AI technology has achieved a breakthrough in the detection of melanoma, the most deadly type of skin cancer, with 100 percent accuracy. Scientists have spent three years fine-tuning the software to improve its early detection capabilities. The AI system was able to identify all 59 cases of melanoma, which often appears as an irregularly shaped mole with different colors. It also detected 189 out of 190 total skin cancer cases, including non-melanoma cancers. This development has significant implications for diagnosing and treating skin cancer, as early detection greatly increases the chances of successful treatment.
Melanoma is responsible for more than 57,000 deaths worldwide in 2020 alone, making it the most fatal type of skin cancer. However, non-melanoma cancers such as squamous cell carcinoma and basal cell carcinoma, although less deadly, are more common and account for more deaths overall.
Artificial intelligence has faced controversy in healthcare due to privacy and ethics concerns. However, medical researchers are increasingly embracing AI as a tool to streamline administrative duties and expedite diagnosis, ultimately giving patients a better chance of survival.
In this study, dermatologists from the UK trained the AI system using patient data from consultations and images of their cancers. They tested the AI’s ability to detect cancers and pre-cancerous growths using more than 22,350 participants over two and a half years. The AI software accurately diagnosed all 59 cases of melanoma and 99.5 percent of all skin cancers, demonstrating significant improvements compared to previous versions.
Dr. Kashini Andrew, the lead author of the study, highlighted the rapid improvement and learning capability of AI. He emphasized the potential of AI technology to free up more time for patients requiring urgent attention.
Dermatologists follow the ABCDEs when examining the skin for cancerous growths: asymmetry, border, color, diameter, and evolving. Most melanomas appear as asymmetrically shaped moles with uneven edges. Other signs of melanoma include irregular borders, multiple shades of color, and size larger than a pea or a pencil eraser. Melanoma grows in two phases, horizontal and vertical, with the latter being more dangerous and invasive.
Non-melanoma skin cancers have different manifestations, such as waxy or red bumps, patches of red and scaly skin, or small pale growths. Despite the accuracy achieved by the AI software, the researchers cautioned that AI should not replace the expertise of dermatologists and can only be used as an additional tool in the diagnosis of skin cancer.
The researchers presented their findings at the European Academy of Dermatology and Venereology Congress of 2023 in Berlin, highlighting the promising future of AI in healthcare provision. With further advancements and research, AI technology has the potential to revolutionize the early detection and treatment of skin cancer, ultimately saving lives.