Researchers from Australia and Bangladesh have developed a new artificial intelligence (AI) system designed to make skin disease diagnosis faster, more accurate and more transparent, with a particular focus or rare and underrepresented conditions that clinicians often struggle to identify.
“A breakthrough innovation”
The framework, named DermaViGNet, combines 2 machine learning techniques to analyse skin conditions while also explaining how it reaches its conclusions.
Adelaide University co-researcher, Dr Sabbir Ahmed, from the School of Computer Science and Information Technology, says this is a breakthrough innovation in this space
“Explainable AI plays a crucial role in clinical adoption to ensure model interpretability,” says Dr Ahmed.
“Many AI systems will share a recommendation or decision but cannot share the reasoning behind it, which can be an unacceptable risk in high-stakes fields such as health.
“Our technology incorporates explainable AI techniques by highlighting the specific regions of an image that influenced a diagnosis.
“This transparency can help clinicians better understand, verify and trust AI-generated results before incorporating them into patient care.”
Outperforming the benchmarks
DermaViGNet was reportedly trained and validated using nearly 10,000 dermoscopic images spanning common and rare conditions.
Measured against 8 existing deep learning models, it reportedly “outperformed all of them”, achieving “98% accuracy” in classifying 5 skin diseases:
- vitiligo
- acne
- nail psoriasis
- hyperpigmentation
- Stevens-Johnson Syndrome-Toxic Epidermal Necrolysis.
“A fast and reliable diagnostic aid”Â
“Our technology has the potential to support healthcare professionals by providing a fast and reliable diagnostic aid, particularly in settings where specialist dermatology expertise may not be readily available,” says Dr Ahmed.
“It may also help reduce diagnostic errors and improve outcomes through earlier detection and treatment of serious skin diseases.”
Adelaide University says, “The pioneering study addresses a longstanding challenge in medical AI: the underrepresentation of rare conditions in training datasets.”
“By incorporating a broader range of skin diseases, the model demonstrated strong performance across diverse conditions while maintaining high levels of accuracy and interpretability,” says Dr Ahmed.
“This is an important step towards more equitable and clinically useful AI-powered healthcare tools.”
Adelaide University says future research will focus on “expanding the dataset, integrating additional clinical information and further refining the technology for real-world clinical deployment”.
