The healthcare industry: primed to profit from AI and machine learning
Artificial Intelligence (AI) and machine learning in healthcare and life sciences holds the promise of reforming the industry. Generating and harnessing mass pools of data, the healthcare and life science industries are particularly primed to profit from the potential of AI, offering the unearthing of hidden insights in a world of unstructured data. Frost and Sullivan predicts that the health AI market, valued at US$600 million in 2014 will reach a high of US$6.2 billion by 2022.
The innovative analytics strategies behind AI and machine learning are increasingly having more impact on all stages of the life science industry and to a number of the basic processes in this sector. To name a few, the key laboratory, operational and clinical areas of which AI and machine learning are predicted to have a large impact, includes medical imaging analytics, drug discovery, clinical trials and clinical decision support.
Drug discovery: identifying promising drug molecules
With fully loaded costs per successfully commercialised molecule reaching US$2 billion, pharmaceutical companies are excited about the potential of AI to improve the identification of promising molecules at the earlier stages of the drug discovery and development process. Hence, reducing the necessity to make and screen a relevant subset of Active Pharmaceutical Ingredients, subsequently reducing the high attrition rates which are a key contributor to the escalating development prices and diminishing returns on R&D across the industry. Laboratory researchers are using AI and machine learning technology for heavy labour-intensive tasks such as target identification, drug design and compound screening, to reduce timelines and increase the likelihood of accurate identification. There are many examples where large pharma have used AI to successfully identify and accelerate new drug candidates: in May 2018, GlaxoSmithKline signed to collaborate with Cloud Pharmaceuticals on an AI-driven drug design program, whilst earlier that year Pfizer announced a collaboration with IBM Watson’s to utilise their machinelearning platform to aid the discovery of new immunology targets.
Clinical trials: improving patient recruitment times
There are numerous areas within the clinical trial process where insights driven from AI and machine learning tools have the ability to implement revolutionary changes. From improving study design and decision-making, to real-time monitoring and patient recruitment, AI has the potential to process large pools of data at varying risk-points and highlight them early. Additionally, companies could have the ability to predict performance of certain trial sites, anticipate drop outs and even predict outcomes when AI is applied to real-world evidence data. Patient recruitment and clinical trial optimisation are two primary applications of AI use-cases.
Contract Research Organisations (CROs) typically provide development and support services to pharmaceutical and biotech companies during the clinical trial process. The patient recruitment process for clinical trials can be difficult and time-consuming for both the CRO and the patient. A Cognizant report on recruitment forecasts estimated that 80% of clinical trials fail to meet enrollment timelines and one third of phase III study terminations are due to recruitment issues. Traditionally, patients may get trial recommendations from their doctor, providing their doctor is aware of an ongoing trial, or patients undertake their own independent research through ClinicalTrials.gov or patient forums. Several companies are developing AI software which can extract relevant information from a patient’s medical records and web search history and compare it with live trials which are recruiting and then suggest an appropriate match. Deep 6 AI is an example of a company which uses natural language processing to better match patients to clinical trials. Their algorithms are trained to recognise patient medical data and clinical data points which are extracted from health records and aggregated to develop a clinical profile. This clinical profile is then utilised to discover and compare populations and individual patients meeting a defined search criteria.
AI algorithms coupled with machine learning can also be used to mine various data sources like electronic health records, prescribing data and insurance claims. The resulting federated database can then be compared with patients who are currently enrolled in clinical trials to identify subgroups of patients which may be more susceptible to adverse events. The same method can be used for patient enrichment strategies whereby certain population subgroups are selected as most likely to progress to a particular disease state or to respond well to treatment. This can ultimately drive down costs and shorten development timeline while increasing the chance of a successful trial outcome.
Beyond the development process: clinical decision support and medical imaging although drug discovery and clinical studies are key areas which can be widely advanced by AI and machine learning in healthcare, the application of AI is not just limited to these areas.
Looking at the broader healthcare arena, AI is already being used to check data to detect patterns to improve analyses to provide better diagnosis and care for patients which also reduces costs.
Companies such as Ieso Digital Health, based in the UK, are using AI and machine learning to improve the accessibility, affordability and quality of healthcare for people with mental health conditions.
Through the use of an online, secure virtual therapy room which patients can access through their computer or smartphone, therapists are able to deliver cognitive behavioral therapy to patients in real-time through written conversations. Ieso’s algorithms support an outcomes-driven therapist allocation and scheduling system, whereby they assign patients to therapists most likely to delivery a meaningful clinical outcome at the lowest cost. A key feature of the app enabled by machine learning technology includes the review of verbatim transcripts, aiding mandatory therapist quality control faster than face-to-face therapy. Natural language processing analytics enable real-time monitoring of therapist protocol adherence and risk detection, whilst providing guidance to the therapist in relation to clinical decision support. Standard practice is systemised and variation in treatment is reduced, ultimately delivering better than average recovery rates for patients.
In another development, the non-profit company Sage Bionetworks has launched the AI-empowered mPower app, a free, 2-year mobile research study with the goal of understanding the progression of Parkinson’s Disease. Unlike many other conditions, Parkinson’s Disease varies significantly between patients and disease progression has an unknown mechanism of action. The app tracks physical and cognitive activities, symptom, medication, and trigger tracking, allowing patients to learn their symptoms, factors, and how these relate to specific medications. This not only allows the Sage Bionetwork team to monitor and understand unique patterns of the disease over time, but also results in patients gaining a deeper understanding of their own condition. Ultimately enabling more productive conversations with doctors and caregivers. The aim of the study is to create a powerful dataset which has the potential to progress into powerful insights through its AI technology.
The ability of AI to decode medical images has already proven a valuable ally for radiologists and pathologists to accelerate their productivity and accuracy in medical imaging. Multiple studies have indicated that AI tools can perform equally to human clinicians at identifying features in images quickly and precisely. For example, in cardiovascular abnormalities, automating the detection of deviations in commonly-ordered imaging tests, such as chest x-rays, could lead to quicker decision-making and fewer diagnostic errors.
The impact of AI and machine learning in healthcare
AI and machine learning is having and will continue to have a huge impact on all aspects of the life science and healthcare industry. Although the industry as a whole has been slower to implement these types of technologies into their processes in comparison to adjacent sectors, in 2018 we saw a shift in the attitude of key industry leaders with numerous collaborations and M&A activities focused around AI technology. From the drug development process to clinical decision support, AI and machine learning has the potential to transform every key inflexion point spanning the entire healthcare and life science sector. Still in its infancy, AI and machine learning will undoubtedly revolutionise this industry as it stands today, and in many ways which are yet to be discovered.
M&A in life sciences AI and machine learning
Linguamatics acquired by a Tier-1 CRO [Results International acted as financial advisor to Linguamatics]
In January 2019, Linguamatics was acquired by a Tier-1 CRO. Linguamatics is a leading provider of natural language processing (NLP) SaaS solutions to the life sciences and healthcare industries. In completing the acquisition, the tier-1 CRO will improve their current capabilities of uncovering insights to patient outcomes and enhancing their value-based care offerings. It’s intelligent solution generates insights from a wide range of unstructured and semi-structured data, empowering customers to efficiently integrate AI into their operations. In 2018, Linguamatics was recognised by Frost and Sullivan as an Artificial Intelligence Life Sciences Leader.
Precision Therapeutics + Helomics
In June 2018, Precision Therapeutics and Helomics agreed to merge. In doing so Precision gained access to Helomics’ artificial intelligence platform which, when combined with Helomics’ vast tumour database of over 149,000 patient cancer tumours, can produce actionable insights to help Precision’s TumorGenesis subsidiary develop patient-derived tumour models much more efficiently.
Genae + Hilbert Paradox
Medical device CRO Genae bought Hilbert Paradox (HPX), a digital health data management platform in March 2018. HPX’s platform enables (1) isolated digital health data silos to be captured and integrated using data analytics and AI, and (2) the processing of large volumes of data generated from genomics, diagnostic devices, biosensors and wearables to accelerate research.
Download The Bulletin: Issue 71 here