Artificial Intelligence Accurately Diagnoses Brain Tumors in Austrian Study

Did you know that although most (about 71% of) brain tumors are benign? Tumors of the brain remain one of the most feared neoplasias in any organ. Brain tumors are feared because they are easily misdiagnosed since symptoms (such as headache, nausea, and vomiting) are non-specific. However, whether benign or malignant, a tumor arising in the brain results in compression and subsequent damage of normal brain tissues which could eventually result in serious symptoms like neurological damage, seizures, and even death. Although modern means such as magnetic resonance imaging (MRI) make brain tumors easily detectable, these techniques do not accurately classify the brain tumor, a prerequisite for accurate diagnosis of tumors, and establishing an optimal treatment plan.

Brain Tumors

In a study published in Cancers, and conducted by scientists at the Karl Landsteiner University for Health Sciences (KL Krems) in Austria, it was discovered that with the right resources, not only could Artificial intelligence diagnose brain tumors, it could do so more efficiently than experts. With this technique, the rate of misdiagnosis of brain tumors could be drastically reduced.

Radiophysionomics

Radiophysionomics, as the technique is called, is a newly developed technique that uses artificial intelligence to diagnose brain tumors.

To develop this technique, firstly, the team trained nine well-known Machine Language algorithms with data obtained from MRI of 167 previous patients who had one of the five most common brain tumors and had accurate classification using histology. A total of 135 classifiers were generated in a complex protocol. Each of these classifiers was a mathematical function that assigned the material to be examined to specific categories.

After training, the algorithms had a greater success rate not only in detecting but also in classifying brain tumors. They theorized that diagnosis using Artificial Intelligence was superior in areas of accuracy, precision, and misclassification to diagnosis obtained by experts.

The precision of this method arises from the fact that, unlike previous studies, the team also took into account data from physiological MRIs. This data includes details of the layout of the blood vessels of the tumor and tumor angiogenesis, as well as the supply of oxygen to the tumor tissue.

Clinical significance

Accurate classification of brain tumors, a feat that even MRIs find difficult, would be invaluable to every clinician. Usually, due to the non-specificity of the symptoms associated with brain tumors, these neoplasms have often been misdiagnosed as other diseases like Alzheimer’s disease, encephalitis, meningitis, migraines, or even headaches.

By detecting and clearly classifying brain tumors, radiophysionomics would offer a clear diagnosis of the tumors. However, although radiophysionomics offers an easy way of classifying brain tumors, the team’s leader, Prof. Andreas Stadlbauer clearly states that this technique aims to supplement and not replace the clinician’s diagnosis. The study even proved that although Artificial Intelligence was superior in aspects such as precise and accurate classification, human assessments gave better results in matters regarding specificity and sensitivity of the assessment.

Conclusion

Brain tumors are known for being the most commonly misdiagnosed tumors in humans. However, from the results obtained using radiophysionomics, the odds of misdiagnosing brain tumors could be drastically reduced. This makes one wonder what other applications Artificial Intelligence may have in the medical field.