Now Skin Cancers can be Detected Proactively Using AI Tool

0
55

Image Source: MIT

Before a disease can be treated, it must be diagnosed properly. For a long time, doctors have been using visual inspections for the identification of suspicious pigmented lesions (SPLs). Most probably, these lesions indicate skin cancer. Cancer detected at an early stage can substantially reduce the treatment cost and improve melanoma prognosis. Melanoma is a kind of malignant tumor. The tumor is quite dangerous as it makes up 70 percent of all the diseases related to skin cancer throughout the world.

The challenge here is that prioritizing and quickly finding SPLs is quite difficult. It is because sometimes the high volume of pigmented lesions required to be evaluated for potential biopsies. Researchers from different platforms including MIT have used deep convolutional neural networks (DCNNs) to develop a novel artificial intelligence pipeline. Through them, they plan to analyze SPLs using wide-field photography commonly found in most personal cameras and smartphones.

Neural networks also called DCNNs can name images after classifying them so that they can be clustered. These machine learning algorithms fall under the subset of deep learning.

According to Luis R. Soenksen, a medical device expert, the cameras are used to photograph large areas of the bodies of patients. The DCNNs are then used to effectively identify and screen them for early-stage melanoma. The research was conducted by Soenksen along with other MIT researchers faculty members of the MIT Institute for Medical Engineering and Science (IMES).

Image Source: Google

Soenksen elaborates on this by saying that “Early detection of SPLs can save lives; however, the current capacity of medical systems to provide comprehensive skin screenings at scale are still lacking.”

According to the paper published by Soenksen in Science Translational Medicine, an SPL analysis system that uses DCNNs can be used to identify skin lesions efficiently and quickly, the ones which require more in-depth investigation. Moreover, it would also help with the screenings that need to be done during primary care visits. The system uses DCNNs to optimize the classification and identification of SPLs in wide-field images.

The researchers trained the system using AI using 20,388 wide-field images. These images belonged to 133 patients from a hospital in Madrid in addition to the publicly available images. A variety of ordinary cameras were used to capture these images. Researchers worked alongside dermatologists to visually classify the lesions in images for comparisons. Results showed that sensitivity of more than 90.3 percent was achieved in differentiating SPLs from nonsuspicious skin, lesions and complex backgrounds. This allowed time-consuming and difficult individual lesion imaging to be avoided.

LEAVE A REPLY

Please enter your comment!
Please enter your name here