Real-time peri-operative microcalcification detection in superficial breast tissues

Abstract

Surgical management of ductal carcinoma in situ (DCIS), particularly in cases involving suspicious morphology and orientation of microcalcifications, remains a primary treatment option. However, the lack of real-time technical assistance in the form of an intraoperative surgical tool for detection of microcalcifications in the resection margins presents a significant challenge. In the context of breast conserving surgery, ex-vivo imaging of excised breast tissues slices from 12 patients was conducted. By employing a cross-polarized multispectral microcamera setup for tissue visualization an imaging depth of up to 2 mm was achieved. The microcamera provides the clinician with a clear color image with magnification allowing features down to 50 μm to be seen on the resection surface. Mammography images were used for accurate cross-correlation, enabling the identification of microcalcifications in the microcamera images. Detection efficacy of microcalcifications in microcamera images was notably influenced by both calcification clustering and distribution depth within the tissue. Calcifications within the 2 mm range were detected through their distinct optical manifestations in relation to the adjacent tissues. Four independent reviewers-two medical and two technical-achieved an average sensitivity of 77.8%, specificity of 80.0%, and overall accuracy of 79.0%. This study demonstrates the potential of an integrated microcamera and cross-polarized setup for non-invasive, real-time detection of microcalcifications in superficial breast tissues. By focusing on the superficial 2 mm, this approach shows promising results and offers substantial opportunities for future research and clinical applications.

Fig. 1

The steps of the measurement protocol are illustrated with the relevant image name indicated below the illustration. In Step 1 an X-ray of the slice is captured using a mammography system (S_MAM). In Step 2 the PPM system is used to capture a RGB image of the slice (S_SPEC). In Step 3 the PPM system is used to co-locate S_MAM with S_SPEC and dots are projected onto regions of interest to generate S_PPM. In Step 4 the microcamera probe is placed above these regions of interest at a fixed distance of 8 mm from the tissue sample to achieve a field of view of 1.0 × 1.0 cm with upwards of 50 μm magnification. The multi-LED source provides the light used for illuminating the tissue. During acquisition, real-time images in color, greyscale and infrared are shown on the screen.

Fig. 2

Examples of the method for filtering specular reflections from S_MicroCam images. (a) Original greyscale images with microcalcifications present and not present. (b) Blob-like features are detected in the images using a LOG method. The scale bar indicates the sigma of the gaussian filter used, which translates to the size of the blob-like feature detected. Larger detected blobs are circled in red, smaller ones in blue. (c) Specular reflection filtered images are generated. Blobs under a threshold size are filtered out using a median filtering technique.

Fig. 3

Workflow of the co-registration method for identification of microcalcifications in the S_MicroCam images. (a) The point-based image registration is performed by manually selecting corresponding anatomical landmarks in S_PPM and S_MAM. (b) Binary image of the projected measurement locations were transformed into squares too match the FOV of the optical probe. (c) The measurement locations are overlaid on the S_PPM and S_MAM, generating the exact counterpart of the S_MicroCam image. (d) S_ROI and S_MicroCam are co-located based on exactly the same visible surface landmarks on both images. In this step up to 20 anatomical landmarks are selected on each image. Six such points are shown as examples in the figure for readability. The landmarks were selected by a clinician who marked the locations using a custom script. (e) Based on the co-location in step 4, S_MAM is warped to match S_MicroCam, thereby providing a new orientation of the microcalcifications. (f) Lastly, the microcalcifications from S_MAM are projected on S_MicroCam with a circle representing an error of 1 mm.

Fig. 4

S_MAM with corresponding greyscale S_MicroCam images of near infrared light at 850 nm. The yellow box represents the measurement with microcalcifications, whereas the blue box is considered a negative control.

Fig. 5

Nine examples of S_MicroCam images of various microcalcification distributions, accompanied with the corresponding S_MAM. (a) Large clusters of microcalcifications within both S_MAM and S_MicroCam. (b) Small groups of clustered microcalcifications, and (c) absence of microcalcifications in S_MicroCam images due to absence of microcalcifications in the mammography (examples 7 and 8), or suspected distribution at depth more than 2 mm (example 9).