Cytological typing of primary lung cancer: study of 100 cases with autopsy confirmation

Deep convolutional neural network-based classification of cancer cells on cytological pleural effusion images

Lung cancer is one of the leading causes of cancer-related death worldwide. Cytology plays an important role in the initial evaluation and diagnosis of patients with lung cancer. However, due to the subjectivity of cytopathologists and the region-dependent diagnostic levels, the low consistency of liquid-based cytological diagnosis results in certain proportions of misdiagnoses and missed diagnoses. In this study, we performed a weakly supervised deep learning method for the classification of benign and malignant cells in lung cytological images through a deep convolutional neural network (DCNN). A total of 404 cases of lung cancer cells in effusion cytology specimens from Shanghai Pulmonary Hospital were investigated, in which 266, 78, and 60 cases were used as the training, validation and test sets, respectively. The proposed method was evaluated on 60 whole-slide images (WSIs) of lung cancer pleural effusion specimens. This study showed that the method had an accuracy, sensitivity, and specificity respectively of 91.67%, 87.50% and 94.44% in classifying malignant and benign lesions (or normal). The area under the receiver operating characteristic (ROC) curve (AUC) was 0.9526 (95% confidence interval (CI): 0.9019-9.9909). In contrast, the average accuracies of senior and junior cytopathologists were 98.34% and 83.34%, respectively. The proposed deep learning method will be useful and may assist pathologists with different levels of experience in the diagnosis of cancer cells on cytological pleural effusion images in the future.

Sequential use of bronchial aspirates, biopsies and washings in the preoperative management of lung cancers

BACKGROUND

The combination of cytology and biopsies improves the recognition and typing of small cell (SCLC) versus non small cell (NSCLC) lung cancers in the fiberoptic bronchoscopy assessment of centrally located tumours.

METHODS

We studied whether bronchial aspirates performed before biopsies (BA) and washings performed after biopsies (BW) could increase the diagnostic yield of fiberoptic bronchoscopy. A series of 334 consecutive samples taken in patients with suspicious fiberoptic bronchoscopy findings was studied. Two hundred primary tumours were included in the study. The actual diagnosis was based on surgical tissue specimen analysis and/or imaging techniques. The typing used was that of the 1999 WHO/IASLC classification.

RESULTS

The diagnosis of malignancy and tumour typing were analyzed according to the sequential (combined) or single use of tests. Malignancy was assessed by cytology in 144/164 (87.8%) positive biopsy cases and in 174/200 tumour cases (87.0%). BA before biopsies allowed 84.0% of cancers to be diagnosed, whereas BW after biopsies allowed 79.0% of cancers to be found (p = ns). However, combining biopsies with BW allowed 94.0% of cancers to be diagnosed, whereas 82.0% were diagnosed by biopsies alone (p < 0.001). The highest diagnostic yield was obtained with the combination of BA, biopsies and BW, with 97.0% sensitivity. Exact concordance in typing was obtained in 83.8% of cases. The six surgically resected cases (3.0%) with negative cytology and biopsy results included four squamous cell carcinomas with necrotizing or fibrous surface and two adenocarcinomas, pT1 stage.

CONCLUSION

Fiberoptic bronchoscopy may reach a yield of close to 100% in the diagnosis and typing of centrally located, primary lung cancers by combining bronchial aspirates, biopsies and washings.

Utility of Rapid On-Site Evaluation of Transbronchial Needle Aspirates

BACKGROUND

Rapid on-site evaluation has been proposed as a method to improve the yield of transbronchial needle aspiration.

OBJECTIVES

This study investigated whether on-site analysis facilitates routine diagnostic bronchoscopy in terms of sampling, yield and cost.

METHODS

Patients with lesions accessible for transbronchial needle aspiration on computed tomography were investigated. A cytopathologist screened the needle aspirates on site for the presence of diagnostic material. The bronchoscopic sampling process was adjusted according to the results. In 90 consecutive patients with neoplastic disease (n=70; 78%), non-neoplastic disease (n=16; 18%) or undiagnosed lesions (n=4; 4%) we aspirated 162 lung tumours or lymph node sites (mediastinal: 7%; tracheobronchial: 68%; other: 25%). In 90 consecutive patients with neoplastic disease (n=70; 78%), non-neoplastic disease (n=16; 18%) or undiagnosed lesions (n=4; 4%) we aspirated 162 lung lesions (paratracheal tumours or lymph nodes: 7%; tracheobronchial lymph nodes: 68%; other: 25%).

RESULTS

The diagnostic yield of needle aspiration was 77 and 25% in patients with neoplastic and non-neoplastic lesions, respectively. Sampling could be terminated in 64% of patients after needle aspiration had been performed as the only diagnostic modality, and on-site analysis identified diagnostic material from the first site aspirated in 50% of patients. Only in 2 patients (2%) diagnostic aspirates were not recognized on site. On-site analysis was cost effective due to savings for disposable diagnostic tools, which exceeded the extra expense for the on-site cytology service provided.

CONCLUSIONS

Rapid on-site analysis of transbronchial aspirates is a highly useful, accurate and cost-effective addition to routine diagnostic bronchoscopy.

Accuracy of touch imprint cytology in diagnosing lung cancer

A 3-year study assessed the diagnostic accuracy of touch imprint smears in the diagnosis of lung cancer. Touch imprint smears were prepared from 90 computerized tomographic-guided core needle lung biopsies. Cytological diagnosis of touch imprint smears were correlated with the histological diagnosis of the corresponding core needle biopsy specimen, which was taken as the gold standard. The sensitivity, specificity, positive predictive value and negative predictive value of imprint smear results were 89%, 100%, 100% and 68%, respectively. There were no false positives, and all patients with small cell lung cancer were correctly diagnosed with this technique. Imprint cytology can be used to provide a rapid, preliminary diagnosis of lung cancer.

Small cell lung carcinoma (SCLC): a clinicopathologic study of 100 cases with surgical specimens

Separation of small cell lung carcinoma (SCLC) from nonsmall cell lung carcinoma (NSCLC) is a critical distinction to be made in the diagnosis of lung cancer. However, the diagnosis of SCLC is most commonly made on small biopsies and cytologic specimens, and practicing pathologists may not be familiar with all its morphologic guises and frequent combination with NSCLC elements, which may be seen in larger specimens. Following the most recent WHO classification of lung tumors and with the hope of identifying prognostic markers, we examined in detail the histology of 100 surgical biopsies or resections with a diagnosis of SCLC from the AFIP and pathology panel of the International Association for the Study of Lung Cancer (IASLC). Multiple clinical and histologic features were studied by Kaplan-Meier analysis. Neuroendocrine architectural patterns, including nested and trabecular growth, with peripheral palisading and rosette formation were common in SCLC. Necrosis and apoptotic debris was prominent in all cases, but crush artifact was infrequent. Cell size in surgical biopsy specimens appears larger than in bronchoscopic biopsy specimens and occasional cells may show prominent nucleoli and vesicular nuclear chromatin, but this does not preclude the diagnosis of SCLC. A high percentage of cases (28%) showed combinations with NSCLC, with large cell carcinoma the most common, followed by adenocarcinoma and squamous cell carcinoma. Because of the frequency of a few scattered large cells in SCLC, we arbitrarily recommend that at least 10% of the tumor show large cell carcinoma before subclassification as combined SC/LC. However, combined SCLC is easily recognized if the additional component consists of other NSCLC subtypes such as adenocarcinoma or squamous cell carcinoma, so no percentage requirement is needed. Stage remained the only predictor of prognosis.

Bronchoscopy and needle biopsy techniques for diagnosis and staging of lung cancer

Lung cancer is the leading cause of cancer deaths in the United States. The individual therapeutic approach and prognosis depends on accurate diagnosis and staging. Flexible bronchoscopy (FB) and transthoracic needle biopsy (TNB) are the most widely used techniques for this purpose. This article provides a critical overview of indications, diagnostic yield, and limitations of bronchoscopy and TNB in the diagnosis of lung cancer.

The utility of transbronchial (Wang) fine needle aspiration in lung cancer diagnosis

We evaluated our experience with transbronchial fine needle aspiration (TBNA) in cancer diagnosis over a period of 1 year. A total of 51 aspirates were performed by specialist chest physicians in the presence of a cytopathologist who made on spot evaluation of Diff-Quik smears for adequacy and guided the aspirator for additional sampling if necessary. Two clusters of at least 10 malignant cells were required on the Diff-Quik smears to render an on the spot positive diagnosis of malignancy. Aspirates showing atypical cells or few malignant cells not fulfilling the above criteria were placed in a suspicious category and additional material was requested. The TBNA results were correlated with the transbronchial biopsy when available.

Fine-needle aspiration as the initial diagnostic modality in malignant lung disease

Cytologic detection of lung cancer is accepted, accurate, and time-honored. Typically, cytologic workup of a radiologic abnormality proceeds sequentially from sputum to bronchoalveolar cytology, and, if necessary, to fine-needle aspiration biopsy (FNA). Initial use of FNA in lung cancer diagnosis is controversial, but increasingly popular. We therefore decided to objectively assess current practice in cytologic lung cancer diagnosis at our institution. All pulmonary cytologic diagnoses for 1993 and the first half of 1994 were retrieved. Positive diagnoses were then used to access all patient data. Patients were stratified according to the specimen from which the first positive diagnosis was obtained. Of 542 pulmonary cytology specimens, 15% were sputa, 65% were bronchoalveolar, and 20% were FNAs. One hundred sixty-one of 172 malignant diagnoses were first diagnoses. Three percent of first malignant diagnoses were made from sputa, 47% were from lavages, and 50% were from FNAs. Although FNAs comprised just 20% of all pulmonary cytologies, 50% of all new malignant cytologic diagnoses were made by FNA. Initial use of FNA is successful, has a high diagnostic yield and low complication rate, and offers the most direct approach to diagnosis.

Prospective study of combined use of bronchial aspirates and biopsy specimens in diagnosis and typing of centrally located lung tumours

OBJECTIVE

To determine the diagnostic accuracy of examining bronchial secretions in pulmonary cytopathology and whether cytology and histopathology can complement each other in routine practice among lung specialists.

DESIGN

A prospective study comparing 1225 cytological and biopsy results, conducted during 1987-93. Tumours were confirmed by histopathology, imaging techniques, or clinical outcome and imaging techniques combined.

SETTING

11 lung or internal medicine units, France.

SUBJECTS

1128 patients (874 men; 254 women) aged 65.3 (SD 13.7) years who underwent fibreoptic bronchoscopy for various pulmonary symptoms.

RESULTS

Exact concordance between cytological and biopsy results was obtained in 1036/1187 (87.3%) satisfactory specimens. In all 574 lung tumours were diagnosed. One case (0.08%) was a false positive cytological diagnosis in a patient with tuberculosis. Patients with lung cancer were more likely to have positive cytological results than positive biopsy results (P < 0.001). Agreement in tumour typing was observed in 375/424 (88.4%) cases, when non-small cell carcinomas, small cell carcinomas and undifferentiated carcinomas were separated. In the 11 patients with squamous cell carcinomas in situ, eight (72.7%) of the carcinomas were diagnosed cytologically as squamous cell. Unsatisfactory material was obtained in only 20 (1.6%) and 19 (1.6%) cases by cytology and biopsy respectively. Examinations had to be repeated in 86 (7.6%) patients.

CONCLUSIONS

Examination of bronchial secretions complements histopathology in both diagnosing and typing lung tumours and could be performed more systematically in patients undergoing fibreoptic bronchoscopy.

Accuracy of bronchial aspiration cytology in typing operable (stage I-II) pulmonary carcinomas

The accuracy of bronchial aspiration cytology in typing resectable (stage I-II) lung cancer has been investigated in 100 cases, comparing preoperative cytologic features with pulmonary tumor histology seen at surgery. The accuracy has been 100% for small-cell carcinoma (two cases), 98.8% for squamous-cell carcinoma (86 cases), and 91.6% for adenocarcinoma (12 cases). The overall accuracy rate has been 98%. No case of undifferentiated large-cell carcinoma has been identified. It is suggested that the high accuracy in cytologic typing of operable lung cancer is basically related to adequate preservation of differentiation features, thus allowing for correct identification of most non-small-cell carcinoma. Moreover, the absence in this study of any large-cell carcinoma, compared with its frequency in advanced stage series, would indicate that such a histotype reflects excessive dedifferentiation of an original squamous or glandular form.