Speed and accuracy: thanks to Artificial Intelligence, CT scans will soon become a safer exam. On the Journal of Medical Imagingindeed, research has just been published which shows how, thanks to an algorithm, it will be possible to perfect the levels of radiation to be administered to patients. A group of researchers, medical physicists and radiologists from the Department of Physics and Astronomy of the University of Florence, the Careggi University Hospital and the Local Health Authority Toscana Centro are working on the project, led by Sandra Doria of the Institute of Chemistry of Organometallic Compounds of the National Research Council of Florence (Cnr-Iccom). The Higher Institute of Health and the Bruno Kessler Foundation of Trento also collaborated on the project, using the computational resources made available by Uniser Pistoia.
Metaverse: the new frontier of medicine and health services
by Irma D’Aria
The ‘power’ of the Tac
Computed tomography is one of the most powerful and consolidated diagnostic tools among those available to modern medicine. However, the manual analysis of the images that are produced through this methodology requires a lot of time and their quality is directly proportional to the amount of X-ray radiation to which a patient must be subjected for the purpose.
Digital health: public and private together for a more efficient health system close to the citizen
by Irma D’Aria
The creation of the algorithm
Researchers have succeeded in automating the process of assessing image quality in computed tomography (CT) exams by using artificial intelligence to reduce radiation to the patient. “Our group – he explains Sandra Doria (Cnr-Iccom), coordinator of the research – has created an algorithm, analyzing the data generated by the visual examination that several radiologists have carried out on CT images of a phantom, created for the purpose of replicating the characteristics of human tissues and the presence of artificial injuries. Subsequently, two artificial intelligence models were developed which were trained and tested through the use of previously collected images and responses from doctors”.
My doctor is an algorithm
by Paola Mariano
Reduce the amount of radiation
These models could represent an automatic assessment strategy of the quality of a CT image, which will allow to optimize the radiation dosage, in order not to expose patients to an excessive amount of X-rays. “During treatments or diagnostic procedures, a patient must be exposed to minimal levels of radiation, according to the known principle ‘as low as reasonably achievable’ (ALARA). With this in mind, medical personnel must find a compromise between exposure to X-rays and obtaining good quality images, also to avoid erroneous diagnoses”, continues Doria.
The comparison between AI and doctors
The results obtained through this study are very promising: “Our models – concludes Doria – can accurately identify an object inserted in the phantom, as a radiologist would be able to do. We hope, in the near future, to be able to apply these models on a larger scale and make assessments even faster and safer, greatly simplifying the process of optimizing the radiation dose used in CT protocols. This aspect is essential to reduce the risks to the patient’s health and to optimize the timing of medical evaluations”.
Other searches in progress
It is certainly not the first time that Artificial Intelligence has been used in the medical field and just as regards the CT scan, some researchers at the University of Massachusetts have recently developed a system capable of predicting the risk of lung cancer with a 94% within one year and 81% within six years. Sybil’s work is to digitally analyze a computed tomography and from this she can predict what the risks of lung cancer are. “Sybil provides a risk score, not a diagnosis, so it’s very helpful to identify which patients need to be followed up closely or screened for cancer,” he explains. Anniversary of Sequistprofessor of medicine at Harvard Medical School, one of the authors of the study, published in the Journal of Clinical Oncology.
The development of the artificial intelligence was based on the data of 15,000 participants for a total of 35,001 CT scans used to develop the model and another 6,282 to test it, noting how many people had actually shown suspicious lesions in the CT scans following the first one and comparing them with the results shown by Sybil. The results proposed by Artificial Intelligence proved to be correct in 94% of cases after one year, 86% after two years and 75%.