Researchers have discovered a new way to analyze lung cancer samples using digital pathology, a technology that allows pathologists to view and analyze whole slide images of tissue samples on a computer. The study, led by a team of Spanish scientists, aimed to assess the reliability of pathologists in determining the percentage of tumor cells in non-small cell lung cancer (NSCLC) samples using digital image analysis. Their findings suggest that digital pathology can help improve the accuracy of tumor cell percentage estimation, which is crucial for molecular testing and targeted cancer treatments. This work highlights the potential of digital tools to enhance cancer diagnosis and personalized medicine. Non-small-cell lung carcinoma, Digital pathology
Revolutionizing Lung Cancer Diagnosis with Digital Pathology
Lung cancer is one of the most deadly cancers, but advancements in molecular testing and targeted therapies have offered new hope for patients. A crucial aspect of this progress is accurately determining the percentage of tumor cells in a sample, as this information guides the selection of appropriate treatment options.
Traditionally, pathologists have relied on visual examination of tissue samples to estimate tumor cell percentage. However, this method can be subjective and prone to variability, which can impact the reliability of molecular testing and treatment decisions.
Overcoming the Limitations of Visual Assessment
To address this challenge, researchers from Spain conducted a study to evaluate the reliability of pathologists in determining tumor cell percentage using digital pathology. The team trained pathologists from nine different centers to quantify epithelial tumor cells, tumor-associated stromal cells, and non-neoplastic cells in whole slide images (WSIs) of NSCLC samples using an open-source software called QuPath.
The researchers conducted two consecutive ring trials, where the pathologists analyzed a set of WSIs and reported their findings. The first trial revealed poor reliability among the pathologists, with an intraclass correlation coefficient (ICC) of just 0.09. This was largely due to the subjectivity involved in the annotation process and the classification of tumor cells versus tumor-associated stromal cells.
Improving Accuracy with Feedback and Training
After the first trial, the pathologists received feedback on their performance, and the researchers made some adjustments to the digital analysis methodology. In the second trial, the reliability among the pathologists improved, with an ICC of 0.24. However, the overall agreement was still considered poor, suggesting that the subjective nature of some tasks, such as the annotation process, remained a challenge.
Implications for Molecular Testing and Personalized Medicine
The study also examined the relationship between tumor cell percentage and the results of next-generation sequencing (NGS), a molecular testing technique used to detect genetic alterations in cancer samples. The researchers found that cases with low tumor cell percentage (≤ 20%) by visual assessment sometimes presented molecular alterations, but digital analysis often assigned these cases higher tumor cell percentages, suggesting that the digital approach may be more accurate.
This finding is particularly significant, as current guidelines recommend a minimum of 20-30% tumor cells for NGS studies to ensure reliable results. By using digital pathology to more precisely determine tumor cell percentage, clinicians may be able to identify a greater number of patients who are eligible for targeted cancer therapies.
Toward Automated and Objective Tumor Analysis
The researchers acknowledge that the subjectivity inherent in some digital pathology tasks, such as the annotation process, remains a challenge. To address this, they plan to explore the use of artificial intelligence (AI) solutions to automate the annotation of tumor and non-tumor areas, potentially improving the reliability and consistency of the analysis.
Digital pathology, combined with advanced AI algorithms, holds great promise for enhancing cancer diagnosis and guiding personalized treatment decisions. By providing a more objective and reproducible way to assess tumor characteristics, this technology could help unlock the full potential of precision oncology, ultimately leading to better outcomes for lung cancer patients.
Author credit: This article is based on research by Irene Carretero-Barrio, Lara Pijuan, Adrián Illarramendi, Daniel Curto, Fernando López-Ríos, Ángel Estébanez-Gallo, Josep Castellvi, Sofía Granados-Aparici, Desamparados Compañ-Quilis, Rosa Noguera, Isabel Esteban-Rodríguez, Ignacio Sánchez-Güerri, Ana Delia Ramos-Guerra, Juan Enrique Ortuño, Pilar Garrido, María Jesús Ledesma-Carbayo, Amparo Benito, and José Palacios.
For More Related Articles Click Here
