3D texture analysis for classification of second harmonic generation images of human ovarian cancer

Abstract

Remodeling of the collagen architecture in the extracellular matrix (ECM) has been implicated in ovarian cancer. To quantify these alterations we implemented a form of 3D texture analysis to delineate the fibrillar morphology observed in 3D Second Harmonic Generation (SHG) microscopy image data of normal (1) and high risk (2) ovarian stroma, benign ovarian tumors (3), low grade (4) and high grade (5) serous tumors, and endometrioid tumors (6). We developed a tailored set of 3D filters which extract textural features in the 3D image sets to build (or learn) statistical models of each tissue class. By applying k-nearest neighbor classification using these learned models, we achieved 83–91% accuracies for the six classes. The 3D method outperformed the analogous 2D classification on the same tissues, where we suggest this is due the increased information content. This classification based on ECM structural changes will complement conventional classification based on genetic profiles and can serve as an additional biomarker. Moreover, the texture analysis algorithm is quite general, as it does not rely on single morphological metrics such as fiber alignment, length, and width but their combined convolution with a customizable basis set.

3D second harmonic generation imaging tomography by multi-view excitation

“…For example, the dermis and ovarian stroma in normal tissues have 3D mesh-like morphologies [9,10], and SHG will not visualize the entire structure. We previously have shown that 3D texture analysis provides significantly improved discrimination between ovarian tumors over the analogous 2D approach even using the incomplete information in 3D stacks [8]. Similar results were found by Georgakoudi and co-workers using a different image analysis approach [11].…”

“…Using these types of metrics, SHG imaging has been used for revealing extracellular matrix (ECM) structural changes in a wide range of diseases such as cancers, connective tissue disorders, and fibroses [2–7]. For example, we have shown that different ovarian tumors are morphologically distinct from normal tissues, as well as benign tumors and tissues at high risk for developing high grade cancer [8]. …”

MicroRNA‑138 inhibits SOX12 expression and the proliferation, invasion and migration of ovarian cancer cells

“…Ovarian cancer accounts for only 4% of malignant tumors in women, but it has the highest mortality rate among all gynecological malignant tumor types (3,4). Recurrence and early metastasis of tumor cells are the primary reasons for the poor prognosis of patients with ovarian cancer (5).…”

Automated classification of multiphoton microscopy images of ovarian tissue using deep learning

“…Recently, the approach was generalized to 3-D texture analysis and to classify SHG images from six different ovarian tissue types. 14 These studies demonstrate the potential of machine learning-based evaluation of SHG images for improved diagnostic accuracy of ovarian cancer detection.…”