An Australian breakthrough in AI-supported immunofluorescence imaging is leading to more accurate predictions of how a melanoma patient may respond to immunotherapy, increasing the opportunity for safer, more effective care.
This new research, led by Dr Priyanka Rana, PhD, and Dr Sidong Liu, PhD, from the Australian Institute of Health Innovation, Macquarie University, is published in the IEEE Journal of Biomedical and Health Informatics and developed in conjunction with Melanoma Institute Australia.
The research shows the new AI-supported model delivers a 20% improvement over current leading techniques for analysing advanced multiplex immunofluorescence images, a key step toward more accurate prediction of patient responses to immunotherapy.
Immunotherapy is a common approach in melanoma treatment, using drugs to activate the body’s immune system to combat cancer. However, many people do not respond as expected to immunotherapy or can experience serious side effects, leading to complications with treatment.
Dr Rana said this technology had the potential to meet the urgent need to support clinicians in predicting how a patient was likely to respond to immunotherapy, enabling them to personalise a treatment plan.
The study introduces COMIL (Channel Optimisation with Multi-Instance Learning)—an AI framework designed to predict immunotherapy response in melanoma patients by analysing multiplex immunofluorescence (mIF)-based spatial proteomic images that capture multiple biomarkers within the tumour microenvironment.
By learning intricate relationships between immune and tumour biomarkers, COMIL captures dependencies often missed by conventional approaches.
Tested on melanoma patient samples from Melanoma Institute Australia, COMIL demonstrated superior accuracy in predicting immunotherapy response, supporting more personalised treatment strategies.
Dr Liu noted that the study was co-developed with the translational research team at Melanoma Institute Australia, led by Associate Professor James Wilmott, whose clinical expertise and collaboration were instrumental in shaping the translational pathway from algorithm design to real-world immunotherapy applications.
“COMIL is a major step toward personalised melanoma management, and we thank Melanoma Institute Australia patients and their families for their contributions to our study,” Dr Rana said.
Professor Helen Rizos, Professor of Cancer Research at the Macquarie University Medical School, said this study shows how AI can turn complex tumour immunofluorescence images into powerful predictors of immunotherapy response. She said that by analysing melanoma images from the Melanoma Institute Australia, the new deep-learning model helps identify the patients who are most likely to benefit from treatment.
“It’s a key step toward choosing the right first-line therapy for every patient and delivering truly personalised melanoma care,” Professor Rizos said.
Melanoma Institute Australia’s Associate Professor Wilmott said the project was an important improvement in the development of AI-based tools to assess the tumour microenvironment.
Further development of such tools may help improve the identification of immunotherapy resistance, he said.
“Teams at Melanoma Institute Australia, from the Personalised Immunotherapy Program, are working to translate these innovations into clinically applicable assays to guide treatment decisions and trial enrolment,” Associate Professor Wilmott said.
Ongoing support from colleagues at the Melanoma Institute Australia and the Charles Perkins Centre, University of Sydney, is also gratefully acknowledged, Dr Rana said.
Beyond melanoma, COMIL’s approach offers broader potential for AI-driven precision pathology across other cancer types.
Source: Australian Institute of Health Innovation
Access the preprint article in the IEEE Journal of Biomedical and Health Informatics here.
