Purpose: Complete tumor removal during cancer surgery remains challenging due to the lack of accurate techniques for intraoperative margin assessment. This study evaluates the use of hyperspectral imaging for margin assessment by reporting its use in fresh human breast specimens. Experimental design: Hyperspectral data was first acquired on tissue slices from 18 patients, after gross sectioning of the resected breast specimen. This dataset, which contained over 22.000 spectra, was well correlated with histopathology and used to develop a support vector machine classification algorithm and test the classification performance. In addition, we evaluated hyperspectral imaging in clinical practice by imaging the resection surface of six lumpectomy specimens. With the developed classification algorithm, we determined if hyperspectral imaging could detect malignancies in the resection surface. Results: The diagnostic performance of hyperspectral imaging on the tissue slices was high; invasive carcinoma, ductal carcinoma in situ, connective and adipose tissue were correctly classified as tumor or healthy tissue with accuracies of 93%, 84%, 70% and 99%, respectively. These accuracies increased with the size of the area consisting of one tissue type. The entire resection surface was imaged within 10 minutes and data analysis was performed fast, without the need of an experienced operator. On the resection surface, hyperspectral imaging detected 19 out of 20 malignancies that, according to the available histopathology information, were located within 2 mm of the resection surface. Conclusion: These findings show the potential of using hyperspectral imaging for margin assessment during breast-conserving surgery to improve surgical outcome.

Complete tumor removal during breast-conserving surgery remains challenging due to the lack of optimal intraoperative margin assessment techniques. Here, we use hyperspectral imaging for tumor detection in fresh breast tissue. We evaluated different wavelength ranges and two classification algorithms; a pixel-wise classification algorithm and a convolutional neural network that combines spectral and spatial information. The highest classification performance was obtained using the full wavelength range (450-1650 nm). Adding spatial information mainly improved the differentiation of tissue classes within the malignant and healthy classes. High sensitivity and specificity were accomplished, which offers potential for hyperspectral imaging as a margin assessment technique to improve surgical outcome.

The Journal of Biomedical Optics (JBO) publishes peer-reviewed papers on the use of novel optical systems and techniques for improved health care and biomedical research.

Optical technologies can potentially be used for tissue characterization in, for example, the detection of positive resection margins during cancer surgery. These technologies are based on the principle that light that has undergone interaction with the tissue contains information regarding the composition, morphology, and microvasculature of the tissue which can be used to discriminate different tissue types. Validation of these optical measurements is done by comparing the experimental results with the gold standard.¹ Currently, the gold standard for tissue characterization is an assessment of microscopy sections, which are stained with hematoxylin and eosin (H&E), by a pathologist. However, the tissue in the H&E sections, used for histopathological assessment, is profoundly deformed compared to the appearance of the tissue during the optical measurements.^{2 – 4} These deformations are introduced when the tissue is processed into H&E sections as explained and shown in Fig. 1. Part of these deformations is related to tissue shrinkage due to formalin fixation and tissue compression and stretching when the tissue is embedded into paraffin blocks. Consequently, overlaying digitalized H&E sections over images that reflect optical measurements of the fresh, unfixed tissue is prone to registration errors.^{5 – 8}

Developing algorithms for analyzing hyperspectral images as an intraoperative tool for margin assessment during breast-conserving surgery requires a dataset with reliable histopathologic labels. The feasibility of using tissue slices hyperspectral dataset with a high correlation with histopathology for developing an algorithm for analyzing the images from the surface of lumpectomy specimens was investigated. We presented a method to acquire hyperspectral images from the lumpectomy surface with a high correlation with histopathology. The tissue slices dataset was compared with the dataset obtained on lumpectomy specimen and the wavelengths with a penetration depth up to the minimum sample thickness of the tissue slices were used to develop a tissue classification algorithm. Spectral differences were observed between tissue slices and lumpectomy datasets due to differences in the sample thickness between both datasets; wavelengths with a high penetration depth were able to penetrate through the thinner tissue slices, affecting the captured signal. By using only wavelengths with a penetration depth up to the minimum sample thickness of the tissue slices, the adipose tissue could be discriminated from other tissue types, but differentiating malignant from connective tissue was more challenging.