PŪTAHI RANGAHAU/AUT RESEARCH CENTRE

Medical Imaging AI

The Medical Imaging AI group at AUT focuses on developing advanced deep‐learning solutions to improve medical image interpretation, especially under challenging data conditions such as limited labels and heterogeneous multimodal inputs for applications in brain and clinical diagnostics.

Exploring cutting-edge machine learning approaches for healthcare imaging, including efficient training methods with limited data, multi-modal analysis combining diverse imaging techniques with clinical information, and robust adaptation across different medical devices and patient populations to enhance diagnostic accuracy and clinical deployment.

  • Label-efficient deep learning
    Few shot, meta learning, synthetic augmentation and test time adaptation for medical imaging.
  • Multimodal image analysis
    Cross-modality analysis of MRI, CT, fundus, histology and clinical metadata, especially for early disease detection.
  • Robust domain adaptation
    Handling distribution shifts across imaging devices or patient cohorts.

Core members: Catherine Shi, Yanbin Liu, Boris Bacic
Funding/fellowship: Google Cloud Research Credits 2023, Lambda Research Program 2025, DAAD AINet Fellowship (AInet Fellows 11/2024 - AI for Science)

Our publications

  • ECCV24, Unsupervised Dense Prediction via Normalized Cuts: https://www.youtube.com/watch?v=a8HIgnt4Ps0&ab_channel=YanbinLiu
  • NeurIPS 24, Topology Preserving Reservoirs (TPR): https://neurips.cc/virtual/2024/poster/92938
  • Priyadarshini N, Ojeda Y, Shringare S, Bacic B, Rathee M,. Deep learning framework for pre-surgical risk assessment of mandibular wisdom teeth: Implications of image enhancements. 2025. 6th International Conference on Sustainable and Innovative Solutions for Current Challenges in Engineering and Technology Artificial Intelligence and Sustainable Computing. Algorithms for Intelligent Systems Artificial intelligence and sustainable computing: Proceedings of ICSISCET. 20242:29-39 Springer. 10.1007/978-981-96-3337-1_3
  • Bačić, B., Claudiu Vasile, Feng, C., & Ciucă, M. G. (2024, 13-15 Dec.). Towards nation-wide analytical healthcare infrastructures: A privacy-preserving augmented knee rehabilitation case study. Presented at the meeting of the Conference on Innovative Technologies in Intelligent Systems & Industrial Applications (CITISIA 2024), Sydney, NSW. [In print][ArXiv.org Preprint: https://arxiv.org/abs/2412.20733 and the source code on GitHub https://github.com/bbacic/tnwahi-appakrcs?tab=readme-ov-file and https://github.com/claudiunz/tnwahi-appakrcs]
  • Feng C, Bačić B, Li W,. SCA-LSTM: A deep learning approach to golf swing analysis and performance enhancement. 2025. 31st International Conference on Neural Information Processing (ICONIP 2024). Neural Information Processing: Proceedings Part XI:72-86 Springer Nature Singapore. 10.1007/978-981-96-6606-5_6
  • Chen C, Gao J, Bačić B, Zhou S. Enhancing the starting performance of elite swimmers through eight weeks of lateral entry training. 2024. Sports Biomechanics. 10.1080/14763141.2024.2400531
  • Rathee M, Bačić B, Doborjeh M. Hybrid machine learning for automated road safety inspection of Auckland Harbour Bridge. 2024. Electronics13(15) : 3030 MDPI 10.3390/ELECTRONICS13153030
  • Bacic B, Feng C, Li W JY61 IMU sensor external validity: A framework for advanced pedometer algorithm personalisation. 2024. 42nd International Society of Biomechanics in Sports Conference ISBS Proceedings, 42(1):60-63 https://commons.nmu.edu/isbs/vol42/iss1/28/
  • Boris Bačić. Predicting golf ball trajectories from swing plane: An artificial neural networks approach, Expert Systems with Applications 2016Expert Systems with Applications, 65:423-438. Elsevier. https://doi.org/10.1016/j.eswa.2016.07.014
  • Bačić B, Hume PA. Computational intelligence for qualitative coaching diagnostics: Automated assessment of tennis swings to improve performance and safety. 2018. Big Data, 6(4):291-304.  10.1089/BIG.2018.0062
  • Boris Bačić. Echo state network ensemble for human motion data temporal phasing: A case study on tennis forehands. 2016. 23rd International Conference ICONIP 2016Neural Information Processing. ICONIP 2016 LNCS vol. 9950:11-18 Springer Verlag. 10.1007/978-3-319-46681-1_2

Video presentations

  • 15 Nov 2023 The Digital Pulse: Transforming Healthcare: Feng C, Bacic B, Li W Towards open-source IMU sensors for activity monitoring Alison Mackie (NZ IoT Alliance), Boris Bacic (AUT), Hanie Yee (Alimtery), Sam Madanian (AUT),   and Josh Wilson (Spark).  Transforming healthcare with IoT streaming - Panel discussion (Start-End time 41-60 minutes). 15 Nov 2023 The Digital Pulse: Transforming Healthcare. Published by: New Zealand IoT Alliance https://iotalliance.org.nz/event/the-digital-pulse-transforming-healthcare/
  • Backswing Length and Swing Path Tendencies from Amateur Golfers (received "ISBS 2020 People's Choice" award, published by: International Society of Biomechanics), (10:29min) https://www.youtube.com/watch?v=iiQSJwhb9mU