Connect with us


The power of machine learning for diagnosis, treatment and preventive healthcare

By Steve Pogrebivsky, president and co-founder of DataGroomr



Machine learning

“There’s a real compelling case to be made for technology helping people be smarter about people,” says Rosalind Picard, Sc.D., founder and director of the Affective Computing research group at the MIT Media Lab.

If that technology is machine learning (ML), that insight might even apply to doctors. At the very least, ML can augment the intelligence of doctors.

Artificial intelligence (AI), and ML in particular, has already started to revolutionise medicine across all three areas that define healthcare: diagnosis, treatment, and disease prevention. Take a look at the emerging landscape thanks to ML:

  • Sustained behavioural change—the holy grail of preventive medicine—is now feasible with ML-driven personalised nudges. ML can learn and predict the most effective nudges for any given patient who needs to improve health behaviors like exercise, nutrition and medication adherence. In this way, ML can improve healthcare delivery and even clinical outcomes by focusing on preventive health interventions.
  • Expedited therapeutic development by using ML for drug repurposing. This approach, which identifies new uses for existing drugs, can swiftly address challenging diseases and emerging pandemics, as The Lancet reported with the arrival of COVID-19.
  • Personalised neuromodulation therapy that precisely targets a patient’s neural networks. Increasing research shows that neural networks are highly individualised. Thanks to ML and large-scale brain scan data (specifically, functional MRI), neuromodulation may soon target the exact source of depression, for instance, or the specific executive function weakness in the frontal lobe responsible for attention problems. This is just one example of how ML may play a role in precision medicine.

Let’s break this down to see how prevention, diagnosis, and treatment may benefit from ML.


It’s well recognised that behaviuoral and social determinants of health are the main forces supporting our health.

Factors like diet, exercise, mental wellness and access to care account for roughly 60 per cent of our health, compared to about 30 per cent for genetic input, according to a recent study on the role of AI in precision medicine, published in Clinical and Translational Science.

Steven Pogrebivsky

While insurers, large companies, clinicians and public health experts have long emphasised prevention in healthcare, behavioural determinants–namely, habits–are challenging for people to change in a sustained way.

Machine learning promises to conquer that challenge through a variety of ways including:

  • Behavioural phenotypes”–with ML, health interventions can be tailored for each patient based on personality traits like extroversion and UX patterns like optimal timing during the day, points out a report in the Harvard Business Review.
  • Detection of subtle changes–physiologic changes may be imperceptible to patients and their families but ML leverages wearables and other digital medical devices to convey important health indicators.
    With patients’ permission, these health indicators can be transmitted to healthcare providers.
    Currently, wearables track heart rhythm, blood pressure, oxygen saturation, sleep habits, and other indicators.
    Verbal signals like voice inflections and non-verbal signals like facial expressions—behavioural biomarkers—may soon be added to the dataset that ML uses to identify patterns and anomalies and to generate insights and tailored health interventions.
  • Deep personalisation–patient-centred nudging is becoming powerful thanks to ML.
    Hyper-personalisation is “the lifeblood of customer loyalty,” says Salesforce’s Executive VP Neeracha Taychakhoonavudh, and in healthcare it’s key to real and lasting patient engagement.
    A healthcare platform that uses ML is constantly adjusting to patients’ needs, syncing with those needs and anticipating them, and tailoring the most effective health interventions for sustained behavioural change to improve health outcomes.
    All of this is accomplished without the need for excessive data input by the user since ML, by definition, constantly learns from, adapts to, and responds to the user.
    One key point to note, however, is that clean data is central to the success of deep personalisation because clean data is necessary for patient trust.
    For instance, duplicate records can vex an Electronic Health Records system and undermine patient trust; to clear up that problem and maintain patient trust, health systems can use a ML-driven deduplication app like DataGroomr, which cleanses Salesforce data continually with minimal human intervention.

A few AI platforms featuring deep personalisation include:

AllazoHealth, which predicts and aims to prevent non-compliance by analysing a large dataset and the patient’s own daily habits.

Lirio, which leverages insights from behavioural scientists to curate content for an individual user based on his or her needs. Hyper-personalised messaging is Lirio’s goal in order to nudge patients toward timely screenings and to optimise engagement.

WoeBot Health is a mental health chatbot designed to improve a patient’s mental health by incrementally learning from conversations, adapting talk-therapy strategies, maintaining a conversational manner using natural language processing, and offering 24/7 support.


Applying machine learning to medical diagnosis can improve diagnostic accuracy.

Lately, experts are finding that an algorithm that uses “counterfactual inference” trained on causal reasoning (causes behind diseases) is far superior to a simpler associative algorithm based on symptoms and patient history.

In a recent study in Nature Communications, researchers found that the causal-reasoning algorithm achieved “expert clinical accuracy” rivaling the top 25 per cent of doctors.

Machine learning in medical diagnostics is also being deployed to:

  • Use pattern recognition and segmentation techniques on medical images (retinal scans, pathology slides, etc.) to enable faster diagnoses and tracking of disease progression
  • Use molecular data to classify patients into subtypes at the molecular level in order to provide more targeted treatments
  • Diagnose conditions like autism and Alzheimer’s disease at an early stage, enabling better management and treatments


Besides the use of ML for repurposing existing drugs, ML is also being deployed to:

  • Predict pharmaceutical properties of molecular compounds and targets for drug discovery
  • Identify patients most likely to respond to certain drugs, such as cancer drugs
  • Help physicians predict surgical complications

Machine Learning is Perfectly Calibrated to the Future of Healthcare

As the healthcare industry harnesses the power of ML, we’ll benefit in ways we can only glimpse now and can attempt to study. The paradigm is shifting, but we’ve already conditioned ourselves with positive experiences of ML in other venues.

Spotify and Facebook, for example, have spoiled us with their “knowing” recommendations. We now have cultivated expectations for engagement and response that we can bring to healthcare.

Machine learning can be an expert ally in the quest for health when smart analytics can make all the difference.

Steve Pogrebivsky, the president and co-founder of DataGroomr, is an expert in data and content management systems with over 25 years of experience.

Steve has founded several technology companies, including MetaVis Technologies, which built tools for Microsoft Office 365, Salesforce, other cloud-based information systems, and Stelex Corporation, which provided compliance and technical solutions to FDA-regulated organisations.

Steve holds a BS in Computer/Electrical Engineering and an MBA in Information Systems from Drexel University. You can follow Steve on Twitter @pogrebs or LinkedIn

Continue Reading


  1. Pingback: Stretchable chip analyses health data with brain-mimicking AI

  2. Pingback: Congenica partners with Avesthagen to grow India presence

  3. Pingback: Roundtable Recap: Mental Health in the Digital Age | Health Tech World

  4. Pingback: Bipolar platform aims to speed up diagnosis | Health Tech World

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending stories