• AI
  • Medical imaging’s the big gainer from AI and ML

    AI’s not new. It emerged in the 1960s. A blog from PLOS Speaking of Medicine says advertising hyperbole has led to scepticism and misunderstanding of what’s possible with machine learning (ML) and what’s not with. The blog sets about providing an accessible, scientifically and technologically accurate portrayal of ML’s current state in clinical translation.

    Medical imaging workflows are seen as benefiting most in the short-term. ML algorithms automatically processing two or three-dimensional scans to identify clinical signs of conditions, such as tumours and lesions, and determining likely diagnoses have been published. Some are progressing through regulatory steps toward the market. 

    Many use deep learning. It’s a form of ML based on layered representations of variables, ML’s neural networks. It’s benefited ophthalmology. A major UK eye hospital has used deep learning to deal with a clinically-heterogeneous set of three dimensional optical Coherence Tomography (CT) scans. Referral recommendations reached or exceeded experts’ decisions.

    Radiologic diagnoses are another ML beneficiary. An algorithm detected 14 clinically important pathologies from frontal-view chest radiographs. They included:

    PneumoniaPleural effusionPulmonary masses and nodules.

    ML’s performance matched practicing radiologists. Another 

    There’s several other clinical activities where ML can benefit healthcare. They include: 

    Triage and preventionClustering for discovery of disease sub-typesAnomaly detection to reduce medication errorsAugmented doctors.

    The blog’s an advance report. Its final version’ll be in PLOS Medicine at the end of December. It’s a valuable guide for Africa’s health systems’ eHealth strategies. An initial step’s to lay down data foundations.

  • Babylon’s AI is embedded in Rwanda’s primary care, and other countries

    Succeeding with UHC depends on extra healthcare resources. It depends on efficient and effective use of resources too. eHealth’s part of the solution. 

    Babylon Health uses AI to improve access to primary care. It’s planning to expand into chronic care. In England, it’s restricted geographically to London.  Regulations seem to prevent Babylon’s AI from providing diagnoses. Instead, it provides health information.  This could change as hard evidence becomes available.

    A review, reported in Digital Health, says Babylon’s AI claims lack convincing evidence. Babylon Health doesn’t agree. A report claims its AI beat human doctors’ average score of 72% in a range of 64% to 94%. Babylon scored 81% in a Royal College of General Practitioners (RCGP) exam using a representative sample of questions from the final assessment for GPs in training. The results have not been peer-reviewed.

    In Rwanda, it’s called Babyl. It uses AI to provide:

    Consultations with doctors and nursesLab testsPrescriptionsReferrals.

    It’s a core UHC component for the country. Where access isn’t feasible with conventional healthcare, Babyl seems crucial in meeting health and healthcare needs.

    Babylon’s AI uses machine learning created by a team of research scientists, engineers, doctors and epidemiologists. They have access to large data volumes from the medical community. Learning’s continuous through feedback from Babylon’s experts.

    It comprises: 

    A knowledge graphA user graphAn inference engineNatural Language Processing (NLP). 

    It seems like a solution for all Africa. On its own, it may not be enough. Increasing referrals may need investment in extra healthcare capacity.

  • AI, blockchain, cold chain and motorbikes improve blood donations and save lives in Nigeria

    Blood shortages are common in many health systems. An initiative in Nigeria uses mHealth to create a community of voluntary blood donors, and connects hospitals with blood banks, and blood banks with donors. Life Bank, a Lagos start-up also provides a discovery platform on for hospitals to order blood

    LifeBank delivers requested blood in less than 45 minutes, in a WHO Blood Transfusion Safety compliant cold chain. An article in Disrupt Africa says it’ll add other medical products such as oxygen, vaccines and rare drugs to its services.

    Its founder, Giwa-Tubosun, began a non-profit service to encourage people to donate blood. She then moved on to address supply shortages and poor logistics. Two main goals are:

    Increasing access to bloodReducing the number of Nigerian women who die from birth complications.

    LifeBank’s resources include:

    AIBlockchainCold chainmHealthMotorbikes.

    These combine to provide information about blood availability and avoid health workers’ wasted time and frustration seeking blood products. They also minimise ineffective blood transports that result in bacteria proliferation and consequences of health complications.

    Supporters include:

    Co-Creation Hub (CcHub) in 2016 that raised pre-seed fundingEchoVC Partners, a venture capitalistParticipation in Merck’s Lagos-based satellite accelerator this yearSelection for MIT Solv2018 that added grants and access to other resources.

    Its impact is considerable. To date, LifeBank’s delivered some 11,000 products for over 400 hospitals. Over 6,300 people are registered as voluntary blood donors, with over 20% donating blood in the last two years. The result: over 2,100 lives saved.

    A challenge is convincing blood bank partners to use LifeBank. As this is  overcome, it’s it easy to envisage LifeBank eventually operating across Africa.

  • How far into the future should eHealth strategies look?

    By definition, eHealth strategies are about investing in the future. They’re also about taking existing eHealth investments forward, either by switching, enhancing and rolling out further. In 2006, Rosabeth Kanter identified several lesson for innovation strategies. They included an “innovation pyramid” where:

    Not every innovation idea has to be a blockbusterSufficient numbers of small or incremental innovations can lead to big gainsBig bets at the top that get most of the investmentA portfolio of promising midrange ideas in test stageA broad base of early stage ideas or incremental innovations.

    The last one’s relevant for a perspective set out in an eBook from Oracle. Technology Takes Healthcare to Next Level proposes strategies for disruptive technologies of:

    AIBlockchainChatbotsIoT. 

    Each one offers promise for healthcare. Combined, Oracle sees the sum of the parts as greater than the whole. Combining blockchain and IoT allows frictionless data exchange. AI and machine learning put data in motion with minimal human intervention. AI tools can study blockchain’s large volumes of data to find patterns that need responses

    For Africa’s health systems, investment in ICT foundations and patients’ clinical and demographic data’s needed to. The strategic challenge is to choose between sequential investment and progress in an innovation pyramid where these four technologies start their journey. While leaving the disruptive technologies into the future, it can defer the costs. It will also defer the benefits.

     

  • Robots could be good for your health

    In his book The Rise of the Robots, published by One World, Martin Ford proposes social and economic scenarios for robots that are good for output, but not so good for people. He sees significant upheaval and displacement from employment across a wide range of commercial and industrial activities and across middle and low income families. The drop in income, so spending power, will degrade economies.

    Simultaneously, robots aren’t paid and don’t spend money. He sees this as exacerbating the social and economic impact.

    Healthcare’s the activity that’s different. He sees the robots marching into healthcare that’s already over-stretched as needs and demands continuously outstrip supply. Four roles are crucial:

    Artificial intelligence in medicineHospital and pharmacy roboticsRobots that care for the elderlyUnleashing the power of data. 

    For low and middle income countries and health systems, sustained investment in robots could be part of the solution. They can improve healthcare professionals’ productivity and help to meet demand.

    They should find a place in Africa’s eHealth strategies. Small scale investment will lay out a trajectory for the future.

  • AI helps to predict cancers’ trajectories

    Many years ago, people in the UK referred to cancers as “a growth.” While it might have lacked scientific precision, it encapsulated cancers’ changing characteristics. The country’s Institute of Cancer Research (ICR) at London’s Royal Marsden Hospital, and part of University College London (UCL), says tumours’ constantly changing nature’s one of the biggest challenges in treating cancer, especially when they evolve into drug-resistant forms.

    It reports that its ICR scientists, working with colleagues at Edinburgh University have used AI to identify patterns in DNA mutations in cancers. The information can forecast future genetic changes to predict how cancers will progress and evolve. The technique, Repeated Evolution of Cancer (REVOLVER), predicts cancers’ next moves so doctors can monitor tumour’s progress and design the most effective treatment for each patient. 

    Three organisations financed the research, published in Nature Methods. They were the Wellcome Trust, the European Research Council and Cancer Research UK. Their support for REVOLVER’s created what’s seen as a powerful AI tool. It’s revealed previously hidden mutation patterns located in complex data sets.

    Teams from ICR and the University of Edinburgh working with colleagues from the  Birmingham University, Stanford University and  Queen Mary Universities London found a link between some sequences of repeated tumour mutations and survival outcomes. It suggests that repeating patterns of DNA mutations could be prognoses indicators. This can help to specify future treatment.

    AI success stories provide material to consider in Africa’s new eHealth strategies, to support leading specialist hospitals to set up a wide range of AI initiatives. They could focus on Africa’s current and emerging health and healthcare priorities.

  • Will AI and Blockchain converge to enhance health analytics?

    While AI and Blockchain are seen by some to offer powerful tools, a view’s emerging that combining them offers significantly more potential for Big Data and health analytics. Or, is it just another dose of eHealth hype? An article in Health IT Analytics  says in the US, AI and Blockchain are now tools of choice for developers, providers and payers in improving their eHealth infrastructure.

    But, it acknowledges that both are near their hype curves peaks. Some providers and payers are reluctant to invest heavily at their maturity stages. Concerns over security, utility and Return on Investment (ROI) are justifications for some organisations to defer investment, leaving others to provide evidence that combining AI and Blockchain can succeed in secure the large data sets and exchanges that Big Data needs for innovative analytics.

    Access to data’s one obstacle. Most data resources are held securely and privately by several institutions. Opening them can create cybersecurity vulnerabilities. Despite this, ideas are fermenting of using Blockchain to produce metadata about the datasets available at several organisations. It can also provide secure, peer-to-peer data exchange. Blockchain can be a pointer to where full data sets are stored, allowing for discoverability without requiring data sets to move each time a transactions completed.

    This strategy enables organisations to keep sensitive data, such as Protected Health Information (PHI) and Personally Identifiable Information (PII) off Blockchain. It’ll reduce risks of breaches. Instead, minimal but sufficient data should be held in Blockchain.

    These comprise complex decisions and projects. It seems premature for Africa’s health systems to pursue combined AI and Blockchain strategies in the medium term. There are other eHealth priorities to address, such as using mHealth to support remote health workers with access to test results and improving their co-ordination with colleagues.

    If the AI and Blockchain are converging in healthcare, Africa’s health systems can watch trajectories and learn from them. If they deliver a significant proportion of their potential, a challenge for Africa’s health systems may be to avoid a sudden disruption to their eHealth strategies and plans. While this can be costly, missing new eHealth opportunities has a cost too, often of missed benefits. 

  • AI passes a stiff test at London’s Moorfields Eye Hospital

    England’s Grand National run at Aintree is gruelling. It has 30 fences, two with open ditches, in a distance of 2.25 miles that’s completed twice. AI has just moved up the field in the eHealth equivalent. 

    An AI project at London’s Moorfields Eye Hospital with Google’s DeepMind has accurately diagnosed eye conditions from scans. As ophthalmologists’ workloads and their complexities increase, diagnostic imaging is expanding faster than specialists can interpret the results. AI already has a constructive reputation in classifying two-dimensional photographs of some common diseases it’s reached the performance of expert clinicians in a real-world clinical pathway with three-dimensional diagnostic scans. 

    At Moorfields, a novel, deep learning architecture is now applied to a clinically heterogeneous set of three-dimensional optical coherence tomography scans from patients. The research found that after training on 14,884 scans, AI’s referral recommendations of sight-threatening retinal diseases reached, and sometimes exceeded that of experts. 

    Other benefits include:

    Tissue segmentations produced by the architecture are device-independent representationsReferral accuracy’s maintained when using tissue segmentations from a different devicePrevious barriers to wider clinical use without prohibitive training data requirements across several pathologies have been removed.

    After training, the algorithm assigned diagnoses to 1,000 patients’ scans whose clinical outcomes were already known. The same scans were shown to eight clinicians. Four were leading ophthalmologists, four were optometrists. They classified the diagnoses into four referral types,  urgent, semi-urgent, routine and observation. AI performed as well as two of the world's leading retina specialists. The error rate was 5.5%. More strikingly within this performance, the algorithm didn’t miss any urgent cases.

    The impact of the project’s global. For Africa’s health systems, the challenge’s entering the AI Grand National and making sure they don’t fall at any of the daunting fences. It offers an eHealth strategic scenario that extends what is now relatively conventional EHRs and mHealth. AI can extract more value from them than originally imagined.

  • AI needs faster data access for researchers and analysts

    Maximising AI’s potential for clinical research and breakthroughs needs access to large data volumes to train then deploy AI models. A white paper by International Data Corporation (IDC), sponsored by: Pure Storage, says Hard Disk Drives (HDD) are too slow for the task. It says All-Flash Arrays (AFAs) are faster and more accurate. 

    An AFA’s a Solid State Disks (SSD) storage system with several flash memory drives. Instead of searching for data on spinning HDDs, SSDs have no moving parts, so are faster to access. The Tech Republic has an entry-level guide on AFAs. It says they’re disrupting traditional data storage resources. 

    IDC’s white paper emphasises AI as a learning process where researchers and analysts need prompt access to data for clinical projects. It has two main benefits:

    Shortens the clinical innovation time from desk to bedsideAttracting and retaining scarce clinical researchers and data scientists who look for leading-edge AI investment and infrastructure to succeed.

    Improved data response times with AFA benefits clinical teams that need access to clinical data for direct patient care too. Faster response times help to improve their productivity and efficiency. They also help to minimise eHealth frustrations and improve job satisfaction. 

    As eHealth foundations are vital parts of eHealth strategies, Africa’s health systems should consider SSDs along with expanding network capacity and connectivity capacity.  

  • Stethee reinvents the stethoscope with AI

    The worlds first Al enabled stethoscope system has been launched by M3DICINE Inc.

    The design itself is revolutionary and operates as easily as the traditional stethoscope. However, it allows users to listen to the lung and heart sounds with a more sophisticated amplification and filtering technology. Heart and respiratory sounds captured are sent via Bluetooth to the Stethee Android or iOS App which enables a wider range of diagnostic capabilities.

    The Stethee system comes in three core products:

    FDA cleared Stethee Pro for medical and healthcare professionalsStethee Vet for veterinarians and animal professionalsStethee Edu developed specifically as an education and research tool

    The technology platform behind the Stethee AI engine , named “Aida” can analyze the heart and lung sounds to build a unique personal biometric signature.  In addition to this, Aida automatically tags geo-location and environment data to each sample in real time.  This offers a completely new dimension of data analytics for public health planning by allowing one to understand what effects environmental factors such as pollution, temperature or humidity have on our heart and lungs.

    Aida also analyzes this encrypted and anonymised data in order to learn and report back quantitative clinically actionable data to vets, doctors and other healthcare professionals. Not only does it identify and analyze heart sounds and respiratory activity but also patterns that may indicate a disease condition. The data is represented in real time in the Stethee App, therefore making it easy to understand vital signs.

    The potential for the Stethee to be used in remote rural areas is quite vast because its relatively easy to use and results can be shared and analyzed promptly by a medical specialist anywhere in the world. This is invaluable to the improvement of patient care, more especially to remote rural areas where access to screening services or a cardiologist is very difficult.