Surgeons can quickly analyse brain tumour mutations with a new AI technology, which, in less than 90 seconds, can check for genetic abnormalities in brain cancer tumours.
Researchers have found a way to check for genetic changes in dangerous brain tumours in under 90 seconds using artificial intelligence, which could improve the efficiency of glioma diagnosis and treatment.
Michigan Medicine’s neurosurgeons and engineers have developed DeepGlioma, an artificial intelligence (AI)-based diagnostic screening system that uses rapid imaging to analyse tumour specimens taken during surgery and detect genetic mutations more quickly. The newly created approach found mutations utilised by the World Health Organisation to establish molecular subgroups of the condition with an average accuracy of over 90 per cent in a study of more than 150 individuals with diffuse glioma, the most common and lethal primary brain tumour.
What are the symptoms?
Patients with intracranial tumours may experience different symptoms and disease stages. Some show clinical signs early in their development, whereas others have a minimal impact despite their size. One possible explanation for the diagnostic delays experienced by 30 per cent of the UK patients with brain tumours is the disease’s heterogeneous degree of presentation.
Improvements in AI will lead to more accurate early diagnosis. Because of the secretion of various tumour-specific molecules within the neoplastic microenvironment that permeates the blood-brain barrier and enters the wider circulation, all of the multiple pathologies that fall under the umbrella term “brain tumour” leave a unique fingerprint on routine blood tests. Because of their subtlety, these shifts in standard blood tests are well-suited for ML analysis. It has been observed that ML models do better than clinicians when it comes to identifying haematological problems. For example, Podnar et al. employed a machine learning algorithm to detect brain tumours at symptom onset to determine minor changes in routine blood testing.
How about an MRI scan?
Neuroimaging, including MRI, is still the gold standard for diagnosing brain tumours. A natural language processing ML algorithm created by Brown et al. read MRI brain requests and then selected the optimal MRI brain imaging sequence to produce the highest quality images for clinical usage. The influence of AI may begin long before radiological images are created. By a large margin, the ML algorithms surpassed the radiologist’s sequence selection. While radiologists typically make decisions about imaging sequences, there is room for inaccuracy in the protocoling process.
Likewise, radiologists’ time and picture interpretation are disrupted by sequence questions for radiographers during working hours. Standardising the MRI sequence protocol using ML-based sequence-determining algorithms may increase the therapeutic usefulness of the images produced. Furthermore, the researchers also found that the ML technique fared better than radiologists’ sequence selection in extremely rare situations like glioblastoma multiforme.
How do surgeons approach it?
Brain tumour patients usually start with surgery to remove as much as possible. So, it is because the boundaries between the tumour and healthy brain tissue can be more clearly defined, and the tumour itself can be more accurately diagnosed using a tumour sample taken and analysed during surgery.
However, the surgeon and patient must wait for the results of the intraoperative pathology investigation because the sample must be processed, stained, and examined by a pathologist. A recent study demonstrates that brain tumours can be accurately diagnosed during surgery in less than three minutes by combining advanced imaging technology with Artificial Intelligence (AI). The method could also accurately differentiate between tumour and healthy tissue.
Why is molecular categorisation important?
As the benefits and hazards of surgery vary across brain tumour patients based on their genetic makeup, molecular categorisation is becoming increasingly important in diagnosing and treating gliomas. For example, Astrocytomas, a diffuse glioma subtype, can gain five years with total tumour resection. However, access to molecular testing for diffuse glioma is limited and only sometimes available at brain tumour treatment centres. Furthermore, when it is accessible, the turnaround time for results, according to Hollon, can be days or even weeks.
Conclusion
Before DeepGlioma, physicians lacked a way to distinguish diffuse gliomas during surgery. Deep neural networks with stimulated Raman histology from the University of Michigan provide real-time imaging of brain cancer tissue.
Furthermore, patients with diffuse glioma have few options, even with the best quality of care. Malignant diffuse gliomas have a dismal prognosis, with a median survival period of about 18 months. Although improving treatment options for gliomas is crucial, only 10 per cent of people with the disease participate in clinical trials, and enrollment is generally restricted based on genetic subgroups.