BrainSpeech: Speech as biomarker in brain cancer v1

• Research type

  Research Study

• Full title

  Assessing Speech as a prognostic biomarker in Patients with Primary and Metastatic Cancer using Machine Learning A Phase II feasibility study

• IRAS ID

  276803

• Contact name

  Matthew Williams

• Contact email

  matt.williams3@nhs.net

• Sponsor organisation

  Imperial College London

• Clinicaltrials.gov Identifier

  37487 (pending), ISRCTN submission

• Duration of Study in the UK

  3 years, 11 months, 26 days

• Research summary

  Summary of Research
  Primary and secondary brain tumours are a large source of disability and death. They are difficult to treat and prognosis remains poor with few innovations in the last decade. This population suffers from a dual problem of diagnostic uncertainty with Magnetic Resonance Imaging (MRI) in assessing disease status together with a high physical and cognitive disability burden requiring assistance to even attend such scans.
  To improve diagnostic yield from MRI, BrainSpeech assesses speech which is affected in a third of patients with the most common primary brain tumour at outset. Detailed speech assessment and analysis takes time to perform due to the need for transcription. However, speech is a non-invasive, near-patient modality that is both rich in information and easy to access (e.g. over phone).
  BrainSpeech attempts to overcome these issues by comparing standardised speech tests against spontaneous speech generated from a prompt (“Tell me about your cancer journey so far”). A combination of modern machine learning methods which have worked well to diagnose other neurological pathologies (Parkinson’s disease, dementia) with accuracies in excess of 90% will be deployed in order to fully analyse patient in speech in hopes of developing a new biomarker to track disease progress.
  50 patients with new or known brain cancer will be recruited and have their free speech alongside cognitive measures assessed as they progress through their treatment. Patients have the option to have their main caregiver paired as a healthy volunteer control.
  We will be able to compare these changes with standard measures of disease progression in order to measure how feasible changes in speech can be used as a marker to track the up-to-date status of their disease. The study will also use an additional form of imaging, tractography, in order to better understand why speech changes in the brain.

  Summary of Results
  Early termination of study.
  Study was approved just as COVID began.
  The study was designed to test speech in person. This was deemed unworkable in the COVID climate. No satisfactory remote workarounds were found (no standardisation of remote equipment, audio quality etc).
  Terminated prior to any patient/volunteer recruitment

• REC name

  Wales REC 4

• REC reference

  20/WA/0047

• Date of REC Opinion

  25 Mar 2020

• REC opinion

  Further Information Favourable Opinion