2023 World Conference on Lung Cancer (Posters)-P1.20. Improved Diagnosis of Lung Neuroendocrine Tumors: Role of Proliferative Activity and Deep Learning

P1.20. Improved Diagnosis of Lung Neuroendocrine Tumors: Role of Proliferative Activity and Deep Learning - PDF(Abstract)

Back to course

Pdf Summary

This presentation at the WCLC 2023 conference focuses on the improved diagnosis of lung neuroendocrine tumors (LNETs) using artificial intelligence (AI) and deep learning. The researchers evaluated the use of proliferative activity and deep learning algorithms to enhance the accuracy of LNET diagnosis.<br /><br />The study involved six thoracic pathologists reviewing a large cohort of 263 LNET cases based on histopathological examination (HE) alone, as well as using the Ki-67 index and phosphohistone H3 (PPH3) expression. The goal was to assess interobserver variability and determine the effectiveness of deep learning algorithms.<br /><br />Logistic models were created to predict diagnosis based on the majority consensus of the pathologists. A mixed model was also developed to simulate a "super-pathologist" and evaluate the impact of the World Health Organization (WHO) criteria, Ki-67, and PPH3 count. These models were then compared to deep learning using a network specifically designed for LNET analysis.<br /><br />The results showed that unanimous agreement on LNET diagnosis based on the WHO criteria was only achieved in 63% of cases. High Ki-67 and PPH3 indices were associated with a worse prognosis and an increased likelihood of atypical carcinoid (AC) diagnosis. The use of deep learning algorithms improved the assessment of proliferative activity and revealed its spatial heterogeneity on whole slide images.<br /><br />The study highlights the limitations of the WHO criteria for LNET diagnosis and emphasizes the relevance of assessing proliferative activity for accurate diagnosis and prognosis. The researchers propose new tools and guidelines to improve the determination of LNET prognosis.<br /><br />This research demonstrates the potential of artificial intelligence and deep learning in enhancing the diagnosis and management of lung neuroendocrine tumors.

Asset Subtitle

Sylvie Lantuejoul

Meta Tag

Speaker Sylvie Lantuejoul

Topic Pathology & Biomarkers: Artificial Intelligence in Pathology

Keywords

lung neuroendocrine tumors

artificial intelligence

deep learning

proliferative activity

interobserver variability

Ki-67 index

phosphohistone H3 expression

pathologist consensus

atypical carcinoid diagnosis

prognosis determination