Learn to recognize the symptoms
During the operation, the surgeon takes a small sample of tissue from the edge of the tumor. A special machine called a portable SRH imaging system takes pictures of the tissue sample right in the operating room. A technician can easily use this machine with a simple touchscreen.
This new system, called FastGlioma, is an open-source artificial intelligence-based system designed to detect brain tumor infiltration in real time during surgery.
The research suggests that by identifying these hidden cancer cells, the technique could help prevent the return of more aggressive tumors and completely eliminate less aggressive ones.
For most glioma patients, surgery is the first line of treatment. While surgeons strive for complete tumor removal, this can be challenging, particularly when the tumor is located in delicate or inaccessible areas of the brain.
After taking a photo of the tissue sample and dividing it into smaller pieces, the AI system reassembles the image. The resulting image highlights areas that are likely cancerous, similar to a heat map. This helps doctors quickly identify potential tumor areas.
But, with the help of AI, researchers hope they can change this reality. As detailed in a study published in November 2024, a new AI-based diagnostic system can "improve the ability to identify tumors and hopefully improve survival due to the added tumor being removed," explains co-author Dr. Shawn Hervey-Jumper.
Each patch is examined and scored based on how likely it is to contain cancer cells. Patches that show more of the tumor get a higher score. The AI system learns from millions of other images and it gets better at recognizing cancer cells by studying these images and figuring out patterns. This helps it give more accurate scores to the tissue pieces.
This color-coded image shows the level of tumor infiltration. The AI model analyzes the entire image and assigns a score between 0 and 1, where 0 is low infiltration and 1 is high infiltration. This score matches the categories used by expert doctors to classify tumor infiltration.
FastGlioma leverages this approach to generalize across patient demographics, healthcare systems, and World Health Organization (WHO) brain tumor classifications with minimal supervised training.
Trained based on over 11,000 tumor specimens and four million microscope images, researchers say the system can differentiate between healthy and cancerous cells in just 10 seconds.
Surgeons use specialized techniques, such as neuroimaging and brain mapping, to guide the surgical intervention. Brain mapping delineates areas of the brain responsible for critical functions, enabling surgeons to avoid damage to vital brain areas.
Approximately 80,000 new cases of primary brain tumors are diagnosed annually in the US, accounting for 1.4% of all cancers. Roughly 36% of these tumors are malignant.
More so than other cancers, brain tumors can have devastating, long-lasting impacts on patients' physical, cognitive, and psychological well-being. Despite significant advancements in treating other cancers, progress in brain cancer survival rates has been limited.
Another deep learning model offers real-time identification of cancerous and non-cancerous brain tissue during surgery. Detailed in the January 2023 issue of Biomedical Optics Express, this model uses information about light and tissue texture, as shown by a special imaging technique called optical coherence tomography.
Sources: (Health Tech World) (Cleveland Clinic) (Mayo Clinic) (EndBrainCancer.org) (BrainTumor.org) (Nature.com)
See also: A groundbreaking discovery in cancer drug development
In the study, neurosurgeons analyzed tissue samples from 220 patients. They found that FastGlioma was only wrong 3.8% of the time when identifying remaining cancer cells, while traditional methods were wrong 24% of the time.
Even with meticulous surgical techniques, some cancerous cells may remain. For brain tumors, complete removal is paramount, as any residual tissue can rapidly lead to faster regrowth.
According to the study, the system can identify cancerous tissue invisible to the naked eye, allowing surgeons to remove it immediately or target it with post-surgical therapies.
Gliomas, the most prevalent type of brain tumor in adults, are masses of abnormal cells that arise within the brain or spinal cord.
The tissue sample is placed on a special slide and put into a machine called an SRH imager. This machine can quickly take pictures of the tissue under a microscope, without the need for any special preparation. This means that the tissue isn't damaged during the process.
Brain surgery demands unparalleled precision and constant innovation to safeguard healthy tissue and maximize patient recovery.
Surgical interventions, often involving scalp incisions, are performed to treat conditions affecting the brain, spinal cord, and peripheral nerves.
Even small amounts of residual tumor can dramatically reduce quality of life, shorten survival, and place a heavy burden on healthcare systems. Unfortunately, rates of residual tumor have not improved in the past two decades, and corrective surgeries and post-surgical treatments cost over a billion dollars annually in the US.
Similar to foundation models like GPT-4, medical AI models can be trained on massive datasets and adapted to various clinical tasks.
The SRH machine takes a picture of the tissue sample and divides it into these small patches. Imagine a main picture divided into tiny squares, like a pixelated photo. Each square is a small image, or a "patch."
The tool, which is open-source and patented by UCSF, has not yet been approved by the Food and Drug Administration (FDA). However, researchers are optimistic. "FastGlioma could revolutionize neurosurgery by improving the overall care of glioma patients," said co-author Dr. Todd Hollon.
As researchers have demonstrated, this technology is faster and more accurate than current methods for tumor detection. They envision its potential applications, with Hollon believing it could be used to diagnose other types of brain tumors in both children and adults.
Researchers believe the tool is not only fast and affordable but can also accurately identify other tumor cells. They plan to test the system on a wider range of cancers, including breast, lung, prostate, and head and neck cancers, as outlined in the study.
AI is also being used to improve MRI interpretations for glioma patients. A new model can distinguish between real tumor growth and false positives, especially in uncertain cases, as reported in the Journal of Neuro-Oncology in September 2022.
Artificial intelligence (AI) is revolutionizing the way we live, and now it’s set to change the way we fight cancer. A study led by the University of California, San Francisco (UCSF) and the University of Michigan recently unveiled a groundbreaking new AI tool that can detect hidden brain tumor cells during surgery, potentially leading to more effective treatments and better outcomes for patients.
Brain tumors are notoriously difficult to treat, and survival rates haven’t improved significantly in recent years, but this new AI technology could be a game-changer.
Curious? Click on to learn more about how this AI development can revolutionize cancer treatment.
Revolutionary AI tool helps surgeons spot invisible cancer
AI breakthrough improves brain tumor detection
HEALTH Brain
Artificial intelligence (AI) is revolutionizing the way we live, and now it’s set to change the way we fight cancer. A study led by the University of California, San Francisco (UCSF) and the University of Michigan recently unveiled a groundbreaking new AI tool that can detect hidden brain tumor cells during surgery, potentially leading to more effective treatments and better outcomes for patients.
Brain tumors are notoriously difficult to treat, and survival rates haven’t improved significantly in recent years, but this new AI technology could be a game-changer.
Curious? Click on to learn more about how this AI development can revolutionize cancer treatment.