An AI model capable of predicting cognitive impairment risk in patients undergoing brain tumour resection has been developed.
It offers a new avenue for mitigating the unforeseen cognitive consequences often observed post-surgery.
Lars Smolders, PhD candidate in the Department of Mathematics and Computer Science, created the model to address a long-standing challenge in glioma surgery.
He said: ‘Patients frequently struggle with complex cognitive functions, such as concentration and task execution, following glioma surgery.’
While deficits like hemiparesis or visual field loss are commonly anticipated, the impact on higher-order cognitive functions remains challenging to predict.
The AI model Smolders and his colleagues have introduced utilises preoperative MRI scans to analyse the brain’s white-matter connectivity.
He said: ‘By focusing on structural details in major white-matter tracts visible on MRI, we can estimate how resilient a patient's brain may be to surgical trauma.’
This insight could equip surgeons with crucial preoperative information about potential cognitive outcomes, guiding surgical planning and patient counselling.
Historically, cognitive prognostication after glioma surgery has been elusive despite its profound impact on quality of life.
Smolders’ model now offers a promising predictive framework, allowing the identification of patients at heightened risk for cognitive decline.
However, the model requires validation on a broader patient population before widespread clinical integration.
He said: ‘The conventional approach has been to correlate lesion location with neurological sequelae, yet we discovered this method offers little predictive power for cognitive outcomes.’
This realisation prompted the development of a more sophisticated predictive methodology.
Smolders added: ‘The brain’s complexity continually humbles me. Predicting individual cognitive trajectories requires far more nuance than lesion mapping alone can provide.’
He believes his model has successfully demonstrated predictive capacity and can transform surgical decision-making.
‘What excites me most is that we can quantify a brain’s vulnerability to surgical insult using only preoperative MRI data.’
With further refinement and validation, this model could help neurosurgeons anticipate and reduce the risk of cognitive impairment, ultimately improving patient outcomes in glioma treatment.
