Machine learning and diffusion-weighted imaging can effectively classify the characteristics and diagnosis of common types of pediatric brain tumors. As a result, the tumor can be characterized and treated efficiently.
Why is it difficult to classify pediatric brain tumors?
Brain tumors in a particular part of the brain, known as the posterior fossa, are the largest cause of death from cancer in children. However, three types of brain tumors might exist in this area, and it is often challenging to characterize them efficiently and quickly.
Currently, to characterize the tumors, radiologists make a qualitative assessment of the MRI. However, overlapping radiological characteristics often make the task difficult for them to distinguish the type of tumor without biopsy confirmation.
Diffusion-weighted imaging uses specific advanced MRI sequences and software that generates images from the resulting data using the diffusion of water molecules. It all results in contrast in the MR image. Then, it is possible to extract an Apparent Diffusion Coefficient (ADC) for better studying and understanding of the tumor.
The Study… and the Breakthrough
The study, which led to the above discovery, involved 117 patients from five different primary treatment centers across the UK. The scans were taken from twelve different hospitals and eighteen different scanners. The resulting images were analyzed and the region of interest was drawn by an expert scientist in pediatric neuroimaging and an experienced radiologist.
The values obtained from the analysis of Apparent Diffusion Coefficient maps from the above image regions were then fed to AI algorithms. It successfully discriminated the three most common types of pediatric posterior fossa brain tumors without any invasion.
The overall process is not only exciting from the viewpoint of science and technology but also for a child and their family. These artificial intelligence techniques will hopefully be made widely available to help doctors with a high level of diagnostic accuracy.