In the last decade, artificial intelligence has been integrated into our lives in unique ways, from the personal assistant used by smartphones and programs that create unconventional art for collectors, to the facial recognition technology used by security at some airports. It is therefore not surprising that artificial intelligence has come to be incorporated in the field of health, where the possibilities for the development of this technology are vast. With around 10 million people dying from cancer each year in the world, and the survival rate of cancer patients can be greatly increased by early diagnosis of the disease, using artificial intelligence to save as many patients as possible brings a real gain. humanity.
What exactly are the ways in which artificial intelligence can help diagnose cancer in its early stages, in the case of imaging tests?
- With the help of artificial intelligence, even the smallest tumors can be identified;
- Through artificial intelligence, CTs can be analyzed directly in 3D format;
- Artificial intelligence can be used to perform routine interpretations, allowing radiologists to focus only on difficult cases that require more attention.
Artificial intelligence and small tumors
Artificial intelligence can help identify small tumors (such as those under 10mm) that can be missed by the human eye. This is possible due to the way the technology analyzes the imaging data.
First of all, through artificial intelligence it is possible to examine a pixel-by-pixel image, an analysis that is difficult to perform at this level just by visualizing the image itself.
Second, the perspective from which artificial intelligence interprets an image (a slide from an MRI / CT) is different from the human one. The artificial intelligence algorithm does not have to be “learned” to recognize the characteristics by which a tumor can be identified, but is left to discover its own set of parameters that indicate the presence of a tumor. These parameters are obtained after analyzing a large number of images in which the tumor has already been identified. Because of these characteristics, details that seem insignificant at first may actually be part of a pattern of cancer identification before it becomes visible to the human eye. Therefore, the patient receives a temporal advantage in diagnosing the disease.
Direct analysis in 3D format of a CT
Another aspect by which the analysis of the imaging examination with the help of artificial intelligence can make the identification of a tumor easier to achieve is the format in which the information provided by a CT is interpreted. In the case of direct analysis by a radiologist, the data of a CT are analyzed in turn in two-dimensional slides. In order to have a three-dimensional perspective, the doctor has to mentally assemble the 2D images to obtain an image in space, in the form of a three-dimensional “puzzle”.
The artificial intelligence algorithm interprets the information of a direct CT in 3D format. Thus, the most accurate volumetric dimensions of the tumor are taken into account in its analysis. In addition, new patterns of tumor identification can be discovered through artificial intelligence, this time depending on the characteristics of the volume.
Performing routine interpretations
One way to propose early diagnosis of cancer in several countries is through screening programs aimed at people at risk for certain types of cancer. Such programs have been shown to significantly improve the chances of survival of those diagnosed. A study by Erasmus University Medical Center in Rotterdam showed that a 10-year lung cancer screening program managed to reduce patient mortality by 25%.
Unfortunately, these efforts to detect cancer early come at a high cost of resources, both material and human. With the help of artificial intelligence, the high volume of work required by such screening programs can be shared between the algorithm and radiologists.
This technology can be used to perform an initial triage of patients, separating them into two categories: cases that do not present a high risk and cases that present. In this situation, the algorithm can be used to analyze routine cases, leaving only the more complex cases to a specialized examination.
This collaboration saves both human and financial resources and time. So it becomes feasible for the program to be able to cover as many patients as possible in a very short time. This increases the chances of early diagnosis and cure of cancer.
Successes in the field of artificial intelligence
The fact that artificial intelligence can bring significant benefits to the field of imaging has been confirmed in several situations.
In the US, a team of researchers at Northwestern University in Chicago has created an artificial intelligence algorithm to detect lung cancer. With its help, they were able to identify in CTs nodules that indicated the early stages of cancer, with an accuracy of 30%.
Researchers at University College London, in collaboration with London’s Institute for Cancer Research, have also created an algorithm for lung cancer detection. In this case, they focused their work on detecting the first signs of remission of lung cancer after initial treatment.
Also, a team of researchers from Imperial College London focused on the detection of breast cancer. Their algorithm improved the case analysis, decreasing the rate of false-positive diagnoses by 1.2% and the rate of false-negative diagnoses by 2.7%.
The results of the research teams are promising, as they validate how artificial intelligence can contribute to the early diagnosis of cancer.
Collaboration between doctors and artificial intelligence
Artificial intelligence programs can provide support in diagnosing cancer, but they cannot replace radiology specialists. They are designed to be used as tools that can make doctors’ work easier and more accurate.
Computers are excellent at local tasks, focused on a single topic. Instead, people are best suited for global tasks, such as analyzing all information (imaging tests, blood tests, family history) from a broad perspective, in order to provide a definitive diagnosis.
In addition, radiologists are able to learn much more quickly the characteristics of a rare type of cancer, making them the most able to interpret imaging cases of high difficulty. Artificial intelligence can instead help doctors by diagnosing routine cancers.
Therefore, a collaboration between radiologists and artificial intelligence can only bring benefits, both for doctors (diagnostic assistance, releasing too much work) and for patients (early diagnosis of cancer, due to the increased possibility of screening programs and the ability to identify subtle signs of cancer).
Article written by Teodora Aslău