Development of a computerized morphometry application for assessment of the tumor fraction in colon carcinoma tissue samples

Pathologist Computer-Aided Diagnostic Scoring of Tumor Cell Fraction: A Swiss National Study

Tumor cell fraction (TCF) estimation is a common clinical task with well-established large interobserver variability. It thus provides an ideal test bed to evaluate potential impacts of employing a tumor cell fraction computer-aided diagnostic (TCFCAD) tool to support pathologists' evaluation. During a National Slide Seminar event, pathologists (n = 69) were asked to visually estimate TCF in 10 regions of interest (ROIs) from hematoxylin and eosin colorectal cancer images intentionally curated for diverse tissue compositions, cellularity, and stain intensities. Next, they re-evaluated the same ROIs while being provided a TCFCAD-created overlay highlighting predicted tumor vs nontumor cells, together with the corresponding TCF percentage. Participants also reported confidence levels in their assessments using a 5-tier scale, indicating no confidence to high confidence, respectively. The TCF ground truth (GT) was defined by manual cell-counting by experts. When assisted, interobserver variability significantly decreased, showing estimates converging to the GT. This improvement remained even when TCFCAD predictions deviated slightly from the GT. The standard deviation (SD) of the estimated TCF to the GT across ROIs was 9.9% vs 5.8% with TCFCAD (P < .0001). The intraclass correlation coefficient increased from 0.8 to 0.93 (95% CI, 0.65-0.93 vs 0.86-0.98), and pathologists stated feeling more confident when aided (3.67 ± 0.81 vs 4.17 ± 0.82 with the computer-aided diagnostic [CAD] tool). TCFCAD estimation support demonstrated improved scoring accuracy, interpathologist agreement, and scoring confidence. Interestingly, pathologists also expressed more willingness to use such a CAD tool at the end of the survey, highlighting the importance of training/education to increase adoption of CAD systems.

Fourier transformation based texture analysis for differentiating between hyperplastic polyps and sessile serrated adenomas

Colorectal cancer (CRC) is the third most common type of cancer. One major pathway involved in the development of CRC is the serrated pathway. Colorectal polyps can be divided in benign, like small hyperplastic polyps and premalignant polyps, like the sessile serrated adenomas (SSA) that has a significant potential of malignant transformation. The morphological similarity between these types of polyp, not-infrequently raises diagnostic difficulties. This study aimed to morphologically differentiate between hyperplastic polyps (HP) and SSAs by using automated computerized texture analysis of Fourier transformed histological images. Thirty images of HP and 58 images of SSA were analyzed by computerized texture analysis. A fast Fourier transformation was applied to the images. The Fourier frequency plots were further transformed into gray level co-occurrence matrices and four textural variables were extracted: entropy, correlation, contrast, and homogeneity. Our study is the first to combine this type of analysis for automated classification of colonic neoplasia. The results were analyzed using statistical and neural network (NNET) classification models. The predictive values of these classifiers were compared. The statistical regression algorithm presented a sensitivity of 95% to detect the SSA and a specificity of 80% to detect the HP. The NNET analysis was superior to the statistical analysis displaying a classification accuracy of 100%. The results of this study have confirmed the hypothesis that Fourier based texture image analysis is helpful in differentiating between HP and SSA. RESEARCH HIGHLIGHTS: Colorectal polyps can be divided in benign, like hyperplastic polyps (HP) and premalignant, like the sessile serrated adenomas (SSA). There is a high morphologic similarity between these two types of polyp that not-infrequently raises diagnostic difficulties. The results of our morphometric analysis that were used to build a neural network based model of prediction of the polyp types, have a great clinical importance of identifying SSA polyps which have significant potential of malignant progression as compared to HP.

Morphoproteomic Characterization of Lung Squamous Cell Carcinoma Fragmentation, a Histological Marker of Increased Tumor Invasiveness

Accurate stratification of tumors is imperative for adequate cancer management. In addition to staging, morphologic subtyping allows stratification of patients into additional prognostic groups. In this study, we used an image-based computational method on pan-cytokeratin IHC stainings to quantify tumor fragmentation (TF), a measure of tumor invasiveness of lung squamous cell carcinoma (LSCC). In two independent clinical cohorts from tissue microarrays (TMA: n = 208 patients) and whole sections (WS: n = 99 patients), TF was associated with poor prognosis and increased risk of blood vessel infiltration. A third cohort from The Cancer Genome Atlas (TCGA: n = 335 patients) confirmed the poor prognostic value of TF using a similar human-based score on hematoxylin-eosin staining. Integration of RNA-seq data from TCGA and LC-MS/MS proteomics from WS revealed an upregulation of extracellular matrix remodeling and focal adhesion processes in tumors with high TF, supporting their increased invasive potential. This proposed histologic parameter is an independent and unfavorable prognostic marker that could be established as a new grading parameter for LSCC. Cancer Res; 77(10); 2585-93. ©2017 AACR.