Building trust and busting bias in AI for cancer diagnosis


2 days ago

The Bias Busters team at ATU. From left: Alan Hernon, Saritha Unnikrishnan, Oladosu Oladimeji, Obed Adjei Peprah and Nihal Elfodil. Image: James Connolly

AI research is about ‘pushing the boundaries of what’s possible’, says ATU computer scientist Dr Saritha Unnikrishnan.

Last month, Dr Saritha Unnikrishnan led a team of researchers to victory in the National AI Challenge 2024 for a prototype aimed at addressing bias in artificial intelligence (AI) systems.

The Bias Busters team was commended for their innovative and ethical approach to AI bias. Their system detects bias in qualitative reviews and compares results to the Google Vertex AI model predictions to address potential biases in model predictions and suggest mitigation strategies.

At the time of the win, Unnikrishnan said the idea was to develop a product that would make “a meaningful contribution to the AI community”.

Before joining the ranks of academia, Unnikrishnan worked as a software engineer and as a systems analyst. In 2015, she undertook a PhD in computer vision and machine learning, focusing on pharma applications, at Atlantic Technological University (ATU). She is now a lecturer in computing at ATU.

She is programme chair for the master’s in data science and leads the postgraduate research training programme focused on AI and robotics. She is also a principal investigator in computer vision and machine learning for three strategic research centres at ATU.

“I’ve had the opportunity to wear many hats throughout my career,” Unnikrishnan tells SiliconRepublic.com. Now, in her research career, she is involved in projects that are “pushing the boundaries of what’s possible”, she says.

Here, she tells us more about some of those projects.

Tell us about your current research.

Some very exciting research projects I am currently working on are explainable artificial intelligence (XAI) methods accelerating the trust and generalisability of AI in cancer diagnosis, multimodal AI approaches in breast cancer screening and staging, generative AI approaches for developing virtual contrast enhancement tools for brain MRI, and a digital twin project using multimodal AI and sensor fusion for the optimisation of the spray drying process in dairy manufacturing.

Click here to listen to Future Human: The Series.

I am also in the final year of leading an Enterprise Ireland-funded commercialisation project that aims to bring a spin-out developing AI-assisted image analysis software for a pharmaceutical application.

In your opinion, why is your research important?

My research is trying to address some of the pressing challenges in healthcare and energy-intensive manufacturing by leveraging the power of AI and at the same time addressing some of the well-known AI challenges such as lack of transparency and generalisability.

Societal good and improvement in quality of life are the two main outputs I foresee.

What inspired you to become a researcher?

Curiosity has always been at the core of my inspiration to become a researcher. I cannot recall one specific moment that sparked my journey into research, but there were many occasions where I realised that if you truly want to make an impact in any field you’re passionate about, research is the most powerful tool to do so.

What are some of the biggest challenges or misconceptions you face as a researcher in your field?

One of the biggest challenges I’ve faced as a female researcher in the field of computer science and now data science/AI is the sense of feeling like an outsider, almost like an alien at times. Gaining trust from listeners – whether it’s the general public or during an industry pitch – can often be more difficult for women than for men. The gender imbalance and underlying biases in these fields remain persistent issues.

While there is some positive change happening, with more awareness and efforts toward inclusivity, we still have a long way to go before true equity is achieved. These challenges not only impact representation but can also affect how our work and contributions are perceived.

Do you think public engagement with science and data has changed in recent years?

Yes, there have been many positive changes in that respect since the Covid-19 pandemic. People from all backgrounds have come to recognise the power of data visualisation and analysis tools, particularly how data can be made more meaningful using such tools to provide real-time insights on health, safety and policy decisions.

In my field of AI and data science, this shift in awareness of AI continued with the release of OpenAI’s ChatGPT in late 2022. This event marked a major milestone for AI, as the technology moved from development into widespread real-world deployment, which made it possible for the public to use and understand the potential of AI and at the same time identify the risks and challenges that it may pose if left unregulated. It also opened up broader conversations about the responsible use of AI and its impact on society.

How do you encourage engagement with your own work?

Collaboration across disciplines is key, whether it is with industry, academia or clinicians. I’ve found that European COST actions are a fantastic way to engage with a consortium of world-leading researchers across various fields, allowing us to develop research funding applications together and drive innovative solutions.

In the fields of computer science and data science, there’s a growing shift towards open-source research. I encourage my students to make their research results available to others whenever possible as it promotes trust and encourages further collaboration.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.