New AI Tool Diagnoses Brain Tumors
A Surgical Revolution
Once surgeons get close to the edge of a brain tumor with their scalpel, they face a challenge: determining where the tumor ends and the healthy tissue begins, often referred to as the "margins". The dilemma is whether to cut away some healthy brain tissue to ensure the entire tumor is removed or to preserve the healthy tissue but risk leaving behind potentially dangerous cells.
Scientists in the Netherlands have employed artificial intelligence to aid surgeons in this decision-making. According to a study published on October 10, 2023, in the journal Nature, this method utilizes a computer that scans segments of a tumor's DNA, focusing on specific chemical alterations. These alterations provide a diagnosis of both the type and subtype of the brain tumor. This in-surgery diagnosis, obtained during the early stages of an operation that can span several hours, guided surgeons on how aggressive their approach should be. In the future, this method may even direct doctors towards treatments specifically tailored for a particular tumor subtype.
Recognizing the tumor subtype during surgery is crucial. This breakthrough ensures that surgeons have access to a detailed diagnosis right when they need it most.
Introducing Sturgeon
The innovative learning system named “Sturgeon” was initially tested on frozen tumor samples from past brain cancer surgeries. It diagnosed 45 out of 50 cases accurately within 40 minutes of starting genetic sequencing. In the other five cases, it refrained from giving a diagnosis due to unclear data.
When tested during 25 live brain surgeries, primarily on children, and compared to the traditional method of examining tumor samples microscopically, the new method produced 18 correct diagnoses. In the remaining seven cases, it didn't achieve the set confidence threshold for a diagnosis. Impressively, the diagnosis turnaround was less than 90 minutes, short enough to influence in-surgery decisions.
At present, doctors can either examine brain tumor samples under a microscope or send them for more comprehensive genetic sequencing. Both these methods can be time-consuming. Furthermore, not all hospitals have access to advanced genetic sequencing technology. Even if they do, results can take weeks, forcing treatment to begin without full knowledge of the tumor type.
Sturgeon, however, employs a rapid genetic sequencing technique on just a tiny section of the cellular genome. This allows it to deliver results before a surgeon begins working on the tumor margins. Moreover, its sophisticated model can provide a diagnosis even with minimal genetic data.
To put it into perspective, imagine being shown only 0.01% of the pixels from a random part of an image, yet still accurately identifying the entire picture. That's the efficiency of this system.
Promise & Pitfalls
While the Sturgeon system offers significant advancements, certain challenges persist. Diagnosing some tumors remains difficult. For instance, the surgical samples are minuscule, roughly the size of a kernel of corn. If these samples contain healthy brain tissue, Sturgeon might struggle to detect sufficient tumor-specific markers. To counter this, in the study, pathologists were tasked to examine the samples microscopically. They then highlighted those rich in tumor content, flagging them for sequencing.
Another complexity arises from the heterogeneity of tumor cells within a single patient. The small segment sequenced might not truly represent the entirety of the tumor. Additionally, rare tumors might not match any of the previously classified types. It's also noteworthy that certain tumor types are identified more effortlessly than others.
Despite these challenges, other medical facilities have begun adopting this new method with surgical samples. This suggests its viability even in less seasoned hands. Yet, it's crucial to remember that sequencing and categorizing tumor cells still demand proficiency in bioinformatics. Moreover, it necessitates individuals adept at managing, troubleshooting, and maintaining the technology.
An Enormous Step in the Right Direction
Brain tumors are particularly suited for classification based on the chemical modifications that the new method targets. However, this approach isn't universal; not all cancers can be diagnosed in this manner.
While significant progress has been made in understanding the molecular profiles of tumors, advancements in actual treatment have been more modest. The primary goal for doctors and scientists is to develop therapies that are less harmful to the nervous system. But the journey from gaining deeper insights into tumors to formulating new therapies is complex and challenging.
The introduction of this new method signifies a broader shift towards molecular precision in tumor diagnosis. Research is ongoing.