Managing Stroke using AI in rural India

Qure AI
8 min readJul 5, 2021

--

Dr Justy Antony Chiramal MD, DTMH

(This article first appeared in Gaon Connection)

At 7 am in rural Assam, 45-year-old tea plantation worker Ram Charan* was having breakfast at home when he suddenly fell off his chair and realized that his right arm and right leg could not be moved. Though Ram smoked beedis, he had no other vices and hadn’t reported any major ailments. His family rushed him to the nearby health center, where the doctor rightly diagnosed that he had suffered a stroke and asked for a Brain CT scan to be taken. However, the closest diagnostic center was 25 km away. By the time Ram returned from his scan, it was almost 9 am. He was put on medication to bring his blood pressure of 190/100, down. The attending physician was already overwhelmed by the number of patients in the ER. So he had a quick look at the CT scan but was unsure if Ram could start on any blood thinners. By the time he was referred to the Baptist Christian Hospital, Tezpur (a town close to the Assam — Arunachal Pradesh border) it was 12 pm and had missed the time window for administering the clot removing injection as per the standard Stroke guidelines (American Stroke Association-ASA). The injection and radiology procedures for stroke were also beyond the limited means of the patient and the radiology procedure hospital . Ram was eventually discharged with significant disability and limited function, after Post-Stroke Care including physiotherapy and rehabilitation for 1 month.

In India, more than 4000 people like Ram Charan develop Stroke every day, with a case fatality rate of up to 42%. It is one of the top 5 leading causes of death in the country, and across the world. In the past decade, the number of Stroke cases have almost doubled in low-and-middle-income countries, which shoulder 80% of the world’s stroke burden. This is because of the higher prevalence of risk factors for Stroke such as hypertension, tobacco consumption, and an aging population, as well as the quality of diagnostics. Hypertension is a major risk factor for Stroke. It’s prevalence in the indigenous Assamese population and tea garden workers has been reported to be higher than the national average. Anecdotal reports and hospital-based studies from this region indicate that the disease is a huge burden in Assam.

Physician using Qure’s AI to detect Stroke

Strokes are sometimes called “Brain Attacks”, as they’re usually caused by reduced blood flow to the brain due to blocked blood vessels — circumstances similar to a heart attack. They’re diagnosed through a brain imaging scan to detect bleeding or clotting. If it is the latter, an injection to break up the clot must be administered within 4.5 hours of the symptom’s onset — which now happens in less than 5% of all cases. Further treatments to remove the clot can only be carried out in Comprehensive Stroke Centers, which are mostly found in urban areas. There are well-established Stroke care pathways in most developed nations, but these are fragmented in many low- and middle-income countries.

Another significant issue faced by India is the inability to scale up Stroke care through a specialist-led model as there are only about 2500 neurologists and 10000 radiologists for a population of 1.4 billion. For patients like Ram Charan and many others with access only to smaller hospitals with no specialists, there is a significant delay in the clinical decision-making process, due to lack of expertise. Emerging digital technologies are helping bridge the gap between demand and supply through teleconsultation platforms (eg. TeleStroke) and AI-assisted diagnostics.

The Baptist Christian Hospital (BCH) in Tezpur is a 130-bed hospital that serves communities in both Assam and Arunachal Pradesh. The hospital is equipped with CT machines as it has an established Physician-Led Stroke Unit. However, BCH depends on tele-radiology services for imaging reports, as well as neurologists in referral hospitals. To reduce the turnaround time and improve patient outcomes, BCH has recently adopted and deployed Mumbai-based health-tech firm Qure.ai’s AI technology — qER, which helps triage and automate Head CT scan interpretation, prioritizes cases for radiologists which in turn, enables physicians to attend to critical patients first.

qER is FDA approved and can detect bleeds, fractures, strokes, etc., generating results in under a minute and immediately alerting physicians via notifications on their phones. By drawing the attention of the doctors as well as the remote radiologists to critical abnormalities on CT scans, the reporting time has reduced from more than 60 minutes to less than 5 minutes, which is important in time-sensitive processes like Stroke management. Technologies like qER also provide an additional level of support, boosting the confidence of non-specialists in these settings.

Physicians using Qure’s solution in Emergency Rooms

Qure’s AI technology does not require any complex hardware, is easy to deploy, and has transformed stroke care. Dr. Jemin Webster, a Physician at BCH, Tezpur says, “It is a handy tool when there is no access to radiologists. In a rural setting, while there are centers with CT machines, reports are only available after a few days. Qure’s AI tool can be very helpful in centers with high caseloads in ERs and in settings like ours which do not have neuro-specialists or radiologists.”

As we embark on improvising Stroke Care in India, technologies like Qure.ai’s qER have a major critical role to play in supporting and empowering the first point of contact for acute Stroke patients — smaller hospitals with non-specialist clinicians. Improving the quality and efficiency of stroke care can significantly improve outcomes, reduce long-term disability making a tremendous difference to the patients’ life.

Introduction

COVID-19 is a contagious vascular and respiratory disease caused by SARS-CoV-2, a novel variant of the SARS virus. First identified in December 2019, the World Health Organisation declared COVID-19 a ‘Public Health Emergency of International Concern’ in January 2020 and a pandemic in March 2020. As of November 2020, the disease has been confirmed in approximately 48 million people worldwide.

Be it the first world or the third, rapidly rising COVID-19 infection caseloads have placed health systems under enormous pressure, often threatening to overwhelm available resources. Numerous research studies have shown that imaging procedures are the most efficient method of evaluating the lung condition of patients with COVID-19 as well as the disease’s rate of progression. In resource challenged situations, chest X-rays augmented by Artificial Intelligence (AI) have been used for mass screening for COVID-19.

An NHS Hospital Case

The Royal Bolton Hospital in Farnworth, Greater Manchester is home to the Bolton National Health Service (NHS) Foundation Trust, providing healthcare services for people in the Metropolitan Borough of Bolton and the surrounding areas. The Trust was the first in the United Kingdom to use qXR, an imaging interpretation tool developed by Qure.ai, to help medical staff effectively monitor the extent and progression rate of COVID-19 in patients through the automated analysis of chest X-rays.

Challenge(s)

The initial months of 2020 saw an alarming spike in COVID-19 cases in the UK. It resulted in an immediate impact on the NHS, medical and social care systems as they faced the challenge of ensuring adequate testing. Patients suffering from the illness placed unprecedented demands on acute care services, particularly on intensive care units (ICUs) and the already burdened medical workforce. This resulted in a substantial rise in mortality rates.

As in other parts of the UK, Bolton NHS also faced an influx of COVID-19 patients and needed a system in place to improve efficiency in prognosis and alleviate the workload of the hospital staff. And Qure.ai was able to help Bolton NHS do that with qXR

Solution

COVID-19 clinically presents itself as consolidation, which are accumulations of fluid and/or tissue in pulmonary alveoli preventing gas exchange or ground-glass opacity, and through nodular shadowing. They primarily affect peripheral and lower areas of the lungs.

Qure.ai’s deep learning-based automated chest X-ray interpretation platform — qXR — can detect, localise, and quantify a total of 29 findings, including COVID-19 related lesions. It had been trained and tested using a growing database (over 3.6 Million) of X-rays from diverse sources. The solution gives 91% sensitivity and 77% specificity in predicting COVID-19 changes on chest X-rays. qXR could also monitor the extent and rate of progression of the infection, creating graphs showing the percentage area of lung affected and tracking changes on subsequent X-rays.

Selection Criteria

The COVID-19 pandemic has brought forth an unprecedented number of challenges for the medical community to resolve. Across the world, healthcare organisations have been forced to reevaluate their systems and processes and the NHS is no different. The rise and implementation of technology to enable medical staff to work with better safety and efficiency has been a key development. While it has proved invaluable in resource-starved, developmentally challenged parts of the world, this trend has also impacted other areas in a positive way.

Qure.ai is EU-GDPR compliant and ISO/IEC 27001 certified while qXR is Class 2A CE certified. The Qure server was located onsite and no data was sent offsite. The Bolton NHS team validated the solution on a test set of 11479 CXRs with 515 PCR-confirmed COVID-19 positives.

Results

During the peaks of the pandemic, when PCR testing turnaround time was more than 24 hours, the NHS clinicians relied on qXR to interpret chest X-rays reviews for the rapid assessment and triage patients into high, medium and low priority categories. qXR gave almost instantaneous feedback to clinicians, providing out-of-hours clinical decision support and assisting accurate evaluation by clinicians with less expertise in chest X-ray reviews. By providing workflow support to improve efficiency and increase reporting capacity, qXR can also help mitigate any shortage of trained radiologists in the long-term.

Success and Next Steps

The Bolton NHS Foundation Trust had been exploring the use and potential benefits of Artificial Intelligence in support of medical diagnoses and treatments for a considerable amount of time. However, the COVID-19 pandemic accelerated this process and resulted in the collaboration with Qure.ai. Not only did deploying qXR in the field of radiography help the hospital staff manage the diagnosis and treatment of patients, but also established the reputation of the Bolton NHS Foundation Trust as a leader and pioneer in the development of the use of AI as part of clinical workflow. Now that the tangible benefits of qXR have been documented at such a prestigious institution, Qure.ai is poised to partner with and assist other healthcare organisations to ensure that the benefits of technology are received by every stakeholder.

AWS Infrastructure

  • Qure is tackling artificial intelligence (AI) challenges in the healthcare industry, advancing digital healthcare through medical imaging AI solutions.
  • On AWS, we are using EC2 for heavy processing and easy scalability.
  • AWS provides us with 99.8% SLA which improves our performance. We have automated backups and failover in real-time.
  • Our data is stored in S3 for better security, scalability, performance and data availability. Also, we have RDS for database reliability.
  • We are using cloudtrail which monitors and records account activity throughout AWS infrastructure.
  • We have AWS WAF which is a web application firewall at the perimeter level to secure our web apps & APIs against malicious traffic, web exploits, botnets, etc.
  • AWS environment complies with data privacy and security guidelines.

*Name changed to protect patient confidentiality.

--

--

Qure AI
Qure AI

Written by Qure AI

Qure.ai is a breakthrough AI solution provider that is disrupting the radiology ‘status quo’ by enhancing imaging accuracy and improving health outcomes.