Alphabet’s “X Labs” published a new blog post Monday about the Project Amber project it has been working on for the past three years — and it now opens up its results for other mental health researchers to learn from and hopes to build on. The project is trying to identify a specific biomarker of depression — it didn’t accomplish the task (researchers now believe that a single biomarker for depression and anxiety probably doesn’t exist), but Lab X still hopes its work using electro-encephalograms (EEG) in conjunction with machine learning to try to find a biomarker will help others.
Researchers at Lab X believe that depression, like other diseases, may have a clear biomarker that helps health care providers diagnose depression more easily and objectively, and then hopefully make it easier and more sustained to treat it. In terms of electro-encephalograms, there are precedents for depression patients appearing to have been exhibiting lower EET activity in order to effectively “win” games through research using specially designed games in the laboratory.
These studies seem to offer a path to potential biomarkers, but in order for them to be truly useful in real-world diagnostic environments, such as clinics or public health laboratories, the team at Lab X began to improve the process of collecting and interpreting electro-encephalograms to make them easier for users and technicians to understand.
Perhaps most notable about this quest, and alphabet’s monday post detailing its research, is that it’s essentially a story of years of unsuccessful research — not the side of an unlikely project you’d normally hear from a big tech company.
The X Labs team summed up what it had learned from years of research into three key points about its user research, each of which touched to some extent on the shortcomings of purely objective biomarker detection methods (even if it had worked), especially when it came to mental illness. Here’s what the researchers say:
1. Mental health measurement is still an open question. Although there are many mental health surveys and scales, they are not widely used, especially in primary health care and counselling settings. The reasons go from burden (“I don’t have time to do this”) to doubt (“The usage meter is no better than using my clinical judgment”) to a lack of trust (“I don’t think my clients filled this out” and “I don’t want to reveal so much to my consultants”). These findings are consistent with the literature on measurement-based mental health care. Any new measurement tool must overcome these barriers and create clear value for people and clinicians with life experience.
2. The combination of subjective and objective data is valuable. People with life experience and clinicians welcome the introduction of objective measures, but they are not a substitute for subjective assessment and asking people about their experiences and feelings. The combination of subjective and objective indicators is considered particularly powerful. Objective indicators may validate subjective experience, or if the two disagree, this in itself is an interesting insight that provides a starting point for dialogue.
3. The new measurement technology has a variety of use cases. Our initial assumption was that clinicians might use “brain wave testing” as a diagnostic aid. However, the concept received a lukewarm response. Mental health professionals, such as psychiatrists and clinical psychologists, are confident in their ability to diagnose through clinical interviews. Primary care physicians believe that EEG testing may be useful, but only if it must be performed by a medical assistant prior to consultation with the patient, similar to a blood pressure test. Counsellors and social workers do not make diagnoses in their practice and are therefore not relevant to them. Some people with life experience don’t like to be labeled depression by machines. In contrast, there is a clear and strong interest in using technology as a tool for continuous monitoring — capturing changes in mental health over time — to understand what happens between visits. Many clinicians ask if they can send the electro-encephalogram system home so that their patients and clients can repeat the tests themselves. They are also very interested in the potential predictability of electro-encephalograms, such as predicting who might become more depressed in the future. More research is needed to determine how best to deploy tools such as EEG in clinical and consulting environments, including how to combine them with other measurement techniques such as digital esomerization.
X Labs is working on open source Amber’s hardware and software on GitHub, and has also issued a “patent commitment” to ensure that X Labs does not take any legal action against users of Amber’s associated EET patents using open source materials. What is not clear is that if Amber succeeds in finding a single biomarker of depression, it will be a result, but perhaps in the hands of the wider community, the team’s work in making EETs more easily outside professional testing institutions will lead to other interesting findings.