The breast cancer It is the second most diagnosed cancer in Spain, only behind that of the colon. With data from the latest study by the Spanish Society of Medical Oncology (SEOM), 34,750 women will be diagnosed with the disease in 2022. This is where radiologists appreciate the support that the artificial intelligence as a support to screening of this type of cancer. It would allow them to detect more cancers, would be diagnosed at an early stage, would mean a better prognosis for the patient and the use of less aggressive treatmentsas explained a The newspaper of Spain Dr Esperanza Elías, of the Reina Sofía University Hospital in Córdoba.
Dr. Elías is a specialist in artificial intelligence (AI) applied to breast screening from the Spanish Society of Medical Radiology (SERAM). He details that the development of new AI systems with “deep learning” technology, an unsupervised machine learning approach (that is, training data is required, but does not need to be labeled) that rely on the functioning of the neurological system. These systems, he explains, have improved the algorithms of ‘computer-assisted diagnosis’ (CAD), i.e. computer-assisted diagnoses that assist doctors in the interpretation of multimedia contents that are obtained in the tests to which the patient has been subjected.
The new systems can detect suspicious breast cancer lesions by assigning them a score based on the likelihood of malignancy
Unlike traditional CAD, adds the radiologist, the new systems are capable of detecting suspicious lesions breast cancer in both digital mammography and tomosynthesis – an advanced form of digital mammography to increase early diagnosis – by scoring tests based on the likelihood of malignancy. Besides, they can be used as a support to the radiologist’s readings, facilitating the task, achieving a decrease in delay, increasing cancer detection and reducing false positives and negatives.
Artificial intelligence is also capable of doing this classify mammograms based on the likelihood of malignancy. In the short term, the doctor abounds, will play a very important role in mammograms. Of course, he specifies that more prospective studies conducted in real-world clinical settings are needed. Some, he points out, show that it could be decrease the workload in screening programs up to 70% without reducing sensitivity.
The system, indicates the radiologist, marks the lesions suspected of malignancy, assigns them a score from 1 to 100 which is based on the probability of malignancy and then, depending on the lesion, with the highest score, classifies the complete study into three categories : low, medium or high risk. “High risk accounts for 3% of all studies or women, and 1 in 4 or 5 women of this percentage (or classified as high risk) have been found to have breast cancer,” she says.
The data managed by radiologists refer to the female population of screening age, from 50 to 69 years
The physician abounds that, at low risk, AI “ranks about 70% of studies or women, which is why we say that workload can be avoided by 70%. While intermediate risk assumes a probability of 1 in about 125 women and represents 27% “. Clarifies that the data refer to the female population of screening age, from 50 to 69 years.
The future: radiogenomics
In addition, the specialist guarantees that the AI increases the positive predictive value of patients referred, in particular in the classified by the system as high risk. “It would allow the generalization of the use of tomosynthesis in screening programs, where an increase in breast cancer detection has also been shown, as it would reduce the increase in workload that the tomosynthesis readings assume, because the time taken by the radiologist is more than double that of a digital mammogram, “he points out
The new radiogenomic techniques, still under study, are moving towards the so-called precision medicine
Dr. Esperanza Elías also alludes to radiogenomics, a new artificial intelligence technique that, applied to medical images (breast resonance, digital mammography, ultrasound …) studies the relationship between the phenotypes of the image (the characteristics of the image presenting cancer) and the genome of the tumor. New techniques, she concludes, which are heading towards the so-called medicinal precision: a complex study, with computer and mathematical models, which evaluates aspects such as the interaction of genes, metabolites, proteins and other biological components of each patient.
The specialist alludes to a way of focusing treatment and prevention on groups of individuals, depending on their disease, genetics, environmental factors and lifestyles. It would allow to avoid the execution of biopsies and therefore to avoid the associated complications. Yes, we will have to wait. At the moment, the radiologist concludes, radiogenomics is used in research in Spain, particularly in breast resonance and mammography.