Researchers at the Institute of Computing at the State University of Campinas (Unicamp) developed a computer program capable of diagnosing skin tumors, such as melanoma, for example, with 86% precision through artificial neural networks, which mimic the body's neurons human.
The technique used reads a person's skin and sends information to the computer that identifies whether the lesion found is benign or malignant, this being possible due to the programming carried out by the researchers, which makes the machine have access to a database. images of different types of skin lesions. In the future, these researchers hope that this program will be available in hospitals and clinics helping to speed up the diagnosis of skin cancer.
How the technique is applied
The skin tumor diagnosis technique that is being developed by Unicamp researchers uses artificial intelligence through a process called deep learning, which consists of teaching a machine a certain function through artificial neural networks, which are computer programs that mimic the neurons in the human body.
These artificial neural networks of the computer are able to analyze the skin, taking into account images from a bank with approximately 23, 906 photographs, of different types of skin lesions, which means that the computer compares the appearance of the person's skin with the images of this bank, revealing whether the lesion is benign or malignant.
The study showed that even if the skin site is not fully scanned, the computer is still able to detect the tumor, making an average of 71% of diagnoses, giving the perspective of early identification of some types of skin cancer, and the treatment faster.
What are the next steps
The Unicamp study is funded by Google, through an initiative called the Google Latin America Research Awards (Lara) and the next steps planned by the researchers include increasing the number of images in the photo bank, to improve the accuracy of computer analysis.
The researchers hope that the computer program will be available for installation on cell phones in the future and that through a specific lens attached, it will be possible to diagnose skin cancer.
In addition, the hope of scholars is that in the future the program can be used in hospitals and health centers where they do not have dermatologists and that it can be used to assist and support doctors when diagnosing and deciding on treatment.
What is currently used
The diagnosis of the main type of skin cancer, melanoma, is made by an international method known as ABCDE, in which the letters correspond to important aspects of the signal on the skin, such as:
- Asymmetry: half the signal is different from the other; Irregular edges: outline of the signal is poorly defined; Variable color: presence of several colors in the same lesion (black, brown, white, reddish or blue); Diameter: sign on the skin greater than 6 millimeters; Evolution: changes in size, color and shape of the sign.
Melanoma can appear on any part of the body, be it on the skin or mucous membranes, in the form of a mole, sign or stain, and in people with black skin it is common to appear these lesions on the palms and soles of the feet. Early diagnosis of melanoma allows better treatment results, so it is important to always observe changes in the skin. Here's how to identify signs of skin cancer.
