2021 PDNC Chao Li

Dr Chao Li

2021 Post-doctoral Non-clinical Fellowship

Integrated AI tools for the prediction of tumour invasion using brain MRI scans

Glioblastoma is a highly aggressive type of brain cancer that poses significant difficulties in its treatment. The current standard of care involves surgical removal of the tumour followed by radiation therapy. However, one of the major challenges is the frequent recurrence of tumours in areas adjacent to the treated region, indicating the presence of undetectable tumours using conventional diagnostic methods.

Artificial intelligence (AI) holds great promise in addressing this issue by enabling the effective detection of tumour regions, thus facilitating more targeted and effective surgical procedures and radiation therapy. However, the development of robust AI models often requires access to large volumes of clinical data, which can be challenging to obtain in practical healthcare settings. Moreover, for the sake of patient safety, it is crucial for clinicians to have a clear understanding of the decision-making process of AI models, which is often lacking in black-box AI models.

To overcome these challenges, this project aims to develop novel AI techniques specifically designed to predict tumour invasion based on brain MRI scans. The goal is to create an AI solution for real-world healthcare that is both reliable and trustworthy. The proposed AI model could be seamlessly integrated into existing clinical systems, allowing for practical implementation within healthcare facilities. Furthermore, its performance and accuracy will be rigorously evaluated using the data collected from clinical trials.

The integration of AI tools has the potential to significantly enhance the ability of clinicians to identify hidden tumour regions. By providing precise information about tumour invasion, the AI solution can help guide surgical interventions and radiation therapy, leading to improved patient outcomes. Additionally, once the AI solution is validated, it can be scaled and deployed across multiple healthcare centres, offering a cost-effective approach to benefit the healthcare system and our society at large.


Multi-modal learning for predicting the genotype of glioma

Wei, Y., Chen, X., Zhu, L., Zhang, L., Schönlieb, C.B., Price, S. and Li, C.

IEEE Transactions on Medical Imaging


Structural connectome quantifies tumour invasion and predicts survival in glioblastoma patients

Wei, Y., Li, C., Cui, Z., Mayrand, R.C., Zou, J., Wong, A.L.K.C., Sinha, R., Matys, T., Schönlieb, C.B. and Price, S.J

Brain 146(4), pp.1714-1727