A groundbreaking tool that combines rapid DNA sequencing and artificial intelligence is revolutionizing the field of neurosurgery by significantly improving the classification of central nervous system (CNS) tumors during brain cancer surgery. Developed by a multi-institutional team of researchers in the Netherlands, this tool, known as Sturgeon, enables neurosurgeons to make better decisions about the extent of tumor resection, ultimately benefiting patients’ outcomes.
Before surgery, neurosurgeons have limited knowledge of the tumor type. As the surgery begins, samples of tumor tissue are removed for immediate histological assessment. However, obtaining a definitive diagnosis through DNA sequencing and molecular analysis typically takes a week.
In contrast, Sturgeon can accurately diagnose most CNS tumors within 90 minutes. Armed with this knowledge, neurosurgeons can modify their surgical strategy in real-time, guided by the precise tumor classification obtained from Sturgeon.
During surgery, neurosurgeons often leave a small remnant of tumor tissue to prevent neurological damage. However, if it later turns out that the tumor is highly aggressive, a second surgery may be necessary to remove the remaining tumor. With Sturgeon, this can be avoided as the tumor type is determined during the first surgery.
The effectiveness of Sturgeon was demonstrated through a study published in the journal Nature. The researchers from UMC Utrecht, Amsterdam UMC, and the Princess Máxima Center for Pediatric Oncology developed, trained, and tested Sturgeon. During 25 surgeries, Sturgeon accurately classified 72% of tumors in less than 45 minutes.
Sturgeon employs rapid nanopore sequencing, a technology that reads DNA in real-time, to obtain methylation profiles of tumor samples. Methylation patterns are unique to each tumor type, enabling molecular subclassification of CNS tumors. The neural network classifier used in Sturgeon is patient-agnostic and runs on a laptop computer in just a few seconds.
To overcome the limited availability of methylation datasets, the researchers developed a strategy to generate realistic training data. Sturgeon utilizes these data to simulate thousands of unique nanopore sequencing experiments for each tumor methylation profile. The final Sturgeon models were trained on 36.8 million simulated nanopore runs and validated on an additional 4.2 million.
Sturgeon’s efficacy in diagnosing pediatric CNS tumors was validated using methylation profiles from actual patients. In simulated sequencing experiments, Sturgeon correctly classified 95.3% of samples within 25 minutes and 97.1% within 50 minutes.
While Sturgeon has shown great promise, it does have limitations. It performs best on tumor samples that are well-represented in the training data and struggles with rare tumor types. Additionally, samples containing less than 50% abnormal cells yield lower accuracy, and large tissue samples are required for sufficient DNA concentration.
Despite these limitations, the impact of Sturgeon in improving the classification of CNS tumors during brain cancer surgery is undeniable. As further advancements are made, including applying this method to other tumor types and validating its effectiveness in larger patient populations, Sturgeon has the potential to transform neurosurgery and significantly benefit patients worldwide.
FAQ
What is Sturgeon?
Sturgeon is a tool that combines rapid DNA sequencing and artificial intelligence to classify central nervous system tumors during brain cancer surgery.
How does Sturgeon work?
Sturgeon uses rapid nanopore sequencing to obtain methylation profiles of tumor samples. These profiles are then classified using a neural network algorithm, providing precise tumor classification in real-time.
What are the benefits of using Sturgeon during surgery?
By knowing the tumor type and aggressiveness during surgery, neurosurgeons can modify their surgical strategy to maximize tumor resection while minimizing the risk of permanent neurological damage. This knowledge also helps avoid the need for additional surgeries when the tumor is highly aggressive.
What are the limitations of Sturgeon?
Sturgeon performs best on tumor samples that are well-represented in the training data and may struggle with rare tumor types. It also requires large tissue samples for sufficient DNA concentration and is less accurate when analyzing samples with less than 50% abnormal cells.
How will Sturgeon be further developed?
Future developments of Sturgeon include applying it to other tumor types, prospective validation to demonstrate patient benefit, and conducting studies in larger patient populations.
(Source: Physics World)