The TACERN Study Group has recently conducted a study in order to discover an algorithm that detects tubers during MRIs. Deep learning was employed in an effort to apply this method to rare diseases, many of which face unmet medical needs. This study was published in PLOS and is meant to help those with tuberous sclerosis.
About Tuberous Sclerosis
Tuberous sclerosis is a disorder that is characterized by the growth of noncancerous tumors throughout the body. This disorder contains a wide variety of symptoms, and it has a wide range of severity. Some people may have mild symptoms, which would lead to a diagnosis later in life, while others have very severe symptoms and are diagnosed during infancy. This disease has also been linked to a higher chance of autism.
Tuberous sclerosis is the result of inheritance or an error in cell division. A parent who also has this disorder can pass it down to a child, or a random mutation can occur that results in tuberous sclerosis. Both of these causes have to do with the TSC1 or TSC2 genes, which are the genes that result in tuberous sclerosis.
Symptoms of this disorder are a result of the tumors, which most commonly grow in the kidneys, brain, eyes, heart, lungs, and skin. The size and location of the tumors can affect the severity of the symptoms. These symptoms include skin abnormalities, seizures, cognitive disabilities, behavioral problems, kidney problems, heart issues, lung problems, and abnormalities in the eyes. Depending on the size and location of the tumors, life-threatening complications can occur. Tumors can block the flow of cerebral spine fluid, which then creates an excess of fluid in and around the brain. Tumors in the heart can cause heart issues like dysrhythmia. Lung failure is a possibility if tumors grow in the lungs, and vision damage can happen if they grow in the eyes. There is also a higher possibility of developing malignant tumors in the brain or kidneys.
In order to reach a diagnosis, a physical examination and a discussion of symptoms and family history must occur. Doctors will look for the benign tumors that are associated with tuberous sclerosis. Tests will be ordered in order to locate and assess the severity of the tumors. Genetic tests may be ordered as well. Specialists will often be ordered for the effected organs.
There is no cure for this disorder, but there are treatments available to manage the symptoms. Medications can be prescribed to help many of the symptoms, including seizures. Surgeries can remove growths if they become harmful. Speech, occupational, and physical therapy may also help if the tumors cause issues in these areas. As tumors in the brain can cause developmental delays, educational services and psychiatric or behavior management may also be helpful.
About the Study
228 participants were enrolled in this study, half of which were affected by tuberous sclerosis. Researchers then took MRIs and extracted FLAIR and T2 axial images in order to detect cortical tubers. Using convolutional nueral network (CNN) architectures, they were able to create an algorithm that would automatically detect tubers when patients go in for an MRI. This algorithm has a sensitivity and specificity of .95.
These results have proven that deep learning can be applied to neurological disorders. Researchers hope that other rare disorders can be helped by deep learning as well.