Lung cancer is a devastating disease. According to the World Health Organization, lung cancer is one of the most common causes of death worldwide, accounting for nearly 2.21 million cases in 2020 alone. Importantly, the disease can be progressive; that is, for many, it can begin as mild symptoms that do not raise alarm, before rapidly evolving to a diagnosis that is life-threatening and leads to death. Fortunately, the range of therapies aimed at helping lung cancer patients has grown enormously in the last two decades. However, early detection of cancer remains one of the only means of significantly lowering mortality rates.

A notable achievement in this field is the recent announcement by the Massachusetts Institute of Technology (MIT) and Mass General Hospital (MGH) regarding the development of a deep learning model called “Sybil” that can be used to predict risk. of lung cancer, using data from a single CT scan. The study was formally published in the Journal of Clinical Oncology last week and discusses how “tools that provide personalized cancer risk assessment in the future could target approaches to those most likely to benefit.” Therefore, the study leaders postulated that “a deep learning model that evaluates all volumetric LDCT [Low Dose Contrast CT] the data could be constructed to predict individual risk without requiring additional demographic or clinical data.”

The model starts with a basic principle: “LDCT images contain information that predicts future lung cancer risk beyond currently identifiable features, such as lung nodules.” Therefore, the developers sought to “develop and validate a deep learning algorithm that predicts future lung cancer risk up to 6 years from a single LDCT scan and assess its potential clinical impact.”

Overall, the study has been remarkably successful so far: Sybil can predict a patient’s future lung cancer risk with some accuracy, using data from a single LDCT.

Undoubtedly, the clinical applications and implications of this technology are still immature. Even the study leaders agree that significant work will need to be done to figure out exactly how to apply this technology in actual clinical practice, specifically with regard to developing a degree of confidence in the technology, with which clinicians and Patients will feel safe trusting it. the outputs of the system.

However, the premise of the algorithm is still incredibly powerful and implies a potential game changer in the realm of predictive diagnostics.

Diagnostic measures have never been so powerful before. The fact that a tool can use just one CT scan to predict long-term disease function could solve many problems, the most important of which is enabling early treatment and decreased mortality.

Experts, initially blushing, may dismiss systems like these, pointing out that no AI system could match clinical judgment and prowess well enough to replace a human doctor. But the purpose of systems like these is not necessarily to replace physician expertise, but to potentially increase physician workflows.

A system like Sybil could easily be used as a recommendation tool, flagging potentially concerning CT scans to a physician, who could then use their own clinical judgment to agree or disagree with Sybil’s recommendation. This would likely not only improve clinical performance, but could also act as a secondary ‘verification’ process and possibly improve diagnostic accuracy.

Undoubtedly, there is still much work to be done in this field. Scientists, developers and innovators have a long journey ahead of them, not only to refine the algorithm and the system itself, but also to navigate the nuanced terrain of introducing this technology into real clinical applications. However, the technology, the intent, and the potential it holds with respect to improving patient care, if developed in a safe, ethical, and effective manner, certainly holds promise for the diagnostic generation to come.