Use cases

Sharpen tumor insights with HSI

robina-weermeijer-3kgf9r-0ohs-unsplash

Use Case Description

Brain and central nervous system (CNS) cancers, particularly glioblastoma (GBM), are among the most aggressive and fatal forms of cancer. Effective surgical removal of the tumor is crucial to improve patient survival. While preoperative planning becomes very advanced, it remains a challenge to translate these plans to the surgical scene, and it is still difficult to distinguish cancerous tissue from healthy brain matter during surgery, especially tumor borders. This uncertainty poses a major challenge for neurosurgeons, as removing too little tumor tissue may lead to recurrence, while removing too much can severely impact brain function and quality of life. 

To address these challenges, this use case proposes an AI-powered, multimodal imaging platform that combines Hyperspectral Imaging (HSI) and additional acquired medical image information to enhance tumor detection and surgical precision.This cutting-edge approach captures both spatial and spectral information, offering a deeper understanding of tissue composition. Combined with AI-driven data analysis, the system can automatically detect biomarkers, classify tumor subtypes, and provide quantitative insights for better diagnosis and prognosis. ultimately leading to improved patient care. 

Before surgery, pre-operative plans are carefully established on multi-parametric MRI to identify different regions of the tumor. 

During surgery, surgeons must rely on neuronavigation systems that align preoperative images with the operative field. However, as the brain shifts during surgery, these images quickly become inaccurate. While intraoperative MRI provides real-time updates, it is costly and impractical in most surgical settings. Other methods, such as fluorescence imaging, offer some tumor visibility but needs contrast agents and not all tumors, especially lower grade gliomas, do not respond to these imaging techniques. , including a lack of depth perception and the need for specialized lighting. 

Our AI-powered imaging solution provides real-time, high-resolution visualization of tumor tissue directly in the operating room. By leveraging HSI, in combination with preoperative taken images, surgeons can instantly detect tumor boundaries without the need for contrast agents or additional surgical delays. AI-driven interpretation of these imaging techniques ensures that even the smallest infiltrative tumor regions are identified, leading to a more complete and safer resection. This technology helps surgeons make confident, data-driven decisions during the procedure, ultimately improving patient outcomes. 

By combining advanced imaging techniques with artificial intelligence, this use case aims to enhance every stage of brain tumor management. This innovative approach empowers clinicians with better tools, improves surgical precision, speeds up diagnosis, and enhances patient outcomes. 

Results, highlights & Research Outputs

2025

  • A paper on domain adaptation for oxygenation predictions in esophagus surgeries to remove tumor cells was presented at MICCAI conference in September.

  • At the MICCAI conference we won the AI challenge to automatically segment brain tumors in pre and post op MRI scans

  • The AI analysis of a clinical study to use AI for brain tumor detection in HSI was finalised and presented at the Flanders AI research day in October.

  • A HSI setup was showcased at the demo booth at the flanders AI research day in october.

  • A handheld prototype to capture intraoperative hyperspectral images in UZ Brussel was designed

  • A review paper on ‘the role of AI in medical hyperspectral imaging’ was written

Outputs & Resources

Publications

  • A paper on domain adaptation for oxygenation predictions in esophagus surgeries to remove tumor cells, presented at MICCAI.

  • A review paper on “the role of AI in medical hyperspectral imaging”.

Demonstrators / Applications

  • HSI setup showcased at the Flanders AI Research Day.

  • Handheld prototype for intraoperative hyperspectral imaging in UZ Brussel.

  • AI-based brain tumor detection in hyperspectral imaging.

Team & Contact

Contact

UC Lead: Siri Willems (imec)

Use Case Lead(s)

Siri Willems, imec