Deep Learning Facilitates Distinguishing Histologic Subtypes of Pulmonary Neuroendocrine Tumors on Digital Whole-Slide Images

Abstract

The histological distinction of lung neuroendocrine carcinoma, including small cell lung carcinoma (SCLC), large cell neuroendocrine carcinoma (LCNEC) and atypical carcinoid (AC), can be challenging in some cases, while bearing prognostic and therapeutic significance. To assist pathologists with the differentiation of histologic subtyping, we applied a deep learning classifier equipped with a convolutional neural network (CNN) to recognize lung neuroendocrine neoplasms. Slides of primary lung SCLC, LCNEC and AC were obtained from the Laboratory of Clinical and Experimental Pathology (University Hospital Nice, France). Three thoracic pathologists blindly established gold standard diagnoses. The HALO-AI module (Indica Labs, UK) trained with 18,752 image tiles extracted from 60 slides (SCLC = 20, LCNEC = 20, AC = 20 cases) was then tested on 90 slides (SCLC = 26, LCNEC = 22, AC = 13 and combined SCLC with LCNEC = 4 cases; NSCLC = 25 cases) by F1-score and accuracy. A HALO-AI correct area distribution (AD) cutoff of 50% or more was required to credit the CNN with the correct diagnosis. The tumor maps were false colored and displayed side by side to original hematoxylin and eosin slides with superimposed pathologist annotations. The trained HALO-AI yielded a mean F1-score of 0.99 (95% CI, 0.939–0.999) on the testing set. Our CNN model, providing further larger validation, has the potential to work side by side with the pathologist to accurately differentiate between the different lung neuroendocrine carcinoma in challenging cases.

Updates on lung neuroendocrine neoplasm classification

“…More recently, Iliè and colleagues investigated the adoption of a convolutional neural network (CNN) for the differential diagnosis of surgically resected pulmonary NENs. In a head‐to‐head comparison, AI‐driven algorithms distinguished the different entities with high accuracy (0.97 F1‐score, 0.93 AUC), with a degree of sensitivity and specificity comparable to conventional assessment, but also showed a slightly higher agreement than that of pathologists, suggesting a beneficial role of this model in assisting pathologists in the diagnostic work‐up 77 …”

“…In a head-to-head comparison, AI-driven algorithms distinguished the different entities with high accuracy (0.97 F1-score, 0.93 AUC), with a degree of sensitivity and specificity comparable to conventional assessment, but also showed a slightly higher agreement than that of pathologists, suggesting a beneficial role of this model in assisting pathologists in the diagnostic work-up. 77 Besides supporting morphological analyses, DIA can also be employed for standard assessment and quantification of prognostic biomarkers (e.g. Ki-67 proliferation index and mitotic counting using PPH3based IHC).…”

Deep Learning for Lung Cancer Diagnosis, Prognosis and Prediction Using Histological and Cytological Images: A Systematic Review

“…The studies of Yang and Kosaraju et al were the only ones that included LCNEC in the classifiers representing the realistic diagnostic practice for a pathologist. Ilié et al applied a DL algorithm for distinguishing SCLC, LCNEC, and atypical carcinoid (AC) [30]. A number of 150 H&E WSIs were included, and the model was in great agreement when compared to expert and general pathologists, achieving an AUC of 0.93.…”

The Third Joint Meeting on Lung Cancer of the FHU OncoAge (University Côte d’Azur, Nice, France) and the University of Texas MD Anderson Cancer Center (Houston, TX, USA). Understanding New Therapeutic Options and Promising Predictive Biomarkers for Lung Cancer Patients

“…In this Special Issue, we have brought together several articles focusing on recent topics related to lung cancer in the fields of fundamental, translational, and clinical research [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 ]. Basic research studies are essential to improve our knowledge of lung cancer carcinogenesis.…”

“…The increase in the number of biomarkers to be rapidly identified in NSCLC patients in daily practice requires the use of next-generation sequencing techniques [ 4 ]. An integrative approach will undoubtedly become necessary in the future, associating molecular genetic analyses with tumor tissue and/or fluids (from blood or other origins), multiplex immunochemical analyses, and algorithms based on artificial intelligence [ 3 , 4 , 24 , 25 , 26 ].…”