Hyperspectral imaging for tissue classification after ovarian cancer surgery

Introduction/Background* The most important prognostic facttor for the survival of advanced-stage ovarian cancer is the completeness of cytoreductive surgery (CRS). Therefore, an intraoperative technique to detect microscopic tumours would be of great value. The aim of this pilot study is to assess feasibility of near infrared (NIR) hyperspectral imaging (HSI) for the detection of malignant ovarian cancer, using ex vivo tissue
samples collected during CRS.

Methodology In this pilot-study, patients with proven or suspected ovarian cancer planned for CRS were enrolled in the
study. Hyperspectral images with 25 spectral bands were
acquired from the resected tissues in the wavelength range of
665-975 nm. All hyperspectral data were processed by image
calibration and min-max normalisation, glare removal, feature
selection and linear support vector machine (SVM) classifier
training. The performance of the classification was evaluated
by leave-one-out cross-validation.

Result(s)* Ten patients who underwent cytoreductive surgery
for advanced-stage epithelial ovarian cancer (EOC) were
included in the study, from which 26 tissue samples were
imaged, with a total of 26.446 data points that were matched
to histopathology. Samples included tissue of the ovaries, fallopian tubes, uterus, omentum and/or part of the intestines.
Overall, HSI combined with the SVM classifier was capable to
discriminate tumour tissue from non-tumour tissue with a sensitivity of 0.81, specificity of 0.75, area under the curve of
0.83, and Matthew’s correlation coefficient of 0.41.

Conclusion* This pilot study shows that hyperspectral imaging
is a promising technique to discriminate ovarian carcinomas
from the surrounding tissue. Hyperspectral imaging can scan a
whole area, is fast, non-contact, non-invasive and can be used
inside the operation room.