distal colon than in the proximal colon, with the highest in rectum (83.5%±3.6%) and the lowest in cecum (53.0%±1.0%)(P<0.001). However, RCWs among the sigmoid, descending and transverse colons were not significant (P=0.189). Of the three main macroscopic types, polypoid had the highest RCW (77.5%±1.3%), and ulceration had the lowest RCW (61.6%± 11.5%)(P<0.01). The difference between polypoid and infiltration was not significant (P=0.237). The RCW of well-differentiated CRC was significantly higher than that of poorly differentiated CRC (P<0.001).Part twoAccuracy of the first40images was95.00%(observer A)、92.50%(observer B)、90.00%(observer C) for each inexperienced observer. There was no significant change in accuracy for any of the observer in the consecutive sets of images (observerA, P=0.812; observer B,P=0.859; observer C,P=0.697).The overall accuracy of inexperienced group was92.50%,94.17%,95.83%in the three sets of images respectively, and there were no significant changes among the overall accuracy of inexperienced group (P=0.366)。The accuracy of inexperienced group are slightly lower than that of experienced group in each set of pictures, but the difference was not statistically significant(set one P=0.085;set two P=0.086;set three,P=0.264).The overall accuracy was95.85%,94.17%,92.50%for inexperienced observers in all images. The kappa values of the inter-observer agreement was0.888during the assessment.Conclusions1. CLE has the potential to enable an immediate diagnosis of CRC and the degree of differentiation of CRC during ongoing endoscopy in vivo.The factors affecting the feasibility of diagnosis including the location, gross morphology and degree of differentiation of thelesion.2. Accurate post hoc interpretation of CLE confocal images can be learned quickly. CLE is a promising technology with high diagnostic accuracy in the diagnosis of CRC that can be achieved by inexperienced observers after a short training and learning.