Systemic sclerosis is an autoimmune and systemic disease that mainly attacks the skin and other internal organs.
Against this background, a timely diagnosis plays an extremely important role when designing effective therapy and management plans. With the help of a artificial intelligence, this crucial task could be facilitated.
There is a close connection between artificial intelligence and medicine. This technology has made important contributions to the development of this discipline and other related scientific fields. However, in the special case of this rare disease, the lack of large data samples to train machine learning algorithms has made it difficult to develop specific solutions that can be used in settings with limited resources for this purpose, such as hospitals. .
Early diagnosis with low-cost technical deployment
A proposal recently put forward by eight researchers from the Department of Biomedical Engineering at the University of Houston, indicates, based on a preliminary study, that it is possible to set up an AI system that can contribute to the early diagnosis of systemic sclerosis, through a platform implemented with a standard laptop (2.5 GHz Intel Core i7 processor).
The material resources required to assemble this diagnostic instrument (technically, a common computer) imply a considerably lower investment than the general prices of equipment of this class for the clinical field.
With the presented neural network, which can be run without problems on equipment such as the one mentioned, it would be possible to immediately differentiate between images of healthy skin and skin with systemic sclerosis.
This contribution to the acceleration of the evaluation can considerably facilitate your care and therapy if the diagnosis is obtained at an early stage of the disease.
As is often the case with diseases of these characteristics, timely diagnoses are a challenge, even for the most experienced professionals. For this reason, the contribution of this advance, despite being in an early phase of research, opens a light of hope for the facilitation of this constant challenge for doctors in the area.
The technique used to give life to this system was deep learning, which consists of grouping several algorithms in layers, composing a neural network, which is capable of making its own decisions based on the training received, which in this case was a Intensive process of five hours of submission to the baseline data sample.
According to the research team at the University of Houston that is behind this study, after making various adjustments, the proposed network managed to achieve 100% precision in its work with a set of training images, 96.8% precision with another set of validation images and 95.2% with a third set of test images.
The report The details of this work were published in the IEEE Open Journal of Engineering in Medicine and Biology.