Awaiting the support of artificial intelligence in lung cancer

Lung cancer (LC) remains the deadliest cancer (1). The only chance to cure the disease is its early detection before the onset of clinical symptoms. The disease is ‘silent’ until the advanced stages when an effective cure is unavailable.

Currently, a broad discussion is ongoing in Europe regarding the implementation of LC screening. It has been proved that people from the high-risk group can benefit from this intervention (2,3). However, small nodules in the lungs detected on low-dose computed tomography (LDCT) scans need a computed tomography (CT) follow-up. They do not have the typical features of malignancy, such as large tumors, and it is impossible to unequivocally determine, without a control examination, whether they are benign or malignant.

Until now, in Europe, only Croatia implemented LC screening and Poland is during the 3^rd round of the Pilot National Lung Cancer Screening Program and the ongoing debate between the political and medical environment with the hope for soon implementation.

But how to do it, when the healthcare sector is facing many problems? Among them, one of the most important, is the small number of specialists, including radiologists and thoracic surgeons. The system is heavily loaded and cannot keep up with patient service, moreover, the number of imaging studies is growing exponentially.

We have entered the era of artificial intelligence (AI) use in healthcare and the clinical potential of big data-based solutions is a fact and holds the promise to beat this disease. A very promising AI field is radiomics—the process of converting radiological images into quantitative data. Radiomics can be used to obtain not only morphological and structural data of the analyzed lesion but also their biology. But how to discover and describe all this information with numbers? How to find information about cell division in the CT scan density? To obtain the working material the workflow must include besides a huge amount of data also image acquisition, preprocessing, region of interest (ROI) segmentation, features extraction from ROI, study, and machine learning (ML).

Extracted features using ML disclose patterns that mirror cancer biology. Created radiomic model enables visual assessment and supports clinical decision-making based on selected radiomic signature (RS). Moreover, clinical data can be incorporated as an additional value into prognostic and predictive models which are potentially powerful and, most importantly, non-invasive prognostic tools (4,5).

AI application in LC detection and management could strengthen the LC screening potential. A considerable number of studies using the collections of big data sets are being conducted worldwide to identify predictive biomarkers of nodule malignancy to support clinical decision-making and predict the disease’s further course. Radiomics could be used also to assess the treatment response and recurrence probability.

The use of radiomics in real-world medicine is very desirable, however, its limitations should be kept in mind. First of all, the analyzed group of images/patients has to be big enough to assure the accuracy of the results. A lot of attention should be put on repeatability and reproducibility which are at least scanner and observer-dependent.

In a recent study published on Translational Lung Cancer Research, Volpe et al. present us the “Impact of image filtering and assessment of volume-confounding effects on CT radiomic features and derived survival models in non-small cell lung cancer”. The paper is an explorative modeling study with particular importance suggesting mindfulness in using radiomics to evaluate small (volume) lung nodules (6).

The major goal of this work relates to tumor volume which is a recognized prognosticator. The impact of image filtering on the volume dependence and prognostic value of radiomic features has been assessed with the use of the non-small cell lung cancer (NSCLC) open-source radiomic database and tools (pyradiomics and python model implementation). The use of a publicly available database and tools can be considered advantageous as it is a proof-of-concept study (6).

Quantification of the feature/volume dependency across multiple preprocessing methodologies and volume groups (high vs. low) has been performed as well as an assessment of the impact of these variations on survival model performance has been studied (6).

In order to assess the impact of volume on survival prediction, two models, one without and the second including volume assessment have been created for each preprocessing method. Moreover, the trained baseline model (only clinical variables and tumor volume) has been created. This article generates a hypothesis and the authors encourage further analysis of preprocessing and feature-volume correlation on the public data sets and further external validation. Survival models were slightly affected by the imaging preprocessing filters (6).

The authors built radiomic models which outperformed ones built exclusively on the clinical variables and volume assessment with the largest improvement in the high-volume group. The presence or absence of volume measurements did not generally alter their performance. In early LC, radiomic features may be of less utility (6).

Lancaster et al., have shown that AI as an impartial reader in baseline screening can significantly reduce a radiologist’s workload by 86.7% whilst noncompromising on false-negative results of ultra-LDCT and volume-based management of nodules in LC screening (7). In principle, prognostic models are developed to aid in the most difficult clinical cases, which include small, discrete nodules. However, it could be postulated that AI including radiomics could be a gateway to worldwide nodule assessment standardization, LC screening unification, and LC management (7).

Volpe et al.’s article and other recent papers (8,9) indicate a great need and potential to introduce AI-based support solutions in the evaluation of medical images in everyday clinical practice (6).

The American Cancer Society currently is funding 70 grants on lung cancer research worth $28 million. Ongoing research includes development of the diagnostic tools for the early detection of LC and the discovery of precision medicine to provide personalized targeted treatment (10).

The situation is hampered by the rapidly growing number of tests, the post-coronavirus disease (post-COVID) health debt, and the aging of the population (6). In addition, difficulties are caused by the growing legal responsibility of physicians and the progressive bureaucratization of the health care system. This indicates that without AI support, the waiting time for test results and diagnosis will start to pose a threat of delaying treatment. We are currently observing changes in the everyday world, including rapid technological development, and in a society that has begun to accept the use of these solutions in everyday life.

However, it should be remembered that without the creation of a legal framework, it will not be possible to implement AI-based solutions. Great emphasis should also be placed on the creation of publicly available extensive databases so that research and validation of new systems have a reliable final result.

All stakeholders hope that the widespread adoption of AI applications will allow LC to be detected earlier with greater specificity, LC patients can be treated faster and more accurately, and many lives will be prolonged and/or saved.

Acknowledgments

Funding: None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Translational Lung Cancer Research. The article did not undergo external peer review.

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-23-76/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. International Agency for Research on Cancer. GLOBOCAN Lung Cancer Facts Sheet 2020. Available online: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwjbrcSFyc79AhWH-yoKHftCCD8QFnoECA4QAQ&url=https%3A%2F%2Fgco.iarc.fr%2Ftoday%2Fdata%2Ffactsheets%2Fcancers%2F15-Lung-fact-sheet.pdf&usg=AOvVaw1QY7aoyIfFa7BQo-E9g_uS
2. National Lung Screening Trial Research Team. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med 2011;365:395-409. [Crossref] [PubMed]
3. de Koning HJ, van der Aalst CM, de Jong PA, et al. Reduced Lung-Cancer Mortality with Volume CT Screening in a Randomized Trial. N Engl J Med 2020;382:503-13. [Crossref] [PubMed]
4. Kang CY, Duarte SE, Kim HS, et al. Artificial Intelligence-based Radiomics in the Era of Immuno-oncology. Oncologist 2022;27:e471-83. [Crossref] [PubMed]
5. Wu G, Jochems A, Refaee T, et al. Structural and functional radiomics for lung cancer. Eur J Nucl Med Mol Imaging 2021;48:3961-74. [Crossref] [PubMed]
6. Volpe S, Isaksson LJ, Zaffaroni M, et al. Impact of image filtering and assessment of volume-confounding effects on CT radiomic features and derived survival models in non-small cell lung cancer. Transl Lung Cancer Res 2022;11:2452-63. [Crossref] [PubMed]
7. Lancaster HL, Zheng S, Aleshina OO, et al. Outstanding negative prediction performance of solid pulmonary nodule volume AI for ultra-LDCT baseline lung cancer screening risk stratification. Lung Cancer 2022; Epub ahead of print. [Crossref] [PubMed]
8. Raghu VK, Walia AS, Zinzuwadia AN, et al. Validation of a Deep Learning-Based Model to Predict Lung Cancer Risk Using Chest Radiographs and Electronic Medical Record Data. JAMA Netw Open 2022;5:e2248793. [Crossref] [PubMed]
9. Mikhael PG, Wohlwend J, Yala A, et al. Sybil: A Validated Deep Learning Model to Predict Future Lung Cancer Risk From a Single Low-Dose Chest Computed Tomography. J Clin Oncol 2023; Epub ahead of print. [Crossref] [PubMed]
10. The American Cancer Society. Cancer Facts & Figures 2023. Special Section: Lung Cancer. Available online: https://www.cancer.org/content/dam/cancer-org/research/cancer-facts-and-statistics/annual-cancer-facts-and-figures/2023/2023-cff-special-section-lung-cancer.pdf (accessed 31, January, 2023).