Machine learning-based gene alteration prediction model for primary lung cancer using cytologic images

BACKGROUND

Understanding the gene alteration status of primary lung cancers is important for determining treatment strategies, but gene testing is both time-consuming and costly, limiting its application in clinical practice. Here, potential therapeutic targets were selected by predicting gene alterations in cytologic specimens before conventional gene testing.

METHODS

This was a retrospective study to develop a cytologic image-based gene alteration prediction model for primary lung cancer. Photomicroscopic images of cytology samples were collected and image patches were generated for analyses. Cancer-positive (n = 106) and cancer-negative (n = 32) samples were used to develop a neural network model for selecting cancer-positive images. Cancer-positive cases were randomly assigned to training (n = 77) and validation (n = 26) data sets. Another neural network model was developed to classify cancer images of the training data set into 4 groups: anaplastic lymphoma kinase (ALK)-fusion, epidermal growth factor receptor (EGFR), or Kirsten rat sarcoma viral oncogene homologue (KRAS) mutated groups, and other (None group), and images of the validation data set were classified. A decision algorithm to predict gene alteration for cases with 3 probability ranks was developed.

RESULTS

The accuracy and precision for selecting cancer-positive patches were 0.945 and 0.991, respectively. Predictive accuracy for the EGFR and KRAS groups in the validation data set was ~0.95, whereas that for the ALK and None groups was ~0.75 and ~ 0.80, respectively. Gene status was correctly predicted in the probability rank A cases. The model extracted characteristic conventional cytologic findings in images and a novel specific feature was discovered for the EGFR group.

CONCLUSIONS

A gene alteration prediction model for lung cancers by machine learning based on cytologic images was successfully developed.

The Morphological Analysis of Cells in the Bronchoscopic Brushing and TBNA of Patients with Lung Adenocarcinoma

Biopsy, brushing, and transbronchial needle aspiration (TBNA) are the most common methods for diagnosis of lung adenocarcinoma and are taken during the same diagnostic bronchoscopic procedure. However, it is not clear what the morphological diagnostic criteria of cytology by brushing or TBNA are. A retrospective analysis was performed on 136 patients who underwent video bronchoscopy examination for diagnostic purposes. All the subjects were performed brushing or TBNA and confirmed as lung adenocarcinoma by biopsy or postoperative pathology. An additional 140 randomly selected patients with benign lung diseases were included in the study and used as a control group. The benign cells usually confused with adenocarcinoma cells were ciliated columnar cells, mucous columnar cells, ciliated cuboid cells, and reactive ciliated cells, respectively. The number of cases diagnosed as adenocarcinoma cells, carcinoma cells, suspicious cancer cells, and atypical proliferative cells by cytology was 101, 11, 20, and 4, respectively. The main basis for the interpretation of adenocarcinoma cells is the enlargement of individual nucleus, the arrangements of multistage papillary, and the general enlargement of nuclei, while the main clue for the interpretation of suspicious cancer cells and dysplasia cells comes from escape cells. The results suggested that the degree of nuclear enlargement, multiple papillary arrangement, and escape cells or escape trend cells are important clues for the interpretation of lung adenocarcinoma cells, while the atypical proliferative cells were similar to escape cells or escape trend cells, which were essentially benign cells beside the cancer.

Lung cancer and molecular testing in small biopsies versus cytology: The Logics of Worlds

The 8th Annual National Molecular Cytopathology Meeting, held in Naples, Italy, on December 2 to 3, 2019, addressed updates in diagnostic cytopathology and molecular classifications and specifically focused on lung cancer biomarker testing in cytology samples. Lung cancer continues to be the most commonly diagnosed noncutaneous malignancy in the world. In the majority of patients, lung cancers are frequently identified when they cannot be surgically accessed, and this leads to the use of cytology for a diagnosis and theragnostic testing. The meeting was an international forum for discussing new roles and updates for cytopathology in molecular testing as the basis for provoking new trends and novel approaches. The relevant literature is referenced. The significance of these updates for the practice of pathology in general is discussed.

Comparative Analysis for Diagnostic Yield of Small Cell Lung Cancer by Cytology and Histology During the Same Bronchoscopic Procedure

BACKGROUND

Biopsy, brushing, and transbronchial needle aspiration (TBNA) are the most common methods used for the diagnosis of small cell lung cancer during the same diagnostic bronchoscopic procedure. However, it is not clear which method provides better results.

PATIENTS AND METHODS

A retrospective analysis was performed of 140 patients who had undergone video bronchoscopy for diagnostic purposes. Bronchial brushings were obtained from all subjects. Biopsy specimens were also obtained from all subjects, except for 6 cases that could not be sampled; the TBNA method was used for some special lesions. The results were analyzed separately by histology and cytology.

RESULTS

The diagnostic yield of cytology was significantly greater than that of histology (P < .01) and that of conventional smear preparations in cytology was obviously greater than that of hematoxylin and eosin stains in histology (P < .01). The false-negative results were significantly lower with cytology than with histology (P < .01). Also, the cases of sampling site restriction with cytology were distinctly less than those with histology (P < .05). Stretch deformation of the tissue structure and cell morphology was the main reason for the false-negative results in the histologic diagnosis. The use of TBNA resolved all 4 cases of hilar adenopathy and 2 cases of lesions outside the bronchus. Multiple brushings of the tissue adjacent to cancer tissue and liquid-based preparations of cancerous necrotic tissue can significantly reduce the false-negative results from biopsy.

CONCLUSIONS

The diagnostic yield of cytologic examination of brushings and TBNA for small cell lung cancer was superior to that of histologic examination of hematoxylin and eosin stains and immunohistochemistry.

Diagnostic concordance of non-small cell lung carcinoma subtypes between biopsy and cytology specimens obtained during the same procedure

BACKGROUND

The objectives of this study were: 1) to determine the diagnostic concordance of non-small cell lung carcinoma (NSCLC) subtypes in cytology and biopsy specimens taken during the same procedure and evaluate the causes of discordance; and 2) to determine the frequency of immunohistochemistry (IHC) use for subtyping NSCLC.

METHODS

Biopsy and cytology specimens that were obtained at the same procedure and diagnosed as NSCLC between January 2011 and December 2014 at the McGill University Health Center were identified (n = 226 pairs). The diagnostic concordance between the 2 methods was evaluated. The slides from discordant cases were reviewed, and final diagnoses were made based on IHC, resection specimens, or pathologist discussion.

RESULTS

Concordance in subtype diagnosis was perfect (adeno-adeno or squamous-squamous) in 66.2% of cases and was partial (adeno or squamous vs non-small cell) in 23%; discordance (adeno vs squamous) was observed in 7.8%. Although subtyping was not possible (ie, the final diagnosis was NSCLC, not otherwise specified) in 12.8% of biopsy specimens and 16.3% of cytology specimens, specific subtyping was not achieved in only 3% of cases when both modalities were considered. IHC was used in 47% of biopsy cases and 13% of cytology cases.

CONCLUSIONS

Subtyping of NSCLC can be achieved in most cases (97%) by considering findings in both biopsy and cytology specimens, and concordance in subtyping between cytology and biopsy specimens can be reached in a high percentage of cases (89.2%). Cancer Cytopathol 2016;124:737-43. © 2016 American Cancer Society.

Bronchial wash cytology: A study on morphology and morphometry

Background:

Bronchial wash cytology of lung lesions is a non/minimally invasive procedure utilized for diagnosis of pulmonary lesions.

Aim:

The aim of this study was to evaluate the efficacy of bronchial wash cytology in the diagnosis of bronchopulmonary lesions and assess the role of morphometry in categorizing dysplastic/malignant lesions.

Materials and Methods:

All cases of bronchial wash cytology received from January 2006 to June 2010 were retrieved and reviewed. Cases with adequate clinical data or a subsequent biopsy were selected for the study and cytodiagnosis was correlated with available clinical details. Morphometry was done on alcohol fixed hematoxylin and eosin stained cytosmears using computer assisted Image Pro software.

Results:

One hundred and seventy-six cases of the 373 cases of bronchial cytology received were included for the study. Bronchial wash cytology technique showed high specificity. Cytohistopathology correlation showed 62.06% concordance rate. Cells from normal epithelium, reactive atypia, neoplastic atypia, squamous metaplasia, non-small cell and small cell carcinoma showed a mean nuclear diameter of 7.4 μm, 11.7 μm, 13.9 μm, 13.0 μm, 10.7 μm, and 17.7 μm, respectively, which was statistically significant with P < 0.05. Multiple comparisons between various groups using analysis of variance and Bonferroni tests also showed remarkable statistical significance.

Conclusions:

Bronchial wash cytology has low sensitivity in detecting pulmonary lesions. It can be of value in patients with contraindication for biopsy. Morphometry can be a useful adjunct to cytomorphology, especially in situations where biopsy is contraindicated.

The clinical significance of cytology versus histology-based diagnosis in small cell lung cancer: a retrospective study

OBJECTIVES

The aim of this study was to investigate the clinical significance of cytology versus histology-based diagnosis among patients diagnosed with small cell lung cancer (SCLC).

MATERIALS AND METHODS

Retrospective analysis of medical records of 443 patients with histologically or cytologically confirmed small cell lung carcinoma (SCLC) was performed. All patients received platinum-based chemotherapy regimens. Survival data (overall survival) were compared between patients with histology or cytology-based diagnosis in the overall study population as well as after stratification of patients according to disease stage (limited or extensive) at the time of diagnosis.

RESULTS

Distribution of demographics and clinicopathological characteristics among the two groups ("histology" and "cytology") was similar. No statistically significant differences in the survival curves between the "histology" and "cytology" groups were found in the overall study population (log rank test, p=0.237), as well as in the subgroup of patients with limited disease (log rank test, p=0.474). In contrast, patients with histology-based diagnosis had a statistically significant longer survival as compared to those with cytology-based diagnosis in the extensive disease subgroup (log rank test, p=0.031), but this association was not retained after adjusting the analysis for demographics and clinical characteristics via a Cox regression model (HR=1.18, 95% CI: 0.96-1.44, p=0.110).

CONCLUSION

The results of our study suggest that the type of diagnostic modality employed (histology or cytology-based) for the establishment of a diagnosis of SCLC may not have a significant effect on the overall survival of patients. Further studies are warranted to further investigate this important, yet rather unexplored, issue.

Histo-cytological diagnostic accuracy in lung cancer

OBJECTIVE

The majority of patients with lung cancer are treated on the basis of a diagnosis made from the analysis of a small tumour biopsy or a cytological sample and histotype is becoming a critical variable in clinical workup as it has led to the introduction of newer biologically targeted therapies. Consequently, simply classifying cancers as small cell lung cancers or non-small cell lung cancers is no longer sufficient.

METHODS

From 2009 to 2011, a review of the histo-cytological database was conducted to identify all small biopsy and cytology specimens collected for diagnostic purposes in patients with a thoracic lesion. In total, 941 patients were studied by examining exfoliative and/or aspirative cytological samples. To establish the accuracy of these methods, cytological and biopsy diagnoses were compared with each other and with subsequent resection specimens when available. Moreover, during the diagnostic workup, we examined a validated panel of immunohistochemical markers.

RESULTS

The diagnostic concordance of pre-operative diagnoses with surgical samples was high in both cytology and biopsy samples [κ = 0.71, confidence interval (CI) = 0.6-0.81; P < 0.0001 and κ = 0.61, CI = 0.41-0.82; P < 0.0001 respectively; good agreement] but concordance between cytology and biopsy was moderate (κ = 0.5, CI = 0.43-0.54; P < 0.0001). Immunohistochemistry-aided diagnoses were definitive for histotype in 92.8% of both cytology (206/222) and biopsy (155/167) specimens.

CONCLUSION

We found that lung cancer diagnosis and subtyping of cytology and biopsy samples are highly feasible and concordant; thus, the diagnostic approach to lung cancer does not require more invasive procedures.

Accuracy of fine needle aspiration cytology in the pathological typing of non-small cell lung cancer

BACKGROUND

Histological typing of non-small cell lung cancer (NSCLC) has an increasing clinical relevance due to the emerging differences in medical treatment between squamous and nonsquamous tumors. However, most NSCLCs are diagnosed in an advanced stage, and the diagnosis is often obtained exclusively by cytology either exfoliative or following fine needle aspiration. We investigated the accuracy of fine needle aspiration cytology (FNAC) in NSCLC typing as compared with histology.

METHODS

Over the period 2000-2009, 1182 transbronchial needle aspirate or transthoracic needle aspirate samples were obtained from patients with suspicious thoracic lesions. In 474 patients, a cytological diagnosis of primary NSCLC was obtained, and 186 (39%) of them (108 transbronchial needle aspirates and 78 transthoracic needle aspirates) received a parallel or subsequent histologic diagnosis on endoscopic biopsy (112) or surgery (74).

RESULTS

At cytology, 158 (85%) NSCLC cases were typed (89 adenocarcinoma and 69 squamous cell carcinoma), while 28 (15%) were classified as NSCLC not otherwise specified. At histology, 183 (98%) cases were typed (109 adenocarcinoma, 69 squamous cell carcinoma, 3 adenosquamous carcinoma, and 2 large cell carcinoma), and only 3 (2%) were classified as NSCLC not otherwise specified. Cytological and histological typing was concordant in 137 of 156 (88%) cases (K = 0.755; p < 0.001). The positive predictive value of FNAC in typing NSCLC was 92% for adenocarcinoma and 82% for squamous cell carcinoma.

CONCLUSION

FNAC in expert hands is fairly accurate for typing NSCLC and can be regarded as an acceptable procedure for diagnostic and medical treatment planning purposes in most NSCLC cases, especially when more invasive approaches are unfeasible. In poorly differentiated and doubtful cases, the use of ancillary techniques, such as immunocytochemistry, may be required to improve the diagnostic yield.