Differential diagnosis of thyroid nodule capsules using random forest guided selection of image features

Abstract

Microscopic evaluation of tissue sections stained with hematoxylin and eosin is the current gold standard for diagnosing thyroid pathology. Digital pathology is gaining momentum providing the pathologist with additional cues to traditional routes when placing a diagnosis, therefore it is extremely important to develop new image analysis methods that can extract image features with diagnostic potential. In this work, we use histogram and texture analysis to extract features from microscopic images acquired on thin thyroid nodule capsules sections and demonstrate how they enable the differential diagnosis of thyroid nodules. Targeted thyroid nodules are benign (i.e., follicular adenoma) and malignant (i.e., papillary thyroid carcinoma and its sub-type arising within a follicular adenoma). Our results show that the considered image features can enable the quantitative characterization of the collagen capsule surrounding thyroid nodules and provide an accurate classification of the latter's type using random forest.

Figure 1

The feature extraction workflow: (a) whole slide image acquisition; (b) cropping of image tiles with different sizes (i.e., 1024 × 1024, 512 × 512, 256 × 256) within nodule capsules areas; (c) RGB to 8-bit conversion using four strategies (i.e., 8b, 8bw, RED, EOSIN) and subsequent feature extraction using the histogram, GLCM and Helmholtz analysis; (d) binarization of 8-bit images and subsequent feature extraction (i.e., TC-ratio, fractal dimension).

Figure 2

Flowchart showing the experimental steps of the proposed method.

Figure 3

Typical bright-field images acquired on collagen nodule capsules for the cases of (a) FA, (b) PTC and (c) PTCFA.

Figure 4

Statistical analysis for the case of histogram parameters extracted from the 1024 × 1024 pixels image tiles. (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Figure 5

Statistical analysis for the case of GLCM parameters extracted from the 1024 × 1024 pixels image tiles. (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Figure 6

Statistical analysis for the case of binary image parameters extracted from the 1024 × 1024 pixels image tiles. (*p < 0.05, ***p < 0.001, ****p < 0.0001).

Figure 7

Statistical analysis for the case of Helmholtz analysis parameters extracted from the 1024 × 1024 pixels image tiles. (*p < 0.05, **p < 0.01).