Validation of automated positive cell and region detection of immunohistochemically stained laryngeal tumor tissue using digital image analysis

Abstract

Objectives: This study aimed to validate a digital image analysis (DIA) workflow for automatic positive cell detection and positive region delineation for immunohistochemical hypoxia markers with a nuclear (hypoxia-inducible factor 1α [HIF-1α]) and a cytoplasmic (pimonidazole [PIMO]) staining pattern.

Materials and methods: 101 tissue fragments from 44 laryngeal tumor biopsies were immunohistochemically stained for HIF-1α and PIMO. QuPath was used to determine the percentage of positive cells and to delineate positive regions automatically. For HIF-1α, only cells with strong staining were considered positive. Three dedicated head and neck pathologists scored the percentage of positive cells using three categories (0: <1%; 1: 1%-33%; 2: >33%;). The pathologists also delineated the positive regions on 14 corresponding PIMO and HIF-1α-stained fragments. The consensus between observers was used as the reference standard and was compared to the automatic delineation.

Results: Agreement between categorical positivity scores was 76.2% and 65.4% for PIMO and HIF-1α, respectively. In all cases of disagreement in HIF-1α fragments, the DIA underestimated the percentage of positive cells. As for the region detection, the DIA correctly detected most positive regions on PIMO fragments (false positive area=3.1%, false negative area=0.7%). In HIF-1α, the DIA missed some positive regions (false positive area=1.3%, false negative area=9.7%).

Conclusions: Positive cell and region detection on biopsy material is feasible, but further optimization is needed before unsupervised use. Validation at varying DAB staining intensities is hampered by lack of reliability of the gold standard (i.e., visual human interpretation). Nevertheless, the DIA method has the potential to be used as a tool to assist pathologists in the analysis of IHC staining.

Keywords: Biomarker analysis; Computational pathology; Positive cell detection; Positive region detection; Validation.

Fig. 1

Image preprocessing workflow of PIMO (left) and HIF-1α (right) stained laryngeal tumor biopsies. Isolated tissue fragments are automatically detected (A) and exported into separate image files (B). Corresponding PIMO and HIF-1α fragments are registered to each other (C) and an automatic tissue detection is performed (magenta). Artifacts are manually removed from the annotation.

Fig. 2

Positive cell and region detection for PIMO (top row) and HIF-1α (bottom row). Cell detection is performed on the original image (A,E) within the tissue annotation (magenta). The optical density of the DAB-staining is measured in each cell’s cytoplasm for PIMO and nucleus for HIF-1α fragments. The black arrows in the measurement maps (B,F) show the threshold for positivity that is used to separate positive (red) from negative cells (blue) (C,G). Areas with a high density of positive cells form positive regions (yellow) (D,H).

Fig. 3

Percentage of positive cells as calculated by the DIA (logarithmic scale) versus scored categorically by observers for PIMO (left) and HIF-1α fragments (right). Observers used a semi-quantitative scoring method (0: <1%, 1: 1%–33%, 2: >33%) depicted as horizontal stripes in the graph. Green points represent a categorical agreement between the DIA and observers, orange points a disagreement of one category. Back diamonds indicate the median DIA positivity for each category. Zero was artificially added on the y-axis, as six PIMO fragments and one HIF-1α fragment contained no positive cells according to the DIA.

Fig. 4

Example of positive region detection result where misclassifications are due to small differences in delineation. The observer delineations (A), divided into areas delineated by one observer (blue) and at least two observers (cyan), are compared to the DIA delineation (B) (yellow). The results (C) show the correctly identified regions (green) and misclassifications. False positives (red) are the regions delineated by the DIA, but not by any observer. False negatives (black) are the regions delineated by at least two observers, but not by the DIA.

Fig. 5

Patterns of classification results on PIMO (top row) and HIF-1α fragments (bottom row). The observer delineations (A,D), divided into areas delineated by one observer (blue) and at least two observers (cyan), are compared to the DIA delineation (B,E) (yellow). The results (C,F) show the correctly identified regions (green), false positives (red) and false negatives (black).

Fig. 6

Examples of false positives (red) due to red blood cells (A), and false negatives (black) due to low cell density (B). On the left, the original tissue is shown with the delineated areas, on the right the positive (pink) and negative (blue) cell detections are shown.