New discoveries based on artificial intelligence can help primary medical doctors find the lesions that may become fatal melanoma, help them diagnose, and refer patients to experts as soon as possible. The discovery involved the Spanish.
Facts have proved that artificial neural network systems based on artificial intelligence and machine learning in computers can effectively analyze photos and classify and distinguish images. Skin lesions Can be a Fatal melanoma If you do not find them in time and eliminate them. They are called “Ugly ducklingBecause of its rarity and difference from other spots or freckles on the skin.
A new system developed by an international team of scientists can filter out these markers from photos that can even be taken with a mobile phone.Conclusion of this work involving Spanish dermatologists JoséAntonio Avilés-Izquierdo, Members of the Gregorio Maranón Hospital in Madrid published today in the journal Science Translational Medicine.
Scientists from various countries and research centers participated in the study, including MIT or Harvard University, although all the photos used for “teaching computers” to distinguish carcinogenic lesions suspected of benign lesions correspond to patients. One of the hospitals treated by the dermatology service of Gregorio Marañón Hospital Main world reference In the treatment of melanoma.
The team trained their technique 38,283 photosIncluding skin 133 patients, And observed that this method distinguishes suspicious lesions from non-suspicious lesions, with a sensitivity and specificity of 90.3% and 89.9%, respectively.
In another experiment, the strategy also classified the “ugly duckling” lesions on the skin of 68 patients, and the classification generally matched the evaluations of various dermatologists.
Avilés-Izquierdo recalled that this was a research project, Since 2014, And has been designated to basically create a tool based on an algorithm that allows a computer to learn to distinguish skin lesions.
The results show that the platform can help clinicians find suspicious lesions during clinic visits faster and on a larger scale, which is helpful for early diagnosis and treatment.
In a statement to EFE, the Spanish doctor emphasized that the tool will be very useful for the following services: Primary health care Preliminarily screen patients who are susceptible to melanoma and determine that they should be referred to expert.
The most common skin cancer
Melanoma is the deadliest form of skin cancer, but its prospects may be very good for patients whose lesions are still limited in the early stages and have not spread deep into the skin.
To detect them, doctors usually assess large areas of skin based on a set of commonly known standards. “ABCDE”; Acronyms corresponding to the following: edge (usually irregular); color (usually varying and inconsistent); diameter (usually greater than 6 mm); and evolution (alternation of moles or spots).
Scientists emphasize that in most countries, health authorities have begun to implement large-scale skin cancer screening programs to reduce the impact and damage caused by such cancers, and many clinics Lack of necessary tools To assess the lesions of a large number of patients.
reference:
Soenksen R, Luis et al. “Using deep learning to perform dermatologist-level detection of suspicious pigmented skin lesions on dermatologists from wide-field images.” Science Translational Medicine (2021)
.