Validation of the quality of histological images obtained of fresh and formalin-fixed specimens of esophageal and gastric mucosa by laser-scanning confocal microscopy.

BACKGROUND AND STUDY AIMS: We have previously reported the success of a method of virtual histology using laser-scanning confocal microscopy (LCM) in vitro on untreated fresh specimens obtained from the gastrointestinal mucosa. In the present study, we aimed to apply LCM to both fresh and formalin-fixed specimens, without additional treatment, in order to validate and compare the quality of the images obtained. METHODS: We obtained 18 specimens from 11 patients, either by endoscopic biopsy or following surgical resection. First, we observed the fresh, saline-immersed specimen with LCM using the Fluroview microscope (Olympus Co. Ltd., Tokyo, Japan). We then fixed the specimen with formalin and obtained further LCM images 1 hour, 3 hours, and 24 hours after fixation. Three independent observers observed the images and were asked to assess the origin of the samples, the treatment of the samples, the time after formalin fixation, and whether they showed benign or malignant lesions. We used kappa statistics to compare the agreement among the three observers in each of these four areas of interest. RESULTS: Between January and March 2003, we obtained 191 LCM images from 18 specimens. Thirty images were randomly selected for observation. The overall accuracy for differentiating between esophagus and stomach specimens was 96.6 %. The accuracy of differentiating normal from cancerous lesions was 92.2 %. The differentiation between saline-immersed and formalin-fixed specimens was 59.7 % accurate and the assessment of the time interval after formalin fixation was only 37.3 % accurate. The kappa statistics showed that there was strong interobserver agreement on the differentiation of specimen origin and of cancerous from benign lesions. However, there was no agreement among the observers on the method of specimen preparation or on the estimated time interval after formalin fixation. CONCLUSIONS: We concluded that images obtained from fresh specimens using LCM were of a quality good enough to make an accurate diagnosis of upper gastrointestinal carcinoma.