University of Nottingham Ningbo ChinaResearch-CentresAI Research InstituteResearch cases

Research cases

CFFormer: Cross CNN-Transformer channel attention and spatial feature fusion for improved segmentation of heterogeneous medical images Artificial Intelligence

CFFormer Image (Click the image to enlarge)

Medical image segmentation is an important part of computer-aided diagnosis, as it helps identify the size and location of lesions and supports clinical decision-making. However, accurately marking these regions often depends on image quality and the experience of radiologists, which can increase the workload in healthcare settings. Although recent advances have improved performance, many existing methods still struggle with low-quality images, where unclear boundaries and noise make it difficult to detect lesions reliably.

To address this issue, we propose a new model called CFFormer, which is designed to better handle blurred areas and capture important information from the entire image. We evaluate the model across five types of medical imaging and eight datasets, and the results show that it performs more reliably than existing approaches, especially on low-quality images. We expect this work to provide a useful reference for future research in medical image segmentation and contribute to the development of more robust computer-aided diagnosis systems.

Jiaxuan Li
Jiaxuan Li

PhD student at the University of Nottingham Ningbo China, supervised by Professor Sean He. Completed both undergraduate and master’s degrees at the University of New South Wales. Research focuses on medical image segmentation and self-supervised learning.

Publication

Falling out with AI-buddies: The hidden costs of treating AI as a partner versus servant during service failure

AI difference Image (Click the image to enlarge)

This research examines how people interact with AI virtual assistants when things go wrong. Companies often design AI to feel more human-like, sometimes presenting it as a “partner” that works with users, rather than a “servant” that simply follows instructions. While this partner-like framing is usually thought to improve user experience, this study shows that it can have unexpected downsides. Across a series of experiments, the researchers find that when an AI system is seen as a partner (versus servant), users are more likely to blame themselves when a service fails.

This happens because people begin to feel psychologically connected to the AI, almost as if it is part of themselves. As a result, when something goes wrong, the failure feels more personal. However, this also reduces users’ confidence in their ability to use the AI and makes them less willing to use it again. The research also shows that companies can reduce the negative effects of failure by emphasizing that the AI is capable of learning and improving over time. Overall, the study highlights a hidden trade-off: making AI feel more like a human partner can strengthen relationships in good situations, but may backfire when problems occur.

Author List: Bo Huang, Sandra Laporte, Sylvain Sénécal and Kamila Sobol

Bo Huang
Bo Huang

Bo Huang is an Assistant Professor in Marketing at Nottingham University Business School China. His research interests include consumer-technology interaction and services marketing, with a particular focus on service failure. His research work has been published in word-leading journals such as International Journal of Research in Marketing, International Journal of Production Management, Journal of Interactive Marketing, Psychology & Marketing, Technological Forecasting & Social Change, among others.

Publication

×