Comparative analysis of confocal microscopy on fresh breast core needle biopsies and conventional histology

Image-guided biopsy is a well-established clinical procedure for the diagnosis of breast abnormalities [1]. Suspicious lesions within the tissue are correlated with the histological result of the biopsy specimen. Interpretation is critically dependent on the experience of the radiologist and pathologist, quality of the specimen, and artefacts of the fixation and dyeing process [2]. Clinical infrastructure impacts time to diagnosis which may lead to an emotional burden for the patient and delay the initiation of the therapy. If the biopsy specimen does not contain the suspect lesion, it increases the risk of misinterpretation. A routine on-site, easy to use histological evaluation of core needle biopsies has not yet been established. Correlation of clinical and histological image would allow an immediate feedback, and could fasten the turnaround time for tissue biopsies.

In the intra-operative setting, the need for a fast and reliable on-site histological evaluation of breast malignancies is widely acknowledged. Because clear margins are mandatory in breast-conserving surgery [3], surgeons face the dilemma of whether to generate a minimal defect within the breast while performing a precise tumor excision, or to make a wider excision and reduce the cosmetic results [4]. However, in some types of breast cancer, intraoperative assessment of margins is difficult to perform [5, 6]. Intraoperative frozen section analysis (FSA) is the standard microscopic method for margin assessment. This approach clearly has limitations [7, 8] and only investigates the shortest distance from the tumor perpendicular to the margin instead of the surface of the specimen.

In search for alternative methods for rapid intraoperative evaluation of tumor margins, some approaches for high resolution imaging have been recently investigated. Full field Optical Coherence Tomography (FF-OCT), which provides sub-surface images of tissue with a high resolution has been a topic of selected studies [9, 10]. However, practicability in clinical settings is limited, because interpretation of the pictures requires specific training of the pathologist. Tao et al. used nonlinear microscopy (NLM) to recently demonstrate a method with high diagnostic accuracy and sensitivity [11]. So far fluorescence-based confocal microscopy has shown promising results in evaluating tissue morphology of a specimen [12‐14]. The limited field of view is however not suited to cover large surgical specimen surfaces while the process of “mosaicking” multiple small images is time consuming and may impact accuracy [12].

Recently a confocal laser microscopy scanner became available for evaluation. In the present study, we tested the Histolog™ Scanner (HS) to evaluate the device for application in clinical practice. The technique uses a confocal microscope and allows observation of the tissue in a spectrum between gross morphology and a sub-cellular level that approaches the resolution of a regular 5x objective. Histological images of the superficial layers of fresh thick tissue can be generated for a scanning area of 2.5 cm² by a unique raster. HS was recently tested in a clinical study on skin cancer specimens with encouraging results [15].

Similarly to the intraoperative situation, quality and reliability of diagnosis with image guided biopsies relies on the quality of the technique, and experience of the performing clinician and the pathologist. The aim of our investigation was to assess image quality and correct diagnosis of human breast cancer tissue in core needle biopsies by using the HS, and to further evaluate the practicability and reliability of the device. For this purpose, two independent experienced pathologists read and interpreted the images obtained with the HS scanner and the images obtained after digitalisation of the routinely processed core biopsies.

The primary endpoint was to evaluate the correspondence of breast cancer diagnosis between assessment of confocal HS images and gold standard histological images.

In the aim of finding a cost-effective and beneficial intervention for fast evaluation of cancer cells, several on-site techniques have been evaluated [9, 18]. However, weak sensitivity and limited clinical practicability were the main deficits so far [10, 19, 20]. Previous publications documented a high image resolution by confocal microscope scanning, providing a visualization of cellular details equal to H&E standard [13‐15]. During the staining process applying this technique, fluorochromes mainly dye nuclei and intensify the contrast between epithelium and stroma. As far as the evaluation, done by two independent pathologists, of the two image modalities can be interpreted, there was no detectable morphologic effect of the Acridin Orange incubation on the quality of the histopathological sections (Figs. 4 and 5). In contrast to conventional techniques, confocal microscopy uses point illumination and a pinhole to exclude interfering signals out of the specimen focus. Malignant cells with morphological characteristics of the nucleus can be easily identified and distinguished from normal glandular, adipose and stromal tissue. Digital coloring facilitates the histopathological analysis, and consequentially the learning curve for interpretation is usually steep.

This report shows a promising approach to achieve fast and on-site diagnosis of high quality images obtained from the surface of fresh CNB. The method does not interfere with the standard of care workflow and therefore also has great potential to support surgical margin assessment in breast conserving therapy. Using this innovative approach, histology grade images from the entire surface of the resection margins can be obtained without loss of tissue due to freezing and cutting artefacts. Furthermore, the subsequent procedure of fixation, histologic and immunologic examination of the specimen is not affected.

HS images eliminate the need for mosaicking of multiple images, and therefore misinterpretation caused by gaps or overlapping can be excluded. In this study, pathologists without previous training interpreted confocal images without knowledge of the corresponding H&E standard result. However, earlier reports also showed accurate and reliable evaluation of confocal microscopy images by surgeons [12]. Conformity of malignant or non-malignant diagnosis was 95% for scanner images compared to H&E standard and therefore proved to be highly reliable.

Using traditional techniques of H&E staining, misinterpretations of tissue samples may occur mainly due to tumor type, biopsy material and technical issues, as well as due to pathologist’s experience [2]. Two of our mismatches with the HS occurred in a case of invasive lobular carcinoma, which is known to be challenging in the H&E standard method, and often requires additional techniques such as immunohistochemistry [21, 22]. Misinterpretation of the lobular cancer was seen in the analysis of pathologist A, while pathologist B had a narrow sense correspondence for these images. The other two mismatches were found in cases where only one image of the specimen was available to the pathologists.

Employing the HS scanning technique, this study clearly demonstrates that quality monitoring of the biopsy tissue on-site is reliable, which can reduce the rate of false negative results or re-biopsies. As usually both sides of the tissue sample are imaged, we also see an advantage in that approach compared to the routine histological examination.

With a high level of correspondence between two independent pathologists, the present study demonstrates that carcinoma of the breast can be diagnosed with a high accuracy in biopsy specimen using HS images.

The Histolog™ Scanner detects breast cancer in fresh human tissue with accuracy and high reliability. It is simple to use, cost- and time-efficient and has a great potential to be adopted for routine use in intra-operative margin assessment.