According to a story from seas.harvard.edu, Harvard freshman Kavya Kopparapu was reading a story about the late Senator John McCain’s diagnosis with glioblastoma, a deadly brain cancer. She was stunned when she realized that the prognosis for glioblastoma had not improved in nearly three decades. After diagnosis, it is rare for a patient to survive for even a single year. Kavya could not accept that, despite all of the medical advances that have been made, that there had been no real improvements. She soon set out to utilize AI to develop a program that could map the genetic code of a glioblastoma tumor. With this knowledge, doctors could have a better understanding of what treatments could work.
Glioblastoma is a rare brain cancer. It is also the most aggressive cancer to originate in the brain. It is characterized by its rapid progression and poor response to most treatments. In most cases, the cause of glioblastoma is not known. A small number of cases evolve from another type of tumor called an astrocytoma. Risk factors for glioblastoma include genetic disorders such as Turcot syndrome and neurofibromatosis, exposure to pesticides, smoking, and a career in petroleum refining or rubber manufacture. Symptoms of glioblastoma include personality changes, headaches, memory loss, seizures, vomiting, and nausea; patients may lose consciousness in late stages. Treatment approaches include anticonvulsants, steroids, chemotherapy, radiation, and surgery. While a small number of patients can survive for several years, treatment is often ineffective, with the tumor relapsing quickly. Five year survival rate is only three percent. To learn more about glioblastoma, click here.
Using disease data from the National Institutes of Health (NIH), Kavya built an AI platform called GlioVision. This platform is able to scan glioblastoma imaging and identify the tumor’s genetic and molecular characteristics in just a few seconds with flawless accuracy. Older methods often take weeks or months. Training the algorithms meant breaking down extremely high resolution images into bite size fragments.
Nearly 40 percent of glioblastoma cases have a mutation that negates conventional approaches like chemotherapy, meaning that some patients are receiving treatment that inflicts painful side effects with next to no benefit. This highlights why genetic analysis of cancer tumors can be so valuable. The platform is currently in the midst of clinical testing. Kavya hopes to modify GlioVision so that it can be used for other forms of cancer as well.