HOW MACHINES WILL HELP NEUROSURGEONS TACKLE BRAIN DISEASE

MQ Health has established the first Computational NeuroSurgery Laboratory in the world to focus on developing computerised analysis tools in neuroimaging and neuropathology that will improve the diagnostic accuracy of brain disease.

The diagnosis of brain tumours and other diseases is heavily dependent on neuroimaging – in particular MRI. However, the large amount of radiological data generated by MR sequences can be overwhelming to interpret and made more difficulty by the possibility of conditions such as ‘mimic’ tumours, inflammatory disease that can resemble a tumour.

The implementation of multidisciplinary teams (MDTs) has helped significantly to increase accuracy of diagnosis, estimates of prognosis and enhanced decision-making around whether to proceed to surgery.

Now, MQ Health is looking to add the use of Artificial Intelligence (AI) tools to brain disease diagnosis through its world-first Computational NeuroSurgery (CNS) Laboratory. Led by Associate Professor Antonio Di Ieva, the team is investigating the use of computerised analysis tools to aid surgeon-MDT-based evaluations of radiological images – also the subject of Associate Professor Di Ieva’s letter in the November 2019 issue of The Lancet.

“Our method is to develop novel diagnostic, prognostic and therapeutic markers of disease, which can then be applied in the development of AI algorithms,” said Associate Professor Di Ieva, who received the 2019 John Mitchell Crouch Fellowship from the Royal Australasian College of Surgeons to complete the first year of research required. “The long-term goal is to enhance treatment and outcomes for patients.”

The first stage of work began earlier this year, with researchers characterising the ‘fingerprint’ of the brain – the highly detailed architecture of the brain in its entire physio-pathological spectrum, from the normal to the diseased.

“The work involves extracting features from pathology slides and MR images of brain disease in order to objectively compare the pattern expressed in different physio-pathological states,” explained Associate Professor Di Ieva.

“We are doing this using fractal and machine-learning methods, thanks to the expertise of computer scientists, including Dr Carlo Russo, Research Associate at the CNS Lab, and Dr Sidong Liu, research fellow from the Australian Institute of Health Innovation at Macquarie University.

“We are also capturing data on the cognitive processes of surgeons as they review imaging data and identify relevant features of an image to diagnose and develop a treatment plan for a patient.”

Data will then be transferred to a computer in order to ‘teach’ the machine to extract features and characterise patterns of brain disease in the same way a surgeon would through complex algorithms developed.

“The aim is to support, not to replace, clinicians in diagnosis and decision-making by parameters confirming or refuting their diagnostic hypothesis,” said Associate Professor Di Ieva.

The CNS Laboratory at MQ Health builds on Associate Professor Di Ieva’s pioneering application of computational fractal-based analysis to the quantification of features in gliomas completed as his PhD research in Austria, in 2011, as well as his use of fractal geometry to study brain cancer and other diseases of neurosurgical interest.

This earlier work led to the successful use of computational fractal-based modelling to objectively quantify the patterns of brain and pituitary tumours as well as arteriovenous malformations, and to predict the response to treatment of patients affected by AVMs undergoing Gamma Knife radiosurgery treatment.