The lack of sufficient margin assessment during breast conserving surgery results in up to 40% of the cases in incomplete tumor removal. We evaluate the feasibility of hyperspectral imaging as an intra-operative margin assessment technique.
Hyperspectral imaging rapidly collects diffuse reflected light with a large field of view over a broad wavelength range (900-1700 nm). Thereby a 3D hypercube is created that contains both spectral and spatial information of the imaged scene. Measurements are performed on 20 freshly excised breast specimen with a pushbroom camera (900-1700 nm). The specimen is sliced according to standard protocol and one slice, that contains both tumour and healthy tissue, is selected for optical measurements. Histopathology of the measured surface of this slice is obtained afterwards and used for hyperspectral data labelling.
We use a spectral-spatial classifier to discriminate tumorous tissue from surrounding healthy tissue. First, we apply a linear Support Vector Machine (SVM) to obtain a pixel-based spectral classification. As output, we obtain classified pixels and their probability estimates. Second, we use this output as input for the spatial regulation, which is based on Markov Random Fields.
This results in a spectral-spatial classification accuracy of 91%. Spatial regulation mainly affects pixels with a tissue type classification dissimilar to its neighbourhood. Thereby, the spectral classification accuracy is not significantly increased but the ‘pepper-and-salt’ effect, observed after pixel-based classification, is reduced.
Esther Kho, Lisanne L. de Boer, Koen K. Van de Vijver, Henricus J. C. M. Sterenborg, and Theo J.M. Ruers, "Hyperspectral imaging for detection of breast cancer in resection margins using spectral-spatial classification (Conference Presentation)," Proc. SPIE 10472, Diagnosis and Treatment of Diseases in the Breast and Reproductive System IV, 104720F (Presented at SPIE BiOS: January 27, 2018; Published: 14 March 2018); https://doi.org/10.1117/12.2288367.5751476539001.
Conference Presentations are recordings of oral presentations given at SPIE conferences and published as part of the conference proceedings. They include the speaker's narration along with a video recording of the presentation slides and animations. Many conference presentations also include full-text papers. Search and browse our growing collection of more than 14,000 conference presentations, including many plenary and keynote presentations.
Study of self-shadowing effect as a simple means to realize nanostructured thin films and layers with special attentions to birefringent obliquely deposited thin films and photo-luminescent porous silicon