Integrating artificial intelligence into lung cancer screening: a randomised controlled trial protocol.

INTRODUCTION: Lung cancer (LC) is the most common cause of cancer-related deaths worldwide. Its early detection can be achieved with a CT scan. Two large randomised trials proved the efficacy of low-dose CT (LDCT)-based lung cancer screening (LCS) in high-risk populations. The decrease in specific mortality is 20%-25%.Nonetheless, implementing LCS on a large scale faces obstacles due to the low number of thoracic radiologists and CT scans available for the eligible population and the high frequency of false-positive screening results and the long period of indeterminacy of nodules that can reach up to 24 months, which is a source of prolonged anxiety and multiple costly examinations with possible side effects.Deep learning, an artificial intelligence solution has shown promising results in retrospective trials detecting lung nodules and characterising them. However, until now no prospective studies have demonstrated their importance in a real-life setting. METHODS AND ANALYSIS: This open-label randomised controlled study focuses on LCS for patients aged 50-80 years, who smoked more than 20 pack-years, whether active or quit smoking less than 15 years ago. Its objective is to determine whether assisting a multidisciplinary team (MDT) with a 3D convolutional network-based analysis of screening chest CT scans accelerates the definitive classification of nodules into malignant or benign. 2722 patients will be included with the aim to demonstrate a 3-month reduction in the delay between lung nodule detection and its definitive classification into benign or malignant. ETHICS AND DISSEMINATION: The sponsor of this study is the University Hospital of Nice. The study was approved for France by the ethical committee CPP (Comités de Protection des Personnes) Sud-Ouest et outre-mer III (No. 2022-A01543-40) and the Agence Nationale du Medicament et des produits de Santé (Ministry of Health) in December 2023. The findings of the trial will be disseminated through peer-reviewed journals and national and international conference presentations. TRIAL REGISTRATION NUMBER: NCT05704920.

Preliminary results in tracheal replacement using stented aortic matrices for primary extensive tracheal cancer.

Objective: Recent studies have demonstrated the feasibility and favorable long-term results of tracheobronchial replacement using stented cryopreserved aortic allografts. We propose to investigate the outcomes of this emerging technique in the subgroup of patients with extensive tracheal cancer. Methods: This study was based on 13 patients with primary extensive tracheal cancer extracted from the prospective registry TRITON-01 (ClinicalTrials.gov Identifier: NCT04263129), which included 40 patients in total. We analyzed early and late outcomes in this subset of patients. Results: From March 2019 to September 2022, 13 patients were included in the study. There were 9 female and 4 male patients, with a mean age of 53.9 years [36-71 years]. They had tracheal replacement for extended adenoid cystic carcinoma (n = 11), squamous cell carcinoma (n = 1), and mucoepidermoid carcinoma (n = 1). A venovenous extracorporeal membrane oxygenation was used in the 6 last cases. The mean length of resection was 81 mm [50-120 mm]. There was no 30-day postoperative mortality. A complete resection (R0) was achieved in 11 patients. The main late complications consisted of tracheal granulomas related to the stent and requiring repeated bronchoscopies (n = 9), pneumonia (n = 3), airway infection (n = 1), bronchoesophageal fistula (n = 1), mechanical stent obstruction requiring change (n = 2), and mediastinitis treated by antibiotics, drainage, and omentoplasty (n = 1). With a maximal follow-up of 3 years and 7 months, cancer recurrence was observed in 2 patients. All patients were alive at last follow-up except 2 (84.6%). Conclusions: Airway replacement using stented CAA represents a feasible and promising solution for extensive tracheal cancer.

Infections in autoimmune pulmonary alveolar proteinosis: a large retrospective cohort.

BACKGROUND: Autoimmune pulmonary alveolar proteinosis (aPAP) is a rare disease, predisposing to an increased risk of infection. A complete picture of these infections is lacking. RESEARCH QUESTION: Describe the characteristics and clinical outcomes of patients diagnosed with aPAP, and to identify risk factors associated with opportunistic infections. METHODS: We conducted a retrospective cohort including all patients diagnosed with aPAP between 2008 and 2018 in France and Belgium. Data were collected using a standardised questionnaire including demographics, comorbidities, imaging features, outcomes and microbiological data. RESULTS: We included 104 patients, 2/3 were men and median age at diagnosis was 45 years. With a median follow-up of 3.4 years (IQR 1.7-6.6 years), 60 patients (58%), developed at least one infection, including 23 (22%) with opportunistic infections. Nocardia spp was the main pathogen identified (n=10). Thirty-five (34%) patients were hospitalised due to infection. In univariate analysis, male gender was associated with opportunistic infections (p=0.04, OR=3.88; 95% CI (1.02 to 22.06)). Anti-granulocyte macrophage colony-stimulating factor antibody titre at diagnosis was significantly higher among patients who developed nocardiosis (1058 (316-1591) vs 580 (200-1190), p=0.01). Nine patients had died (9%), but only one death was related to infection. INTERPRETATION: Patients with aPAP often presented with opportunistic infections, especially nocardiosis, which highlights the importance of systematic search for slow-growing bacteria in bronchoalveolar lavage or whole lung lavage.

Ultra-Fast Amplicon-Based Next-Generation Sequencing in Non-Squamous Non-Small Cell Lung Cancer.

The number of molecular alterations to be tested for targeted therapy of non-squamous non-small cell lung cancer (NS-NSCLC) patients has significantly increased these last few years. The detection of molecular abnormalities is mandatory for the optimal care of advanced or metastatic NS-NSCLC patients, allowing targeted therapies to be administrated with an improvement in overall survival. Nevertheless, these tumors develop mechanisms of resistance that are potentially targetable using novel therapies. Some molecular alterations can also modulate the treatment response. The molecular characterization of NS-NSCLC has to be performed in a short turnaround time (TAT), in less than 10 working days, as recommended by the international guidelines. In addition, the origin of the tissue biopsies for genomic analysis is diverse, and their size is continuously decreasing with the development of less invasive methods and protocols. Consequently, pathologists are being challenged to perform effective molecular technics while maintaining an efficient and rapid diagnosis strategy. Here, we describe the ultra-fast amplicon-based next-generation sequencing (NGS) workflow used in daily routine practice at diagnosis for NS-NSCLC patients. We showed that this system is able to identify the current molecular targets used in precision medicine in thoracic oncology in an appropriate TAT.

The Need to Set up a Biobank Dedicated to Lymphoid Malignancies: Experience of a Single Center (Laboratory of Clinical and Experimental Pathology, University Côte d'Azur, Nice, France).

Several therapies to improve the management of lymphoma are currently being investigated, necessitating the development of new biomarkers. However, this requires high-quality and clinically annotated biological material. Therefore, we established a lymphoma biobank including all available biological material (tissue specimens and matched biological resources) along with associated clinical data for lymphoma patients diagnosed, according to the WHO classification, between 2005 and 2022 in the Laboratory of Clinical and Experimental Pathology, Nice, France. We retrospectively included selected cases in a new collection at the Côte d'Azur Biobank, which contains 2150 samples from 363 cases (351 patients). The male/female ratio was 1.3, and the median age at diagnosis was 58 years. The most common lymphoma types were classical Hodgkin lymphoma, diffuse large B-cell lymphoma, and extra-nodal marginal zone lymphoma of MALT tissue. The main sites of lymphoma were the mediastinum, lymph node, Waldeyer's ring, and lung. The Côte d'Azur Biobank is ISO 9001 and ISO 20387 certified and aims to provide high quality and diverse biological material to support translational research projects into lymphoma. The clinico-pathological data generated by this collection should aid the development of new biomarkers to enhance the survival of patients with lymphoid malignancies.

Comparison of outcome after stereotactic ablative radiotherapy of patients with metachronous lung versus primary lung cancer.

BACKGROUND: Early-stage lung cancer, primarily treated with surgery, often occur in poor surgical candidates (impaired respiratory function, prior thoracic surgery, severe comorbidities). Stereotactic ablative radiotherapy (SABR) is a non-invasive alternative that provides comparable local control. This technique is particularly relevant for surgically resectable metachronous lung cancer, in patients unable to undergo surgery.. The objective of this study is to evaluate the clinical outcome of patients treated with SABR for stage I metachronous lung cancer (MLC) versus stage I primary lung cancer (PLC). PATIENTS AND METHODS: 137 patients treated with SABR for stage I non-small cell lung cancer were retrospectively reviewed, of which 28 (20.4%) were MLC and 109 (79.6%) were PLC. Cohorts were evaluated for differences in overall survival (OS), progression-free survival (PFS), metastasis-free survival, local control (LC), and toxicity. RESULTS: After SABR, patients treated for MLC have comparable median age (76.6 vs 78.6, p = 0.2), 3-year LC (83.6% vs. 72.6%, p = 0.2), PFS (68.7% vs. 50.9%, p = 0.9), and OS (78.6% vs. 52.1%, p = 0.9) as PLC, along with similar rates of total (54.1% vs. 42.9%, p = 0.6) and grade 3 + toxicity (3.7% vs. 3.6%, p = 0.9). Previous treatment of MLC patients was either surgery (21/28, 75%) or SABR (7/28, 25%). The median follow-up was 53 months. CONCLUSION: SABR is a safe and effective approach for localized metachronous lung cancer.

Lack of correlation between MET and PD-L1 expression in non-small cell lung cancer revealed by comparative study of matched biopsies and surgical resection samples.

INTRODUCTION: Both MET expression and the PD-L1 tumor proportion score (TPS) are companion diagnostics for treatment of advanced non-small cell lung carcinoma (aNSCLC) patients. We evaluated the rate of correlation between MET expression and the PD-L1 TPS in matched biopsies and surgically resected specimens from NSCLC patients. PATIENTS AND METHODS: This retrospective analysis assessed the prevalence and correlation between MET expression (SP44 clone) and the PD-L1 TPS (22C3 clone) by immunohistochemistry together with molecular alterations determined by targeted next-generation sequencing in matched lung biopsy and surgically lung resected specimens from 70 patients with NSCLC. RESULTS: The study found a significant correlation between the MET H-score in surgical samples and matched biopsies (P-value < 0.0001), as well as between the PD-L1 TPS in paired biopsies and surgical samples (P-value < 0.0001). However, there was no significant correlation between the MET H-score or expression subgroups and the PD-L1 TPS in both types of paired samples (P-value = 0.47, and P-value = 0.90). The MET H-score was significantly higher in adenocarcinoma compared to squamous cell carcinoma (P-value < 0.0001). A mutational analysis showed that the MET H-score was significantly higher in NSCLC cases with targetable molecular alterations (P-value = 0.0095), while no significant correlation was found for the PD-L1 TPS. CONCLUSIONS: Our study found no significant correlation between PD-L1 and MET expression in samples from NSCLC patients, highlighting the importance of personalized treatment strategies based on individual expression profiles. These findings provide valuable insight into the development of effective immunotherapy and targeted therapy for NSCLC patients.

Rituximab and mycophenolate mofetil combination in patients with interstitial lung disease (EVER-ILD): a double-blind, randomised, placebo-controlled trial.

BACKGROUND: Standard of care for interstitial lung disease (ILD) with a nonspecific interstitial pneumonia (NSIP) pattern proposes mycophenolate mofetil (MMF) as one of the first-step therapies while rituximab is used as rescue therapy. METHODS: In a randomised, double-blind, two-parallel group, placebo-controlled trial (NCT02990286), patients with connective tissue disease-associated ILD or idiopathic interstitial pneumonia (with or without autoimmune features) and a NSIP pattern (defined on NSIP pathological pattern or on integration of clinicobiological data and a NSIP-like high-resolution computed tomography pattern) were randomly assigned in a 1:1 ratio to receive rituximab (1000 mg) or placebo on day 1 and day 15 in addition to MMF (2 g daily) for 6 months. The primary end-point was the change in percent predicted forced vital capacity (FVC) from baseline to 6 months analysed by a linear mixed model for repeated measures analysis. Secondary end-points included progression-free survival (PFS) up to 6 months and safety. FINDINGS: Between January 2017 and January 2019, 122 randomised patients received at least one dose of rituximab (n=63) or placebo (n=59). The least-squares mean change from baseline to 6 months in FVC (% predicted) was +1.60 (se 1.13) in the rituximab+MMF group and -2.01 (se 1.17) in the placebo+MMF group (between-group difference 3.60, 95% CI 0.41-6.80; p=0.0273). PFS was better in the rituximab+MMF group (crude hazard ratio 0.47, 95% CI 0.23-0.96; p=0.03). Serious adverse events occurred in 26 (41%) patients of the rituximab+MMF group and in 23 (39%) of the placebo+MMF group. Nine infections were reported in the rituximab+MMF group (five bacterial infections, three viral infections, one other) and four bacterial infections in the placebo+MMF group. INTERPRETATION: Combination of rituximab and MMF was superior to MMF alone in patients with ILD and a NSIP pattern. The use of this combination must take into consideration the risk of viral infection.

Accurate Detection of SARS-CoV-2 by Next-Generation Sequencing in Low Viral Load Specimens.

As new SARS-CoV-2 variants emerge, there is an urgent need to increase the efficiency and availability of viral genome sequencing, notably to detect the lineage in samples with a low viral load. SARS-CoV-2 genome next-generation sequencing (NGS) was performed retrospectively in a single center on 175 positive samples from individuals. An automated workflow used the Ion AmpliSeq SARS-CoV-2 Insight Research Assay on the Genexus Sequencer. All samples were collected in the metropolitan area of the city of Nice (France) over a period of 32 weeks (from 19 July 2021 to 11 February 2022). In total, 76% of cases were identified with a low viral load (Ct ≥ 32, and ≤200 copies/µL). The NGS analysis was successful in 91% of cases, among which 57% of cases harbored the Delta variant, and 34% the Omicron BA.1.1 variant. Only 9% of cases had unreadable sequences. There was no significant difference in the viral load in patients infected with the Omicron variant compared to the Delta variant (Ct values, p = 0.0507; copy number, p = 0.252). We show that the NGS analysis of the SARS-CoV-2 genome provides reliable detection of the Delta and Omicron SARS-CoV-2 variants in low viral load samples.

French practical guidelines for the diagnosis and management of idiopathic pulmonary fibrosis - 2021 update. Full-length version.

BACKGROUND: Since the latest 2017 French guidelines, knowledge about idiopathic pulmonary fibrosis has evolved considerably. METHODS: Practical guidelines were drafted on the initiative of the Coordinating Reference Center for Rare Pulmonary Diseases, led by the French Language Pulmonology Society (SPLF), by a coordinating group, a writing group, and a review group, with the involvement of the entire OrphaLung network, pulmonologists practicing in various settings, radiologists, pathologists, a general practitioner, a health manager, and a patient association. The method followed the "Clinical Practice Guidelines" process of the French National Authority for Health (HAS), including an online vote using a Likert scale. RESULTS: After a literature review, 54 guidelines were formulated, improved, and then validated by the working groups. These guidelines addressed multiple aspects of the disease: epidemiology, diagnostic procedures, quality criteria and interpretation of chest CT scans, lung biopsy indication and procedures, etiological workup, methods and indications for family screening and genetic testing, assessment of the functional impairment and prognosis, indication and use of antifibrotic agents, lung transplantation, management of symptoms, comorbidities and complications, treatment of chronic respiratory failure, diagnosis and management of acute exacerbations of fibrosis. CONCLUSION: These evidence-based guidelines are intended to guide the diagnosis and practical management of idiopathic pulmonary fibrosis.

Ultrafast Gene Fusion Assessment for Nonsquamous NSCLC.

Introduction: Gene fusion testing of ALK, ROS1, RET, NTRK, and MET exon 14 skipping mutations is guideline recommended in nonsquamous NSCLC (NS-NSCLC). Nevertheless, assessment is often hindered by the limited availability of tissue and prolonged next-generation sequencing (NGS) testing, which can protract the initiation of a targeted therapy. Therefore, the development of faster gene fusion assessment is critical for optimal clinical decision-making. Here, we compared two ultrafast gene fusion assays (UFGFAs) using NGS (Genexus, Oncomine Precision Assay, Thermo Fisher Scientific) and a multiplex reverse-transcriptase polymerase chain reaction (Idylla, GeneFusion Assay, Biocartis) approach at diagnosis in a retrospective series of 195 NS-NSCLC cases and five extrapulmonary tumors with a known NTRK fusion. Methods: A total of 195 NS-NSCLC cases (113 known gene fusions and 82 wild-type tumors) were included retrospectively. To validate the detection of a NTRK fusion, we added five NTRK-positive extrathoracic tumors. The diagnostic performance of the two UFGFAs and standard procedures was compared. Results: The accuracy was 92.3% and 93.1% for Idylla and Genexus, respectively. Both systems improved the sensitivity for detection by including a 5'-3' imbalance analysis. Although detection of ROS1, MET exon 14 skipping, and RET was excellent with both systems, ALK fusion detection was reduced with sensitivities of 87% and 88%, respectively. Idylla had a limited sensitivity of 67% for NTRK fusions, in which only an imbalance assessment was used. Conclusions: UFGFA using NGS and reverse-transcriptase polymerase chain reaction approaches had an equal level of detection of gene fusion but with some technique-specific limitations. Nevertheless, UFGFA detection in routine clinical care is feasible with both systems allowing faster initiation of therapy and a broad degree of screening.

Deciphering the Impact of HER2 Alterations on Non-Small-Cell Lung Cancer: From Biological Mechanisms to Therapeutic Approaches.

Despite the recent increase in the number of types of treatments, non-small-cell lung cancer (NSCLC) remains the major cause of death from cancer worldwide. So, there is an urgent need to develop new therapeutic strategies. The HER2 gene codes for tyrosine kinase receptor whose alterations are known to drive carcinogenesis. HER2 alterations, including amplification, mutations, and overexpression, have been mainly described in breast and gastric cancers, but up to 4% of NSCLC harbor actionable HER2 mutations. HER2-targeted therapy for NSCLC with trastuzumab, pertuzumab, and trastuzumab emtansine has failed to demonstrate an improvement in survival. Nevertheless, recent data from phase II trials have shed light on promising specific therapies for HER2-mutant NSCLC such as trastuzumab deruxtecan. Herein, we aimed to provide an updated review on the biology, epidemiology, molecular testing, and therapeutic strategies for NSCLC with HER2 molecular alterations.

Molecular Profiling in Non-Squamous Non-Small Cell Lung Carcinoma: Towards a Switch to Next-Generation Sequencing Reflex Testing.

Molecular diagnosis of lung cancer is a constantly evolving field thanks to major advances in precision oncology. The wide range of actionable molecular alterations in non-squamous non-small cell lung carcinoma (NS-NSCLC) and the multiplicity of mechanisms of resistance to treatment resulted in the need for repeated testing to establish an accurate molecular diagnosis, as well as to track disease evolution over time. While assessing the increasing complexity of the molecular composition of tumors at baseline, as well as over time, has become increasingly challenging, the emergence and implementation of next-generation sequencing (NGS) testing has extensively facilitated molecular profiling in NS-NSCLC. In this review, we discuss recent developments in the molecular profiling of NS-NSCLC and how NGS addresses current needs, as well as how it can be implemented to address future challenges in the management of NS-NSCLC.

Airway replacement using stented aortic matrices: Long-term follow-up and results of the TRITON-01 study in 35 adult patients.

Over the past 25 years, we have demonstrated the feasibility of airway bioengineering using stented aortic matrices experimentally then in a first-in-human trial (n = 13). The present TRITON-01 study analyzed all the patients who had airway replacement at our center to confirm that this innovative approach can be now used as usual care. For each patient, the following data were prospectively collected: postoperative mortality and morbidity, late airway complications, stent removal and status at last follow-up on November 2, 2021. From October 2009 to October 2021, 35 patients had airway replacement for malignant (n = 29) or benign (n = 6) lesions. The 30-day postoperative mortality and morbidity rates were 2.9% (n = 1/35) and 22.9% (n = 8/35) respectively. At a median follow-up of 29.5 months (range 1-133 months), 27 patients were alive. There have been no deaths directly related to the implanted bioprosthesis. Eighteen patients (52.9%) had stent-related granulomas requiring a bronchoscopic treatment. Ten among 35 patients (28.6%) achieved a stent free survival. The actuarial 2- and 5-year survival rates (Kaplan-Meier estimates) were respectively 88% and 75%. The TRITON-01 study confirmed that airway replacement using stented aortic matrices can be proposed as usual care at our center. Clinicaltrials.gov Identifier: NCT04263129.

Setting Up an Ultra-Fast Next-Generation Sequencing Approach as Reflex Testing at Diagnosis of Non-Squamous Non-Small Cell Lung Cancer; Experience of a Single Center (LPCE, Nice, France).

The number of genomic alterations required for targeted therapy of non-squamous non-small cell lung cancer (NS-NSCLC) patients has increased and become more complex these last few years. These molecular abnormalities lead to treatment that provides improvement in overall survival for certain patients. However, these treated tumors inexorably develop mechanisms of resistance, some of which can be targeted with new therapies. The characterization of the genomic alterations needs to be performed in a short turnaround time (TAT), as indicated by the international guidelines. The origin of the tissue biopsies used for the analyses is diverse, but their size is progressively decreasing due to the development of less invasive methods. In this respect, the pathologists are facing a number of different challenges requiring them to set up efficient molecular technologies while maintaining a strategy that allows rapid diagnosis. We report here our experience concerning the development of an optimal workflow for genomic alteration assessment as reflex testing in routine clinical practice at diagnosis for NS-NSCLC patients by using an ultra-fast-next generation sequencing approach (Ion Torrent Genexus Sequencer, Thermo Fisher Scientific). We show that the molecular targets currently available to personalized medicine in thoracic oncology can be identified using this system in an appropriate TAT, notably when only a small amount of nucleic acids is available. We discuss the new challenges and the perspectives of using such an ultra-fast NGS in daily practice.

Autophagopathies: from autophagy gene polymorphisms to precision medicine for human diseases.

At a time when complex diseases affect globally 280 million people and claim 14 million lives every year, there is an urgent need to rapidly increase our knowledge into their underlying etiologies. Though critical in identifying the people at risk, the causal environmental factors (microbiome and/or pollutants) and the affected pathophysiological mechanisms are not well understood. Herein, we consider the variations of autophagy-related (ATG) genes at the heart of mechanisms of increased susceptibility to environmental stress. A comprehensive autophagy genomic resource is presented with 263 single nucleotide polymorphisms (SNPs) for 69 autophagy-related genes associated with 117 autoimmune, inflammatory, infectious, cardiovascular, neurological, respiratory, and endocrine diseases. We thus propose the term 'autophagopathies' to group together a class of complex human diseases the etiology of which lies in a genetic defect of the autophagy machinery, whether directly related or not to an abnormal flux in autophagy, LC3-associated phagocytosis, or any associated trafficking. The future of precision medicine for common diseases will lie in our ability to exploit these ATG SNP x environment relationships to develop new polygenetic risk scores, new management guidelines, and optimal therapies for afflicted patients.Abbreviations: ATG, autophagy-related; ALS-FTD, amyotrophic lateral sclerosis-frontotemporal dementia; ccRCC, clear cell renal cell carcinoma; CD, Crohn disease; COPD, chronic obstructive pulmonary disease; eQTL, expression quantitative trait loci; HCC, hepatocellular carcinoma; HNSCC, head and neck squamous cell carcinoma; GTEx, genotype-tissue expression; GWAS, genome-wide association studies; LAP, LC3-associated phagocytosis; LC3-II, phosphatidylethanolamine conjugated form of LC3; LD, linkage disequilibrium; LUAD, lung adenocarcinoma; MAF, minor allele frequency; MAP1LC3/LC3: microtubule associated protein 1 light chain 3; NSCLC, non-small cell lung cancer; OS, overall survival; PtdIns3K CIII, class III phosphatidylinositol 3 kinase; PtdIns3P, phosphatidylinositol-3-phosphate; SLE, systemic lupus erythematosus; SNPs, single-nucleotide polymorphisms; mQTL, methylation quantitative trait loci; ULK, unc-51 like autophagy activating kinase; UTRs, untranslated regions; WHO, World Health Organization.

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

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.