Researchers from Harvard Medical School have created an AI tool named PICTURE (Pathology Image Characterization Tool with Uncertainty-aware Rapid Evaluations) that distinguishes between glioblastoma and primary central nervous system lymphoma (PCNSL), two look-alike brain cancers. Glioblastoma is the most common aggressive tumor, while PCNSL, more rare, originates from immune cells. Misdiagnosis can lead to inappropriate treatment, emphasizing the tool’s importance in neuro-oncology.
PICTURE boasts over 98% accuracy in identifying these tumors through real-time analysis during surgery, aiding surgeons in making informed decisions about whether to operate or opt for alternative treatments like radiation. The AI system assesses critical cancer features and includes an “uncertainty detector” that signals when it lacks confidence in a diagnosis.
Tested across five hospitals, PICTURE outperformed both traditional methods and human pathologists, proving vital in improving diagnostic precision in brain cancer treatment. Future adaptations may extend its capabilities to other cancer types and diverse populations.
Source link
