Revolutionizing Glioma Care: Machine Learning and fMRI Predict Post-Surgical Functional Status, study finds

Researchers found that a machine-learning algorithm with resting-state functional MRI can help clinicians foresee surgical outcomes in high-grade glioma patients. They developed a random-forest classifier that was highly accurate for predicting tumor resection outcomes in brain cancer patients.

The findings were published online in the Journal of Neuro-Oncology on 24 May 2024.

"The capability to forecast postsurgical functional outcomes from the initial diagnosis could be advantageous in surgical planning and for better-informing patients of their likely treatment outcomes," the group noted.

High-grade gliomas, characterized by their rapid growth and invasive nature, present formidable challenges for clinicians tasked with navigating the delicate balance between tumor resection and preserving neurological function. Traditional methods for assessing postoperative functional status have often relied on subjective evaluations and empirical observations, leaving room for uncertainty and variability in treatment outcomes.

Patrick H. Luckett, Department of Neurological Surgery, Washington University School of Medicine, St. Louis, MO, USA, and colleagues aimed to develop models capable of predicting functional outcomes in HGG patients before surgery, facilitating improved disease management and informed patient care.

For this purpose, the research team retrospectively recruited adult HGG patients (N = 102) from the neurosurgery brain tumor service at Washington University Medical Center. All patients completed resting-state functional MRI and structural neuroimaging before surgery.

Demographics, tumor location, measures of resting state network connectivity (FC), and tumor volume were used to train a random forest classifier to predict functional outcomes based on Karnofsky Performance Status (KPS < 70, KPS ≥ 70).

The following were the key findings of the study:

· The models achieved a nested cross-validation accuracy of 94.1% and an AUC of 0.97 in classifying KPS.

· The strongest predictors identified by the model included FC between somatomotor, visual, auditory, and reward networks.

· Based on location, the relation of the tumor to dorsal attention, cingulo-opercular, and basal ganglia networks were strong predictors of KPS.

· Age was also a strong predictor. However, tumor volume was only a moderate predictor.

The research demonstrates how machine learning can accurately classify HGG patients' functional outcomes before surgical, chemical, or radiotherapy treatments. These results were achieved using tumor location, age, tumor size, and RS-fMRI measures.

"By incorporating these models into clinical practice, we stand to enhance patient care, enabling personalized treatment plans that balance quality of life with survival," the researchers wrote. "Such models can drive a more nuanced approach to HGG patient management, prioritizing longevity and post-treatment quality of life."

Reference:

Luckett, P.H., Olufawo, M.O., Park, K.Y. et al. Predicting post-surgical functional status in high-grade glioma with resting-state fMRI and machine learning. J Neurooncol (2024). https://doi.org/10.1007/s11060-024-04715-1