Home > Oncology > WCLC 2019 > Screening, Detection, and Diagnosis > New technologies in lung cancer detection

New technologies in lung cancer detection

WCLC 2019
Emerging technologies for lung cancer detection are on the rise with many new techniques being explored, such as deep learning combined with radiomics. Radiomics represents the extraction of a large amount of quantitative computational features from medical images using data-characterisation algorithms. These features might reflect tumour characteristics that are not seen by the naked eye and can be of prognostic and predictive value. As the presence of a micropapillary or solid component is identified as an independent predictor of prognosis (indicating a more extensive resection), the accurate classification of subtypes still remains difficult in radiology, even when aided by classical radiomics. Wang et al. aimed to explore imaging phenotype using a novel method which combines radiomics with deep learning to predict high grade patterns within lung adenocarcinoma.

Deep learning has achieved great success in the fields of image analysis and computer vision, and may therefore be a useful approach in the field of lung adenocarcinoma. Five different methods were compared to classify the ground-glass opacities (GGOs) for the prediction of the pathological subtypes of high-grade lung adenocarcinomas, including classic machine learning, radiomics with selected features, radiomics + deep learning, and radiomics and deep learning separately. A total of 31 patients with high-grade patterns and 80 patients who lacked such patterns were analysed.

It was found that the methods combining radiomics with deep learning achieve the highest accuracy among all methods, with an overall accuracy of 0.888. This significantly outperformed classic machine learning, and radiomics and deep learning alone (P<0.001; see Figure) [1]. These results, based on a small image dataset, indicate that radiomics and deep learning need to be combined to effectively classify GGO. The exploration of tumour images may contribute to better treatment planning and personalised medicine.

Figure. Classification results of different methods [1]

Another possible new screening method is based on breath analysis by using volatile organic compounds (VOCs). There is a clear need for non-invasive diagnostic biomarkers which are able to identify patients at risk or with early stage cancer and VOCs patterns seem to be a potential tool aimed at early diagnosis and follow-up of these neoplasms. Canito et al. aimed to develop and validate a methodological approach to identify a VOCs breath pattern in order to discriminate between patients with lung cancer, malignant pleural mesothelioma, and healthy subjects [2]. The study population consisted of 28 lung cancer patients, 14 patients with pleural mesothelioma, 5 asbestos-exposed patients, and 20 healthy subjects.

The model was found to have good prediction ability. It resulted in good accuracy of 84% in lung cancer (sensitivity 86% and specificity 83%). Regarding malignant pleural mesothelioma, a negative predictive value of 82%, a positive predictive value of 91%, and a prediction ability of 93% were obtained. Despite the small study population and some technical limitations, these preliminary data support the VOCs analysis as a potentially useful tool in the earlier diagnosis of lung and pleural neoplasms. However, further investigation into breath analysis is warranted, and a prospective study evaluating VOCs in exhaled breath of patients with malignant pleural mesothelioma and asbestos-exposed subjects is currently ongoing.

  1. Wang X, et al. MA10.02. WCLC 2019.

  2. Catino A, et al. MA10.05 WCLC 2019.

© 2023 The Author(s). Published by Medicom Medical Publishers.
User license: Creative Commons Attribution – NonCommercial (CC BY-NC 4.0)

Posted on