Belong.life, a global provider of digital healthcare platforms has launched a world-first AI tool for oncology. Health Tech World hears from the company CTO about its game-changing potential to automate and streamline cancer care.
Nearly two million new cancer cases are projected to occur in the US in 2023 alone, while immediate access to specialist care teams can be limited, especially among rural populations.
In an effort to bring a new level of support to cancer patients, last month, Belong.life launched Dave, the world’s first conversational AI oncology mentor.
The AI tool provides oncology-specific answers to cancer patients’ questions, communicating instantaneously with cancer patients.
With its empathetic tone and an ability to retain long-term memory of previous correspondence, the tool aims to assist clinicians in guiding people through their cancer journey.
One early adopter of the platform is Mark. His cancer journey began in 2018 with a solitary plasmacytoma tumour.
He was declared cancer free after a course of radiation, but a few months later, he was hit with a bad case of neuralgia across his face and scalp.
He was diagnosed with multiple myeloma with 25 tumours in his spine and head. A week later, this number more than doubled to 58.
“My best friend came to visit me,” Mark told Health Tech World.
“He said later that when he saw me, he thought I was a goner and wouldn’t last another week. He thought he was going to have to start working on my eulogy.”
“About this time I discovered and joined the Belong app. I shared my story and immediately received many supportive replies.
“I became hooked [on] the app and decided to help others while sharing my journey. I eventually became more of a mentor.”
“Dave came onboard just before I went through my CAR-T cell infusion. This technology was new to me and Dave helped me become knowledgeable about it and even its success rate.
“Now that Dave is around, I really don’t have to search for answers. Dave’s answers are immediate.
“Before, on the app, if someone had a technical question for a group doctor, it could take hours or days to get an answer, as the doctors had their own stuff to do.
“Dave is there to take a large load off of their shoulders.”
The Dave platform is built and finetuned using machine learning (ML), LLM classification and NLP and was trained on datasets of unique patient-physician and patient-to-patient interactions.
The data was aggregated from Belong’s Beating Cancer Together app which brings together seven years of proprietary interactions and patient journeys on the app, translating to billions of data points.
This gave the platform a comprehensive understanding of the cancer journey, the company said.
Prior to launching Belong.Life seven years ago, the company’s CTO, Irad Deutsch, and his fellow co-founders had built three successful start ups in the big data and analytics industry.
But after his mother was diagnosed with cancer a decade ago, he was inspired to bring his expertise to the healthcare sector.
“We never thought about doing anything in the healthcare space until about 10 years ago,” Deutsch told Health Tech World.
Irad Deutsch
“After she passed away, I shared my experience with my colleagues and then everybody started to share their challenges with their families within different healthcare conditions.
“We said ‘Hey, how come there aren’t any good tools that will help patients in their journey to make educated decisions’.
“So then we came up with Belong and we all shifted into this industry.”
Dave is available to patients as a free service through Belong’s Beating Cancer Together app, where users can also communicate with other patients and with cancer care professionals.
The tool is also available as a SaaS solution for hospitals, providers and patient support programs, which can customise the mentor characteristics and fields of expertise to include in-house clinical guidelines.
“We have live physicians on the platform,” Deutsch said. “Real experts from all around the world answering questions.
“They do not provide medical advice; of course, we cannot do this, but we provide educational support and explain to them how decisions are being made [ to] empower them.
“We have peer-to-peer; we have patients talking to other patients and we have technology that creates a small circle of interest because we believe there’s no two patients alike on the planet.
“Everybody experienced things differently. So, we want to connect patients that are on the same journey together.”
The platform achieves this through more than 3000 machine learning classifiers that analyse the data that patients upload to the platform; all done anonymously.
“Every chat, every post, every document they upload, we analyse how we can connect them with patients very similar to them,” Deutsch added.
For Mark, this peer-to-peer support was invaluable and the introduction of Dave to the platform has added an additional layer of support as he navigates his complex course of treatment.
For patients at the beginning of their journey, Mark believes Dave could provide them with much-needed comfort after the shock of their diagnosis.
A diagnosis comes with a myriad of often distressing questions that go unanswered until patients visit their oncologist. Dave offers a place to turn to in these interim periods.
“So often, cancer patients can become overwhelmed when their doctor shares their diagnosis,” Mark said.
“Having Dave available to ask questions is important as knowledge is power and comfort, and Dave provides it in a sympathetic manner, which is really cool.”
Dave has already been tested by more than 10,000 people with cancer, who have praised it for providing oncologist-like support and empathetic care.
This empathetic tone of voice is learned via patient-to-patient data points while its technical knowledge is derived from doctor-to-patient interactions.
The company also uses medical documents to train the AI model to understand doctor visit summaries.
“People upload blood test results and Dave can tell them a lot of things about the figures,” Deutsch said.
“We also have PROs [patient reported outcomes] which are surveys that we send out periodically to patients to actively collect more data throughout their journeys.”
Since his diagnosis, Mark has gone through various treatments, from chemotherapy to the immune-modulating treatment, Revlimid. Last year, his condition took a turn for the worse.
Tumour growth had attacked his T-6 vertebrate and destroyed it. Threatened with possible paralysis, he underwent further chemotherapy along with other treatments.
Recently, he was offered CAR-T cell infusion. Though not without complications, the treatment has been a success.
He is now in remission, but like many cancer patients, Mark’s journey has been long and tumultuous.
He told Health Tech World that Dave has helped him better understand his latest treatment and the risk that came with it.
“Dave came onboard just before I went through my CAR-T cell infusion,” Mark said.
“This technology was new to me and Dave helped me become knowledgeable about it and its success rate.”
An ethical approach to AI
Users of AI models like ChatGPT and Bing AI are increasingly seeing flaws in responses, known as hallucinations, where the model confidently produces incorrect information without the facts to back it.
In healthcare, this comes with obvious risks to patients. Belong.life has addressed this by using two machine learning models in the backend which work together to self-correct its own answers.
“We understand that we are in a very sensitive area,” Deutsch said. “We cannot afford to make mistakes and hallucinations like GPT does.
“So, we have a self-corrective mechanism that makes sure that it is not doing anything out of scope in terms of safety and ethics.
“We also have a reinforcement mechanism […] that analyses the patient’s response and by doing that it is able to reinforce the model.
“Bottom line what it creates is a very good experience for a user that actually feels like they are communicating with a real expert in the oncology field.”
In cases where the platform cannot provide an answer with confidence, patients are directed to a real expert on the platform.
“Dave can provide you with fast results in seconds to 80 per cent of your questions,” Deutsch continued.
“But then if you have a tougher question, that can go to a real doctor because there is no replacement for a real doctor.”
Having used the platform since its launch earlier this year, Mark said he is impressed with the accuracy of Dave’s responses, describing it as an “amazing resource” and source of comfort for people going through their cancer journey.
“When Dave was first introduced, I tested him with some complicated issues that I already knew the answers to and Dave was spot on,” he said.
“I am really excited that Belong added Dave to the platform. The algorithm is so advanced that people forget they’re talking to a computer.”
Belong.life is now training models for additional AI mentors for neurological conditions, MS and obesity.
More broadly, Deutsch sees AI tools like Dave transforming healthcare services by boosting efficiency and reducing the labour burden on clinicians.
He compares it to an autonomous car.
“In 10 years from now you will have an autonomous service with a doctor holding the wheel, making sure that you’re on the right path,” he said.
