A new AI model that is capable of simultaneous analysis of various skin images is possible, may possibly improve the diagnostic accuracy of both dermatology and non-dermatology professionals.
An international research team led by AI and Machine Learning experts at Monash University developed a multimodal foundation model, called Panderm, Designed as a tool for clinical decision support in dermatology. It can process multiple skin images at the same time and offer diagnostic probability assessments.
The AI model was trained on more than two million skin images of four types: close-up photos, dermoscopic images, pathologies and total body photos, which came from 11 institutions in different countries.
It was trained to perform a wide range of clinical tasks, including total detection of skin cancer and risk assessment, repetition of cancer and metastasis prediction, skin type assessment, mole counting, lesion change, differential diagnosis of various skin disorders and lesional segmentation.
The Panderm -Team involved researchers and doctors of Alfred Health, the University of Queensland, Princess Alexandra Hospital in Brisbane, Royal Prince Alfred Hospital, NSW Health Pathology, the University of Florence in Italy, Medical University of Singapore and Singitara Ai -Singitara Ai -Singitary Ai a Singapore, Singapore Ai Ai -Singitary Ai Aanitititi Ai -Singitary Ai -Singary, Singarian Ai Aanitary, Singalitary Ai -Singitary, Singalitary Ai Aanitary, Singarian Ai Aanitary Ai Aanitary, Singaly Ai Aanitary, Singaly Ai -Medical and Singaly Ai -Medical and Singarian Ai Rijk and Singarian Ai Rionyytary, Singarian and Singaly Ai R ampage and Singarian Ai Rionyyyyya kore Spain.
Findings
The researchers carried out diagnostic performance and three readers’ studies for their model, the findings of which have been published in Nature Medicine.
One of the remarkable findings was that the model then surpassed clinicians when detecting melanoma at an early stage, the most aggressive type of skin cancer, by 10%.
It was also determined that the AI helped to increase the accuracy of dermatologists in diagnosing skin cancer from di -moscopic images with 11% points to 80%.
Another important finding was that Panderm improved the ability of non-dermatologists to identify and differentiate skin conditions, such as inflammatory dermatoses and pigmental disorders, based on photos with 16.5%. These include generalists who carry out routine initial skin reviews: general practitioners, practitioners of general practitioners and assistants for nurses and clinical test.
Interestingly, the model also surpasses existing models (such as Swaverm, SL-Imagenet and Dinov2) when performing different clinical tasks with regard to the assessment of skin cancer and other skin conditions, even when they are trained with only 10% of the labeled data. Tasks include risk ratification of lesions, phenotype assessment, detection of lesion changes and malignancy, multi-class cancer diagnos, lesion segmentation and metastasis forecast and prognosis.
“Given the limited specialist access in first-line care institutions where most skin conditions are initially evaluated, these findings indicate that Panderm is potential to tackle dermatological expertise in health care in health care, both the technical possibilities and clinical applications. More importantly, in both human collaboration studies and in both the Diagosis Studies of the Diagosis Studies and the of the Diagosis Studies and the Diagnosis Studies and the Diagnosis and the Diagnosis Studies and the Diagosis Studies and Diagnosis Even the diagnosis of human diagnosis, “said authors.
“This phenomenon probably stems from the selective integration of clinicians of AI recommendations instead of blind therapy compliance, which represents a balanced clinical implementation in which practitioners retain their diagnostic autonomy, while still benefiting from AI support,” they explained.
Why it matters
Assessing skin conditions in clinical practice includes numerous tasks – from risk assessments and image analysis to monitoring lesions and predicting results. Although AI-driven clinical support tools are available on a large scale for dermatology, they are limited to individual, insulated tasks.
“The absence of integrated AI solutions that can support these different workflows is currently hindering the practical impact of AI in dermatology,” the researchers said.
“Earlier AI models have difficulty integrating and processing different data types and imaging methods, reducing their usefulness for doctors in different real-world institutions,” the university teacher Zongyuan Ge, one of the main co-authors of the study, was quoted in a media release.
According to H. Peter Soyer, another main co-author, the study of the Panderm team has unveiled the potential of a new multimodal foundation model to support skin disease care in low means institutions.
“The Strength of Pandermen Lies in its ability to support existing clinical workflows. It could be particularly valuable in busy or resource-limited settings, or in primary care where access to dermatologists may be limited. Small Amount of Labelled Data, A Key Advantage in various Medical Settings Where Standard Annotated Data is of Limited, “Said the Professor and the Director of the Dermatology Research Center at the University of Queensland.
Moreover, Panderm can be indispensable in the early detection of the deadly and invasive melanoma. “This kind of help could support a previous diagnosis and more consistent monitoring for patients who are at risk of melanoma,” said Victoria Mar, one of the main authors of the study and a professor and director of the Victorian Melanoma service at Alfred Health.
The research team is planning to carry out more clinical evaluations of their Dermatology Foundation model with a focus on guaranteeing fair performance in various patient populations and health care institutions.
The larger trend
It is also in Australia where the The world’s first AI-driven pop-up Skin Care Kliniek was set up to detect skin cancer such as melanoma early. About two of the three Australians would be established a form of skin cancer during their lives. Based on government statistics, around 400,000 cases are reported every year. The nurse pop-up clinic by Health Charity Skin Check Champions wants to achieve that number and raise the skin screening by a quarter with the support of AI.
Outside of Australia, South Korea has recently approved its first local AI-driven smartphone application for diagnosis of skin cancer. The Canofymd Scai from Lifesemantics received regulatory approval in June last year from the Ministry of Food and Drug Safety.
#improve #capacity #nonderma #diagnose #skin #conditions