Efficacy of exponentiation method with a convolutional neural network for classifying lung nodules on CT images by malignancy level.

OBJECTIVES: The aim of this study was to examine the performance of a convolutional neural network (CNN) combined with exponentiating each pixel value in classifying benign and malignant lung nodules on computed tomography (CT) images. MATERIALS AND METHODS: Images in the Lung Image Database Consortium-Image Database Resource Initiative (LIDC-IDRI) were analyzed. Four CNN models were then constructed to classify the lung nodules by malignancy level (malignancy level 1 vs. 2, malignancy level 1 vs. 3, malignancy level 1 vs. 4, and malignancy level 1 vs. 5). The exponentiation method was applied for exponent values of 1.0 to 10.0 in increments of 0.5. Accuracy, sensitivity, specificity, and area under the curve of receiver operating characteristics (AUC-ROC) were calculated. These statistics were compared between an exponent value of 1.0 and all other exponent values in each model by the Mann-Whitney U-test. RESULTS: In malignancy 1 vs. 4, maximum test accuracy (MTA; exponent value = 2.0, 3.0, 3.5, 4.5, 5.0, 5.5, 6.0, 6.5, 7.0, 7.5, 8.0, 8.5, 9.0, 9.5, and 10.0) and specificity (6.5, 7.0, and 9.0) were improved by up to 0.012 and 0.037, respectively. In malignancy 1 vs. 5, MTA (6.5 and 7.0) and sensitivity (1.5) were improved by up to 0.030 and 0.0040, respectively. CONCLUSIONS: The exponentiation method improved the performance of the CNN in the task of classifying lung nodules on CT images as benign or malignant. The exponentiation method demonstrated two advantages: improved accuracy, and the ability to adjust sensitivity and specificity by selecting an appropriate exponent value. CLINICAL RELEVANCE STATEMENT: Adjustment of sensitivity and specificity by selecting an exponent value enables the construction of proper CNN models for screening, diagnosis, and treatment processes among patients with lung nodules. KEY POINTS: • The exponentiation method improved the performance of the convolutional neural network. • Contrast accentuation by the exponentiation method may derive features of lung nodules. • Sensitivity and specificity can be adjusted by selecting an exponent value.

The clinical significance of tertiary lymphoid structure and its relationship with peripheral blood characteristics in patients with surgically resected non-small cell lung cancer: a single-center, retrospective study.

INTRODUCTION: The presence of tertiary lymphoid structure (TLS) in tumor tissues has been reported to be a factor associated with a good prognosis in several types of cancers. However, the relationship between TLS formation and peripheral blood findings remains unclear. The purposes of the study were to evaluate the effect of the presence of TLS on survival and determine the peripheral blood characteristics associated with TLS formation in non-small cell lung cancer (NSCLC) patients. METHODS: A total of 147 consecutive NSCLC patients who underwent lung resection at Fukushima Medical University Hospital between 2013 and 2017 were enrolled. TLS expression was evaluated, and the relationships between clinical parameters and outcomes were analyzed. Peripheral blood mononuclear cells (PBMCs) were further analyzed by mass cytometry to characterize the TLS-positive microenvironment. RESULTS: Forty-six patients had high TLS expression, and the remaining 101 patients had low TLS expression. In stage II to IV patients (n = 35), disease-free survival was longer in the high TLS expression group (p = 0.027). A low neutrophil to lymphocyte ratio (NLR) < 2.75 in the peripheral blood was associated with high TLS expression (p = 0.003). Citrus analysis after mass cytometry assay showed that the number of cells expressing HLA-DR and CD9 in PBMCs was lower in the high TLS expression group. CONCLUSION: High TLS expression is associated with a good prognosis after surgery in stage II and III NSCLC patients. In the peripheral blood, a low NLR and few antigen-presenting cells indicate the presence of TLS in the tumor microenvironment.

[Resistance Mechanisms to Immune Checkpoint Inhibitor and Its Overcome with Focus on ß-Catenin in Lung Cancer].

Although the indications for immune checkpoint inhibitors are expanding rapidly, the disease will eventually progress in many patients. Elucidating and overcoming the resistant mechanisms to immune checkpoint inhibitors is a major challenge. WNT/ß-catenin pathway has long been known as one of the mechanisms involved in cell proliferation and epithelial-mesenchymal transition in cancer development. Recently, it has become clear that WNT/ß-catenin pathway also plays a role in cancer immune escape, as reported in melanoma. We have also studied WNT/ß-catenin pathway as a mechanism of immune escape in lung cancer. In this article, we review how WNT/ß-catenin pathway is involved in immune escape and resistance to immune checkpoint inhibitors, mainly in non-small cell lung cancer. In addition, we discuss how to overcome the tumor immune mechanism caused by WNT/ß-catenin pathway in the context of current combination therapies and therapies in development.

[ß-Catenin Expression in Non-Small Cell Lung Cancer and Therapeutic Effect of Immune Checkpoint Inhibitors].

Recently, ß-catenin mediated immune escape mechanism has been reported in several cancers. We investigated whether ß-catenin is associated with resistance to immune checkpoint inhibitor therapy in non-small cell lung cancer. Non-small cell lung cancer patients expressing high levels of ß-catenin showed poor progression-free survival and overall survival after single agent anti-PD-1 therapy. They had less infiltration of CD8-positive cells and antigen-presenting cells. Microarray analysis also showed low gene expression of CD8A and IFNG. siRNA knockdown of CTNNB1 in the ß-catenin-positive lung cancer cell line LK-2 tended to decrease CTNNB1 and ATF3 expression and increase CCL4 expression. The results suggest that ß- catenin suppresses tumor infiltration by antigen-presenting cells and confers resistance to immune checkpoint inhibitors in non-small cell lung cancer via downregulation of CCL4 production.

Tumor mutation burden and immunological, genomic, and clinicopathological factors as biomarkers for checkpoint inhibitor treatment of patients with non-small-cell lung cancer.

Cancer treatment using immune checkpoint inhibitors is widely used, although biomarkers predictive of response are not well established. However, both the expressions of programmed cell death ligand 1 (PD-L1) and the tumor mutation burden (TMB) hold promise as such biomarkers for immune checkpoint inhibitors; however, its characteristics and clinical and immunological impacts have not been fully analyzed. We, therefore, evaluated the clinical and immunological parameters related to TMB to identify potential new biomarkers. We enrolled 92 patients with non-small-cell lung cancer who underwent surgery at Fukushima Medical University Hospital from 2013 to 2016. TMB of individual tumors was calculated by whole-exome sequencing analysis. Major cancer-related gene mutations were evaluated using panel sequencing. Expression of PD-L1 and abundance of tumor-infiltrating lymphocytes were evaluated by immunohistochemistry using surgical samples. The median TMB value was 60. TMB was significantly higher in men, current or former smokers, and in patients with squamous cell carcinoma, tumor size ≥ 2.8 cm, wild-type EGFR, TP53 gene mutation-positive status, and cyclin-dependent kinase-inhibitor gene 2A mutation-positive status. According to multivariate analysis, TMB was significantly associated with EGFR gene mutation-negative status (p = 0.0111) and TP53 gene mutation-positive status (p = 0.0425). If TMB is identified as a robust biomarker for immune checkpoint inhibitor administration, analysis of TP53 and EGFR mutations may provide a relatively rapid and easy proxy for predicting TMB.

[Monitoring Tumor Infiltrating Lymphocytes by Peripheral Blood in Lung Cancer Patients].

There have been many reports on the association between tumor infiltrating lymphocytes and cancer prognosis. It is known that tumor infiltrating lymphocytes contain not only cytotoxic T lymphocytes but also bystander lymphocytes and immunosuppressive cells. In most of previous reports, tumor infiltrating lymphocytes were defined as CD3 or CD8 T cells. It is generally thought that patients with cancer rich in tumor infiltrating lymphocytes have a good prognosis. Most tumor infiltrating lymphocytes are thought to be cytotoxic T lymphocytes. It is also reported that cancer rich in tumor infiltrating lymphocytes is responsive to immune checkpoint inhibitors. In recent years, several reports revealed clonal replacement in tumor infiltrating lymphocytes after administration of immune checkpoint inhibitors. This change was also detectable in peripheral blood. From the viewpoint of lung cancer treatment, combination of immune checkpoint inhibitors and chemotherapy became the standard therapy. We need to understand the tumor immune microenvironment in order to select the best treatment regimen for each patient. However, it is often difficult to obtain an adequate amount of tissue biopsy sample in standard of care. It is hoped that we can understand the tumor immune microenvironment using the peripheral blood. Thus, studying the association between treatment response, tumor infiltrating lymphocytes, and peripheral blood is considered to be important to research and develop peripheral blood biomarkers in lung cancer.

[Exploring the Possibility of New Biomarkers of Immune Checkpoint Inhibitors for Non-Small Cell Lung Cancer (NSCLC)].

Mutation burden in a tumor, presumably involving neo-antigens in the tumor tissue, is also thought to be one of the better predictors for the efficacy of immune checkpoint inhibitors. However, it is difficult to analyze the mutation burden routinely in the clinic. Here, we describe more convenient factors that can be used as surrogate markers of mutation burden. Ninety-four patients with NSCLC who underwent resection in our institution were recruited for this study. Mutation burden and major gene alterations were analyzed by using next generation sequencing. Several immunological parameters were also assessed using immunohistochemistry. Statistical analysis was performed on mutation burden, major gene alternations, immunohistochemistry, and clinical parameters. The median mutation load was 54 mutations(range, 10-363 mutations). Squamous cell carcinoma, EGFRmutation -negativity, and TP53 alteration-positivity were closely connected with higher mutation burden. Multiple regression analysis showed that mutation burden in the tumor could be associated with EGFRmutation and TP53 alteration status.

FDG-PET in the evaluation of response to nivolumab in recurrent non-small-cell lung cancer.

BACKGROUND: Nivolumab, an immune checkpoint inhibitor, is recently clinically applied to non-small cell lung cancer (NSCLC) treatment, and this causes T cell activation and T cell infiltration to tumor tissue through the blockade of the interaction between programmed cell death 1 (PD-1) and programmed cell death ligand 1 (PD-L1). 18F-fluorodeoxyglucose-positron emission tomography (FDG-PET) sometimes shows false positive because of the recruitment of neutrophils, lymphocytes, and macrophages. To date, there is only one report except our case, which described the correlation between FDG-PET and nivolumab. CASE PRESENTATION: We report on a 75-year-old man on nivolumab treatment for metastatic non-small cell lung cancer. He had undergone right lower lobectomy for lung adenocarcinoma in the right S8 segment 10 months prior to recurrence. Pathological findings revealed invasive adenocarcinoma, pT1bN2M0 stage IIIA. Epidermal growth factor receptor (EGFR) mutation was positive for de novo T790M and anaplastic lymphoma kinase (ALK) rearrangement was negative. Immunohistochemistry was negative for PD-L1. He underwent chemotherapy with a combination of cisplatin and pemetrexed for four cycles but developed progressive disease involving the right hemithorax, multiple lymph nodes, and multiple osseous sites. Nivolumab was instituted as a second-line chemotherapy. After six courses of this immunotherapy, FDG-PET scan showed decreased FDG uptake in each recurrent lesion despite T lymphocyte activation by nivolumab. Serum carcinoembryonic antigen (CEA) level was also remarkably decreased. CONCLUSIONS: Nivolumab's effect on recurrent NSCLC may be monitored by PET; larger studies are needed.

Anaplastic Lymphoma Kinase-Tyrosine Kinase Inhibitor Treatment for Metastatic Non-small Cell Lung Cancer Throughout the Entire Gestation Period.

Non-small cell lung cancer (NSCLC) occasionally develops in younger, fertile patients. This early-onset NSCLC tends to have more oncogenic driver mutations than in aged patients. Among early-onset NSCLC patients, pregnancy is very rare. However, there are some patients who were able to balance tyrosine kinase inhibitor (TKI) administration and pregnancy. Here, we report a case of a pregnancy under alectinib hydrochloride (a second-generation anaplastic lymphoma kinase (ALK)-TKI) administration throughout the entire gestational period for ALK-rearranged metastatic lung adenocarcinoma. The patient was an Asian female in her early 20s who became aware of her pregnancy after diagnosis and the start of alectinib administration. She intended to have the baby despite the necessity of continuing her treatment and the unknown risks involved. A multidisciplinary team (thoracic surgeon, obstetrics, pediatrics, and so on) was organized to support the patient, baby, and family. There were no obvious signs of tumor progression during pregnancy. She gave birth at 41 weeks and one day of gestation. There was no placental metastasis. Alectinib concentration at delivery was 155 ng/mL in maternal blood, 22.1 ng/mL in umbilical cord venous blood, 20.1 ng/mL in amniotic fluid, and 11.8 ng/mL in colostrum. The baby had been exposed to alectinib throughout the entire pregnancy; however, fetal growth curve parameters remained within the normal ranges and the baby developed without anatomical or neurodevelopmental anomalies or fetal metastasis for the first 13 months of age.