Like many other pharma groups Boehringer Ingelheim has jumped on the digital bandwagon. And one of its most promising candidates, known as pathological speech processing (PSP), is on the verge of going into clinical trials less than two years after starting development – neatly demonstrating why enthusiasm for such projects is running high.
PSP uses machine learning to analyse speech patterns to detect brain disorders like schizophrenia and Alzheimer’s disease, and is one of the first contenders to emerge from Boehringer’s BI X digital subsidiary.
The company highlighted other BI X-originated projects during a press trip last week, including the Adam platform for drug design, and the Brass pharmacovigilance system. But according to Heiko Schmidt, the unit’s chief operating officer, PSP is so far the most tangible potential product to originate from the digital lab, which was set up in June 2017 (Interview – Boehringer does digital health, San Francisco style, 26 April 2018).
Right now, PSP is being used as an R&D tool, Allan Hillgrove, head of Boehringer’s human pharma unit, said during a tour of the German company’s digital facilities. But he added that the group’s ultimate aim was to develop PSP as a diagnostic for the general population.
For this to become reality Boehringer would need to carry out prospective clinical studies – and here the company is working on a trial protocol, a spokesperson told Vantage. He would not say when the first clinical study might start, or what disorder Boehringer initially planned to investigate, but much of the group’s presentation focused on PSP’s potential in schizophrenia, so this could be a strong contender.
Unintelligible speech is one of the symptoms of schizophrenia, but Boehringer is using PSP to look for more subtle changes in language in a bid to diagnose the disease earlier and/or identify patients who are about to relapse.
For example, the company is using a method called graph analysis, which evaluates the order in which words are put together in sentences. There are differences in the speech graphs of healthy people versus those with schizophrenia or bipolar disorder, explained Michael Sand, the leader of Boehringer’s CNS division.
“The computer is able to monitor subtle differences in these graphs, and these are very closely correlated with clinical symptoms,” he said.
The company is also using artificial intelligence to evaluate the way people process the meaning of words. Mr Sand pointed to one such analysis in which five out of 34 at-risk people were predicted to go on to develop schizophrenia within 2.5 years; all five did, and the other 29 patients did not. “That very strongly outperforms clinicians’ ability,” he said.
Finally, Boehringer is looking at acoustic markers including “the melody of the language, intonation, the rate and rhythm of speech [and] the length of pauses between words and sentences”, Mr Sand said.
As well as schizophrenia, PSP could be useful in Alzheimer’s, depression and indeed “any brain disease”, according to Mr Schmidt. Earlier diagnosis could lead to earlier treatment – but clearly, in disorders for which there is currently no cure, like Alzheimer’s, the use of PSP could throw up some ethical issues.
Boehringer has so far invested less than €20m ($22.5m) per year in BI X, although its finance chief, Michael Schmelmer, says it is willing to put in more. There is still a long way to go with PSP, but the project’s speedy development so far suggests that this could have been money well spent, and there might soon be more to come from the company’s digital drive.