A tool that helps to detect one of the dangerous type of cancer melanoma, using Artificial Intelligence.
Archa Harikumar H
Luis Ruben Soenksen | credits: MIT Innovation Initiative
One of the dangerous type of cancer melanoma, a type of malignant tumour which is the main reason for more than 70 percent of all skin cancer-related deaths. Before this new tool was introduced, physicians have relied on a visual inspection to spot suspicious pigmented lesions (SPLs) which can be an indication of skin cancer. These early stage detection of SPLs in primary care settings can improve melanoma prognosis and can reduce the treatment cost.
The problem associated with SPL is that quickly identifying and prioritizing SPLs is difficult. Due to high volume of pigmented lesions which should be often evaluated for potential biopsies. Recently researchers from MIT and elsewhere have developed a latest artificial intelligence pipeline, using deep convolutional neural networks (DCNNs) and using them to analyse SPLs through the application of wide-field photography which is common in most smartphones and personal cameras.
DCNNs are neural networks that are used to classify (or “name”) images and then to group them (e.g.: when searching for photo). These machine learning algorithms is a subgroup of deep learning.
How does it work?
Using a smartphone cameras a wide-field photographs of large areas of patient’s bodies is captured. The captured image is then analysed using an automated systems. This automated system detects and analyses all pigmented skin lesions observable in the image. A predefined DCCN examines the pigmented lesions and categorizes them as yellow colour is considered for further inspection and red colour as a need to meet a dermatologist and displays the result in heatmap format. According to Luis R. Soenksen, a postdoc and a medical device expert currently serving as MIT’s first venture builder in AI and Health care said that the program uses DCNNs to quickly and effectively detect and screen for early stage melanoma. He along with MIT researchers, including IMES faculty members Martha J.Gray, W.Kieckhefer Professor of Health Sciences and Technology, professor of electrical engineering and computer science; and James J.Collins, Termeer Professor of Medical Engineering and Science and Biological Engineering was working on it.
Soenksen recently published a paper on this which explains about the development of system for the analysis of a SPLs using DCNNs in order to find skin spots lesions more quickly and efficiently which require more investigation, screenings which can be done during routine primary care visits, or even by the patients themselves. The system utilized DCNNs to enhance the detection and classification of SPLs in wide-field images.
The system was trained by the researchers using 20,388 wide field images from 133 patients at the Hospital Gregorio Marañón in Madrid and also the publicly available images using AI. The images were taken using variety of ordinary cameras readily available to consumers. The images of lesion were visually classified for comparison by the dermatologists working with the researchers. They found out that the system achieved more than 90.3 percent sensitivity in distinguishing SPLs from nonsuspicious lesions, skin, and complex backgrounds by avoiding the need for complex and time-consuming individual lesion imaging. In addition to that, it also includes a new method to extract intra-patient lesion saliency on the basis of DCNN feature from detected lesions.
Our research suggests that systems leveraging computer vision and deep neural networks, quantifying such common signs, and achieve comparable accuracy to expert dermatologists. “We hope our research revitalizes the desire to deliver more efficient dermatological screenings in primary care settings to drive adequate referrals.- Soenksen states
According to researchers, doing so would allow for more rapid and accurate assessments of SPLs and could lead to early treatment of melanoma.
Gray, who is a senior author of the paper explains about, how this important project got developed. The work originated as a new project developed by fellows (five of the co-authors) in the MIT Catalyst program which was a program designed to nucleate projects that solve pressing clinical needs. This work symbolize the vision of HST/IMEs devotee to leverage science to advance human health. This work was keep up by Abdul Latif Jameel Clinic for Machine Learning in Health and by the Consejería de Educación, Juventud y Deportes de la Comunidad de Madrid through the Madrid-MIT M+Visión Consortium.