Lung cancer is a serious disease. According to the World Health Organization, lung cancer is one of the leading causes of death worldwide, accounting for an estimated 2.21 million deaths in 2020 alone. Most importantly, the disease can progress; that is, for many, it may begin as a few symptoms that do not cause alarm, before quickly turning into a life-threatening disease, leading to death. Fortunately, treatment options for patients with lung cancer have greatly improved over the past two decades. However, early detection of the cancer is the only way to significantly reduce mortality.
One notable event in this arena is the recent announcement by the Massachusetts Institute of Technology (MIT) and Mass General Hospital (MGH) regarding the development of a deep learning method called “Sybil” that can be used to predict the risk of lung cancer, based on use data. from just one CT scan. The study was published in the Journal of Clinical Oncology last week, and discusses how “tools that predict a person’s future risk of cancer may target potential benefits.” Therefore, the leaders of the study suggested that “a deep learning method for evaluating volumetric LDCT [Low Dose Contrast CT] data can be developed to predict individual risk without the need for additional demographic or clinical information.”
The model begins with a basic premise: “LDCT images have more predictive value for future lung cancer risk than what is currently known as lung tissue.” Therefore, the developers tried to “develop and validate a comprehensive study method that predicts the risk of lung cancer up to 6 years from a single LDCT examination, and evaluates its impact on disease.”
Overall, the research has been surprisingly successful, so far: Sybil can predict a patient’s lung cancer risk with some degree of accuracy, using data from just one LDCT.
Undoubtedly, the medical applications and effects of this technology are still in their infancy. Even the leaders of the research agree that a lot of work needs to be done to find out how to use this technology in real medical practice – especially when it comes to developing confidence in the technology, which doctors and patients can rely on. results of the system.
However, the core of the algorithm is still very powerful and incorporates game-changing features into the forecasting data.
CT scan, showing metastatic lung cancer. (Photo by BSIP/Universal Images Group via Getty … [+]
Diagnostic methods have never been more powerful. The fact that the tool can use a CT scan to show the activity of the disease for a long time can solve many problems – the most important thing that helps to receive treatment quickly and reduce the death rate.
Pundits, initially embarrassed, may push back against such systems, saying that no AI system can match the judgment and clinical skills enough to replace a human doctor. But the purpose of systems like this is not to replace the medical technology, but to supplement the physical process.
A system like Sybil can be used as a therapeutic tool, calling on a doctor, who can use their clinical knowledge to approve or disprove Sybil’s recommendations. This will not only increase clinical progress, but can also act as a second method of “diagnosis” and increase the accuracy of diagnosis.
Undoubtedly, there is still a lot of work to be done in this arena. Scientists, programmers, and inventors have a long journey ahead of them not only to perfect the actual algorithm and process itself, but also to navigate the hyper-nuanced arena of introducing this technology into real medical applications. However, the technology, purpose, and potential for improving patient care, if developed safely, fairly, and efficiently, are promising for the next generation of patients.
