It can take doctors a few tries to identify a patient’s condition, especially when the symptoms are common. Traditional infectious diseases tests can only detect a small number of diseases , so doctors have to use educated guesswork when choosing which tests to administer.

One startup in America, Karius, has developed a new way of diagnosing infectious diseases. It runs an analysis on a blood sample and compares the data with an extensive database of diseases. Artificial Intelligence (AI) is then used to identify the pathogens present.

Karius’ algorithms are powered by Amazon Web Services (AWS). Here are six ways AI is set to help clinicians improve the quality of care, enhance patient safety, and save precious time.

1. Detect infectious diseases

First, AI can help to find infectious diseases in blood samples. Disease-causing microbes leave DNA traces in our bloodstream, and US-based startup Karius developed a unique test that uses these traces to identify the types of bacteria, viruses, fungi in blood plasma.

Karius analyses blood samples using genomic sequencing machines. It then uploads this data onto AWS Cloud and applies AI algorithms to find which infections are present. This typically takes just a day.

AI comes in handy when processing the sheer volume of genomic data. One genome can have many millions of mutations across hundreds of thousands of individuals. Karius runs on a secure Amazon Virtual Private Cloud and processes tens to hundreds of millions of data points per patient.

This new method of diagnosing infectious diseases is a lot more useful than conventional diagnosis methods. Besides only being able to test for a few specific diseases, traditional tests can be invasive. Doctors may have to take a tissue sample from, say, the heart or the brain, to test for infections there, and repeat the procedure until a diagnosis is confirmed.

These tests may also only show the presence of an infection, but may not provide the necessary detail to choose the most appropriate intervention or dosage for treatment.

A single Karius test can detect more than 1000 pathogens. It doesn’t matter where in the body the infection is located; the test picks up on its traces in the bloodstream. The test can also tell doctors how severe the infection is, based on how much of the microbes are present.

2. Listen for mental distress

Next, Eleos Health, an AWS-based startup partner, uses AI-based voice analysis to improve psychiatric treatments. According to the startup, 45 per cent of clinical improvement is associated with voice-related data and biomarkers.

During telehealth or in-person consultations, AI conducts encrypted voice analysis to look for factors such as listening ratio, length of silences, speech rate, and the number of “I” statements. The algorithm then points out what needs to be changed in a treatment plan to help deliver more effective treatment and improve outcomes.

In another project, two US universities and the University of Pittsburgh Medical Center are working on sensing technologies that can pick up on signs of depression. This is part of the Pittsburgh Health Data Alliance’s and AWS’s collaboration to advance health innovation.

The technology aims to automatically measure subtle changes in individuals’ behavior — such as facial expressions and use of language. A quick and accurate depression marker would help clinicians identify less prominent cases, and measure patients’ responses to treatments better.

This research demands high computational power – researchers have to train machine learning models on tens of thousands of examples across acoustic, visual and written data. AWS services have allowed the team to train their models in just a few days, rather than weeks.

3. Build 3D models

Healthtech firm HeartFlow uses AI to create personalised 3D models of patients’ hearts. Cardiologists use this to better evaluate the impact a blockage has on blood flow and determine the best treatment.

Heart disease is typically expensive, risky and invasive to test for. The procedure can lead to stroke, major blood vessel damage and other serious complications. Yet, more than half of patients who undergo the test turn out to have no significant blockages at all.

HeartFlow’s non-invasive method translates a scanned image of the heart into a digital 3D model, which is then refined by the company’s certified experts. Its machine learning algorithm calculates how blockages are affecting the patient’s blood flow in the heart.

This platform runs on AWS Cloud, which allows HeartFlow to deliver results quickly. It uses Amazon Simple Storage Service to store data, Amazon Elastic Compute Cloud to process it, and Amazon Aurora for data protection.

4. Predict breast cancer

In another project under the Pittsburgh Health Data Alliance-AWS collaboration, a research team is experimenting with using deep-learning systems to predict the short‐term risk of developing breast cancer. The algorithms analyse mammograms to understand the risk factors for breast cancer with the help of Amazon SageMaker, a machine learning tool.

This may help patients take preventive action earlier on. Clinicians can also diagnose it sooner, when interventions are more effective.

5. Find addiction patterns

In 2017, 2.1 million people in the US misused opioids chronically, and 47,600 died from opioid overdoses. Health analytics company axialHealthcare has built an AI platform to identify opioid misuse early, before it turns into addiction.

axialHealthcare’s platform analyses data on prescriptions and claims and uses machine learning to flag high-risk patterns. The platform uses Amazon RedShift and Amazon Elastic MapReduce to store and process high volumes of data.

The company then refers cases that need intervention to its Clinical Consult Services team for follow-up by phone or in-person visits. This is supported by Amazon Connect, a self-service, cloud-based contact centre service.

6. Evolving the patient experience

AWS recently announced a machine learning upgrade of its Amazon Connect call centre tool at re:Invent 2020. Healthcare organisations such as axialHealthcare and aged care facility Juniper use Amazon Connect to manage patient calls, but they will soon be able to do much more.

Amazon Connect can now use machine learning to authenticate callers in real time. Once callers have given their consent to use their voice ID, the tool creates a unique voiceprint for the caller. This will be used to verify their identity the next time they call.

Contact centre staff can also resolve queries more quickly with Amazon Connect Wisdom (preview) and Contact Lens real time analytics. The service analyses the conversation, picks up on the root issue, and combs through the database to recommend helpful information.

Amazon HealthLake (preview) is another tool that will help to improve patient experience. The tool consolidates and standardises health data across complex and separate systems, then stores it on a secure cloud.

This gives doctors a complete view of a patient’s medical history, and hospitals can predict population health trends and optimise hospital efficiency. All this can be done with tools to ensure security and data protection of sensitive data.

AI has become a worthwhile apprentice on the cutting-edge of medical research and treatments. From detecting infectious diseases to predicting cancer, it offers great potential in transforming the future of healthcare.