AI^2 Forum April 2024
This month at AI^2 the downfall of the company that once claimed their goal was “to put accessible, affordable health service in the hands of every person on Earth” was discussed.
University of Leeds
This month at AI^2 the downfall of the company that once claimed their goal was “to put accessible, affordable health service in the hands of every person on Earth” was discussed.
On April 10th the LIDA Health Early Career Researcher (ECR) community held their first event. This was the first of 5 events aiming to empower ECRs with skills to identify knowledge gaps, write project proposals, advance career trajectories, foster collaborations, and expand networks.
The Deep Learning module this week is moving on from simple backpropagation in a fully connected neural nets to consider convolution: the practical this week involves image convolution using 3 x 3 kernels to allow edge detection.
Mary presented at the February edition of Cafe Scientifique. She presented on the use of AI to detect diseases from speech in a presentation titled “Dr Alexa: detecting disease from your voice”. We discussed the detection of throat cancer and dementia and the challeneges that rise from using AI in healthcare. It was great to discuss this work with a diverse audience and see hear how the public views the possible use of AI in the future of healthcare.
We had an unforgettable social night out with medical AI colleagues at Pizza Punks in town, where we enjoyed great pizza after a long exhausting brain-drilling day. And got recharged with new research ideas.
On the 20th of December 2023, the AI^2 forum hosted its second Christmas bake-off! This year, we had a range of exciting entries, from delicious Snow DAG chocolate fudge cakes and Random Forest Brownies to melted Snowpeople Cookies and Gingerbread AI^2 Cookies. Everyone enjoyed having a nibble on a sweet treat or three; the entries were amazing, and judging these bakes was immensely rewarding for our tummies!
Three of our students ran an AI workshop for classes of year 7, 8, and 9 students. In this workshop we explained how decision trees can be made and challenged the students to make some of their own. We then showed the students some decision trees in action when we got an AI to guess their age from the sound of their voice. We also discussed how object detection works and how it can be used (for example self driving cars). We showed an example of live object detection and the students tried different objects to see if the detector could recognise them. We were featured on the school blog here!
One of our students attended the Hamlyn Winter School on Surgical Imaging and Vision at Imperial College London, hosted by the Hamlyn Centre. The week-long course featured technical and clinical talks as well as small lectures and workshops from industry (NVIDIA and MedTronic). In addition to lectures covering state of the art in surgical imaging, robot-assisted imaging and vision algorithms for tracking, scene reconstruction and navigation, the Hamlyn Centre provided participants the opportunity to collect their own data with a variety of surgical robots including the da Vinci surgical system. The week culminated with group presentations of small projects completed over the week.
As part of Bradford Tech Fest, our students ran a number of workshops in schools about the use of generative AI. These sessions, delivered to pupils between the age of 15 and 18, covered how the pros and cons of tools such as ChatGPT, as well as an overview of how image diffusion models work (and where they go wrong!). The sessions rounded up with a lively debate where the students were asked to consider what would do if they were a headteacher of a school forming a policy on the use of AI tools. Should students be allowed to utilise these tools? The verdict – it’s complicated!
The International Conference for Process Mining (ICPM) 2023 commenced on October 23rd. Among the attendees from the University of Leeds were three students—Zoe Hancox, Alexander Coles, and Abdulaziz Aljebreen—accompanied by their lecturer, Owen Johnson. The focus of their participation was the Process-oriented Data Science for Healthcare (PODS4H) workshop, where they shared their innovative research and insights.
In this month’s forum, we had two fantastic guest speakers: Associate Prof. Dave Wong and Dr. Alexa Ruppertsberg.
One of our clinicians in the CDT, Allan Pang, presented his work on ML-driven physiological early warning systems this summer at the Machine Learning for Healthcare Conference at Colombia University, New York City. As the name suggests, this conference lies at the intersection of ML and Healthcare to encourage collaboration between clinicians and ML practitioners. Some take-home reflections following the meeting are the need to consider the broader systemic effect that the ML model will have once implemented rather than just the traditional data-centric metrics. For instance, a positive change in healthcare behaviour is more critical than increasing application engagement. This approach requires a clear definition of the problem that needs solving and clear metrics to monitor progress. If we fail to do this, this can lead to blind optimisation and unforeseen consequences for patients and healthcare staff alike. Of course, not one to miss the opportunity to combine a run and sightseeing, the weather allowed for a Sunday morning run across the Brooklyn and Manhattan bridges to admire the New York skyline!
This week we hosted 6 year 12 students for a summer placement through In2Science. We had a jam packed schedule over the five days they were with us.
At the Child Friendly Leeds Live festival, PhD students, including Victoria Moglia, Morgan Thomas, Oliver Umney, Xin Ci Wong, Zoe Hancox, and Oliver Mills, enthusiastically contributed to the BeCurious initiative. Their collective effort centered around running an engaging stall aimed at sparking children’s curiosity about the realm of artificial intelligence and its applications in healthcare.
In the July 2023 edition of AI^2, we had the privilege of hosting two esteemed guests from NHS England: Dr. Dan Schofield and Dr. Jonathan Pearson. They provided us with a fascinating overview of NHS England and its Digital Analytics Research Team (DART).
In AI^2 the June 2023 edition Zoe and Jason hosted a testathon session where together we explored Pytesting of AI models to improve our code reliability, reduce bugs and increase our error finding efficiency.
On June 22nd, our fellow PhD students had the incredible opportunity to attend and present posters at the N8 Digital Health Community Day held at the University of Leeds, Nexus. This event was a gathering of experts from various fields, providing a platform for knowledge exchange and collaboration in the realm of digital health.
This week two students attended UK Speech. Mary presented a poster about the classification of vocal pathologies. This was a fantastic opportunity to meet with other researcher using speech technologies and learn from some of the top experts in this field.
Dr. Kieran Zucker spoke at the AI^2 forum about the challenges of implementing AI in healthcare. Despite the widespread belief in AI’s potential to revolutionize healthcare, there are several barriers preventing its effective use in clinical settings. The main takeaway points from the talk are as follows:
We were very excited to be invited to take part in National Numeracy Day at the University of Leeds. We had the opportunity to explain the role that numeracy has in our jobs to groups of students from local secondary schools. Each student was encouraged to write down what they had learned about our jobs after speaking to us and some of what they said can be seen in the pictures below.
This year we got the opportunity to show off some of our work at Be Curious. This is an event run by the university which gives the opportunity for researchers to present their work to families. We hosted three different stalls all about the exciting work we do in medical AI. The first stall was all out how wearable sensors can be used to monitor patient health. The second stall allowed families to see how ultrasound could be used in diagnosis and have a go using a model of the throat. The third stall pitted families against an AI in guessing someone’s age based only on their speech. We had a brilliant time showing off our research to the next generation of researchers!
This year we hosted the AI CDTs in Healthcare Conference in York. Students from each of the four CDTs presented their research both with oral presentations and posters. We also had two excellent keynotes speakers - Dorothy Monekosso presented “Why is innovating in digital health challenging?” and Richard Stephens presented “Fleshing Out The Data – Involving Patients In Your Research”. It was a great opportunity to meet with other students doing work in this field to discuss ideas and challenges while working in this area.
Dr Keiran Suchak and Dr Peter Hristov gave a talk titled “What to Expect When You’re Expecting… the End of Your PhD”. We enjoyed soaking up all their fantastic tips and tricks for navigating the challenging journey of completing a PhD. Our speakers inspired and motivated our own PhDs with renewed energy and a sense of purpose. From their practical advice on overcoming writer’s block to their heartfelt reminders to take time to celebrate our achievements and acknowledge those who have helped us along the way, their advice will stay with us as we move forward in our academic and professional career. The slides from their talk can be found in this link and we’ve also captured and compiled some of the insights from this talk below:
In the AI^2 March 2023 edition (TL;DR – four clinicians discussing the challenges and opportunities for data science and AI in their specialties).
This week, three of our PGRs had their first experience discussing their research at an international conference! This included presentations about efficiency in ovarian cancer classification using histopathology data, explainability in the classification of chest x-rays, and a poster about the effective compression of histopathology data. The San Diego conference was a fantastic opportunity to connect with other researchers, discuss new ideas, and learn more about the wider field of AI in medical imaging.
Hamidreza Saligheh Rad gave a talk about his work in Radiology and the impressive companies and initiatives he’s been a part of. Teams then competed to create art that depicted a flying pig drawing a masterpiece. The challenge was to deceive other teams about the source of the art (AI 🤖 or human 🎨 ). Check out some of the creations below and see if you can tell the difference between AI and human-generated art!
Alexia Toumpa was our guest speaker who shared her fascinating work on ‘Prediction of object interactions in the real-world’ and thoughts on her time in industry.
We had an AI-themed Christmas bakeoff and medical quiz to end the year. 10 bakers took part in the bakeoff.
Two of our students attended the Conference on Neural Information Processing Systems (NeurIPS) in New Orleans. Mary presented a poster in the Women in Machine Learning workshop summarising her work using AI to detect voice disease from speech. Bailey presented a preprint of his work at the New Frontiers in Graph Learning workshop. This conference was an amazing opportunity to meet a wide range of researchers and see the newest cutting edge work being done in computer science.