Timely implementation of AI in desperate times like this has never been a miss by India’s digital technology specialists. From DRDO to some of India’s most prominent health tech startups, COVID-19 diagnosis based on radiology imaging augmented with AI-based clinical data has come to fore. This post highlights a few
By Anusha Ashwin
The second wave of the SARS-CoV-2 pandemic that hit India has been ruthless in consuming lives. We have witnessed Covid-19 patients from being mildly symptomatic suddenly slip into fatal lung damaged conditions with want for ventilator support-based treatments. As the infectivity rate of the WHO-categorized ‘Indian Variant of Concern’ Coronavirus is far higher with patients developing vascular complications at a faster rate, the need for hospitalization for a lot more patients across the country became eminent. While the sudden surge in COVID-19 positive cases was turning up at hospitals there was a growing paucity in the supply of medical grade oxygen and essential therapeutics for COVID-19. This situation further complicated the entire treatment protocol as physicians across the country had to deal with an unprecedented, never-before surge in COVID patients seeking medical interventions under hospital settings.
To combat the spread of the second wave of COVID-19, rapid, effective screening, and immediate medical response for the infected patients became a crying need. People with suspected COVID-19 need to know quickly whether they are infected, so they can receive appropriate treatment, self-isolate, or seek medical intervention at a hospital equipped with life-saving equipment.
Currently, the gold standard of COVID-19 diagnosis, RT-PCR of nose and throat swab samples is the most used clinical screening method to determine COVID-19 infection in a person. However, the technique is manual, complicated, laborious and time-consuming with a positivity rate of only 63%. Also, numerous studies have shown that RT-PCR kits are not highly sensitive, and in some cases give false-negative results.
RT-PCR also requires specialist equipment and takes at least 24 hours to produce a result. In 24 hours, the possibility of a patient experiencing severe lung infection with SpO2 levels quickly dropping to below 80% has been witnessed. Physicians working in ICUs have noticed the deterioration of the lung tissues at a faster rate leaving patients breathless, calling for immediate intervention through external life support equipment.
Another drawback in relying only on RT-PCR tests is that the test samples are collected from the upper chest cavity, while in most COVID complicated cases lower chest abnormalities tend to occur. That is why dependency on RT-PCR results was something that the physicians realized early on was not the right approach to saving lives.
So, the want for quick, definitive interpretations of COVID-19 tests made physicians resort to other diagnostic methods that include clinical symptoms analysis, epidemiological history, and positive radiographic images (computed tomography (CT) /Chest radiograph (CXR)) combined with positive serological testing. Among these radiological imaging is now considered a prerequisite and the most important and primary diagnostic tool for COVID-19.
Why chest CT/X-Ray can be an early spotter in Covid complications and why imaging analysis need AI?
As said, the brutality of the second wave has created a need for quick and effective diagnostic tools for COVID-19 emergency room physicians and radiologists to identify those infected patients whose condition is most likely to deteriorate. The COVID-19 diagnostic approach is mainly divided into two broad categories, a laboratory-based approach with serological tests and a chest radiography-based approach.
Covid-19 infected patients are usually classified into (I) asymptomatic infection (II) acute upper respiratory tract infection (III) mild pneumonia, (IV) severe pneumonia, and (V) critical cases in need of a ventilator. As COVID-19 manifests predominantly as a pneumonia-like respiratory process, the radiology techniques most involved in its diagnosis and follow-up are those of chest imaging, mainly chest X-rays and chest CT. These tools are playing an important role in the management of patients that are confirmed or suspected to be infected with the virus. Chest X-ray and chest CT are now the two most common imaging studies for diagnosis and management of COVID-19 patients.
Chest radiography or X-Ray (CXR), although considered a little insensitive for detection of early or mild disease it is useful in triaging patients and monitoring care in those with radiographically detectable pneumonia. Physicians believe triaging COVID-19 suspected patients using X-ray/CT is fast, cost-effective, and efficient. Also, despite the high sensitivity and three-dimensional nature of CT, chest x-rays are considered more useful during the second wave due to its relative speed, low cost, portability, and accessibility, especially in low-resource settings, and with high patient volumes and critically ill patients whose transport for CT might be physically challenging.
But the advantage of chest CT scan is that it reveals characteristic radiographic findings in patients with COVID-19 pneumonia. The hallmarks of COVID-19 infection on imaging at the beginning of the disease are bilateral and peripheral ground-glass opacities with lower lobe predominance. Four to 14 days after symptom onset there is a greater lung involvement with consolidation, and ‘reverse halo sign’ and ‘crazy paving’ signs.
Pulmonologists and radiologists lookout for these certain patterns that show characteristic manifestations of viral pneumonia. Chest CT scanning in patients with COVID-19–associated pneumonia usually shows ground-glass opacification, possibly with consolidation. Some studies have reported that abnormalities on chest CT scans are usually bilateral, involve the lower lobes, and have a peripheral distribution. Pleural effusion, pleural thickening, and lymphadenopathy have also been reported.
Now, with the advent of Artificial Intelligence (AI) in the medical imaging field, diagnosis has gotten easier and also in a relatively shorter time. In the diagnosis stage of COVID-19, AI can be used to recognize patterns on medical images taken by X-Rays or CTs. Using AI, chest x-rays can augment clinical data in predicting the risk of progression to critical illness in patients with COVID-19. Apart from that AI has the potential to aid in the rapid evaluation of CT scans for differentiation of COVID-19 findings from other clinical entities. And, AI-based algorithms can readily identify CT scans with COVID-19 associated pneumonia, as well as distinguish non-COVID-related pneumonia with high specificity and accuracy.
Recently researchers at the NYU Grossman School of Medicine developed a program that used several hundred gigabytes of data gleaned from 5,224 chest X-rays taken from 2,943 seriously ill patients infected with SARS-CoV-2. They created an ML-based program to see patterns by analyzing thousands of chest X-rays. Their computer program was developed with an 80% accuracy prediction ability indicating if COVID-19 patients would develop life-threatening complications within four days.
The NYU researchers of the study, which was published in the journal npj Digital Medicine online on May 12, cited the ‘pressing need’ for the ability to quickly predict which COVID-19 patients are likely to have lethal complications so that treatment resources can best be matched to those at increased risk. So, to address this need, the NYU Langone team fed not only X-ray information into their computer analysis, but also patients’ age, race, and gender, along with several vital signs and laboratory test results, including weight, body temperature, and blood immune cell levels. Also factored into their mathematical models, which can learn from examples, where the need for a mechanical ventilator and whether each patient went on to survive or die from their infections.
The researchers in their paper have reported that the computer program accurately predicted four out of five infected patients who required intensive care and mechanical ventilation and/or died within four days of admission.
Study senior investigator Krzysztof Geras, PhD*, an assistant professor in the Department of Radiology at NYU Langone, says a major advantage to machine-intelligence programs such as theirs is that its accuracy can be tracked, updated and improved with more data. He says the team plans to add more patient information as it becomes available. He also says the team is evaluating what additional clinical test results could be used to improve their test model.
Geras says he hopes to soon deploy the NYU COVID-19 classification test to emergency physicians and radiologists. In the interim, he is working with physicians to draft clinical guidelines for its use.
How India has progressed in its own way in deploying AI
Why go all the way to New York Universities when Indian healthcare startups and government organizations have developed Artificial Intelligence- and Machine Learning-based algorithms that can be paired with medical images to analyze the condition of COVID-19 patients. Timely implementation of AI in desperate times like this has never been a miss by India’s digital technology specialists. From DRDO to some of India’s most prominent health tech startups, COVID-19 diagnosis from radiology imaging augmented with AI-based clinical data has come to fore. Listed here are a few who have made the difference:
The Centre of Artificial Intelligence and Robotics, DRDO has developed an AI-based intelligent COVID detection application software called ATMAN (AI Based Intelligent COVID-19 detector Technology for Medical Assistance).
ATMAN AI is an artificial intelligence tool for Chest X-ray screening as a triaging tool in Covid-19 diagnosis which is a method for rapid identification and assessment of lung involvement. The web-based software can classify images under normal, COVID-19, and pneumonia, using chest X-rays. DRDO claimed ATMAN AI had shown an accuracy of 96.73 percent on digital chest X-rays of RT-PCR positive patients.
The back end of ATMAN has been built with Deep Convolution Neural Network, which is tuned to accurately detect Covid-19 irrespective of limited availability of Covid X-ray images for the system to learn. The software automatically pre-processes the images before passing them to the Neural net to take care of the variant illumination levels of the X-ray images. The software, according to DRDO, is easy to navigate and can be easily accessed over Internet through a variety of devices like mobiles, tablets, laptops or computers.
ATMAN has been tested and validated by the doctors from HCG Centre for Academics and Research, Bengaluru and Ankh Life Care, Bengaluru who have also helped by providing data and relevant medical domain knowledge. ATMAN will be utilized by 5C Network, India’s largest digital network of Radiologists, with support of HCG Academics across India. 5C Network, which is connected to over 1,000 hospitals across India, will now make ATMAN available to state-run and private hospitals, which is much needed during this second wave.
Researchers from the Departments of Computational and Data Science (CDS) and Instrumentation and Applied Physics at the Indian Institute of Science (IISc), in collaboration with colleagues from the Oslo University Hospital and the University of Agder in Norway have developed a new software tool that reveals the severity of lung infections in COVID-19 patients.
The software tool developed by the IISc-led team, called AnamNet can read the chest CT scans of COVID-19 patients, and, using a special kind of neural network, estimate how much damage has been caused in the lungs, by searching for specific abnormal features. The researchers claim that a tool of this kind can provide automated assistance to doctors and therefore help in faster diagnosis and better management of COVID-19.
AnamNet employs deep learning and other image processing techniques, which have now become integral to biomedical research and applications. The software can identify infected areas in a chest CT scan with a high degree of accuracy.
The researchers have trained AnamNet to look for abnormalities and classify areas of the lung scan as either infected or not infected ‒ which is called as ‘segmentation’. The tool can judge the severity of the disease by comparing the extent of infected area with healthy area.
“It basically extracts features from the chest CT images and projects them onto a non-linear space [a mathematical representation], and then recreates the [segmented] image from this representation. This is called anamorphic image processing,” explains Naveen Paluru*, first author and PhD student in the lab of Phaneendra Yalavarthy, Associate Professor at CDS.
These findings were described in a recent study published in the journal IEEE Transactions on Neural Networks and Learning Systems.
The study also compared AnamNet’s performance with other state-of-the-art software tools which perform similar tasks. It not only matched its peers in its accuracy, but also performed just as well using fewer parameters. The neural network was also computationally less complex, which allowed the researchers to train it much faster to detect anomalies.
Another significant advantage of AnamNet is that the software is lightweight with a small memory footprint. This has enabled the team to develop an app called CovSeg that can be run on a mobile phone and hence potentially be used by healthcare professionals. “We felt the need for a lightweight framework that could be deployed as a point-of-care diagnostic device on smartphones or a Raspberry Pi,” says Paluru.
According to the researchers, AnamNet holds promise beyond merely identifying lung infections in COVID-19 patients. The team is currently focusing on making the software more robust to handle COVID-19 scans, but they are also looking to diversify to other common lung diseases like pneumonia, fibrosis, and even lung cancer in the near future.
The software tool is freely available to the public.
*Paluru N, Dayal A, Jenssen HB, Sakinis T, Cenkeramaddi LR, Prakash J, Yalavarthy PK. Anam-Net: Anamorphic Depth Embedding-Based Lightweight CNN for Segmentation of Anomalies in COVID-19 Chest CT Images. IEEE Trans Neural Netw Learn Syst. 2021.
AiBharata – MYAIRA
Bengaluru-located AiBharata https://aibharata.com/, a stealth mode startup focused on developing next-generation Artificial Intelligence (AI) products, has developed an Augmented AI-powered Automated Diagnosis Platform called MYAIRA (My AI Raksha) that can diagnose and detect 15 chest and lung diseases including COVID using chest X-Rays.
Dr. Vinayaka Jyothi, CEO & Founder, AiBharata, says, “MYAIRA is a simple and an easy to operate system. Just upload an X-ray image and get the results. X-Ray analysis results are available in a few seconds even on mobile networks. AI will mark and show the abnormal areas as it is AI will mark and show the abnormal areas as it is trained from more than 2 lakh radiology images obtained from different regions including rural, tier-2, and tier-1 cities. MYAIRA improves the efficiency of medical workflow and clinical decision support even at the rural level. Radiologists can use it as an augmented tool to speed up their decisions; doctors and physicians can use it to get an immediate diagnosis to triage patients and start early treatment; Patients or people can use it as a second-opinion app.”
“We had a field trial at a Taluk Govt Hospital and MYAIRA had 96% Sensitivity in detecting COVID infections and 99% Specificity detecting people without any infections. Due to this high accuracy performance, the doctors started prescribing XRAY along with RT-PCR tests. While they waited for RT-PCR test results for days, they could get MYAIRA results within seconds. As specificity was 99%, if MYAIRA said there was no COVID diagnosis, the chances of the person not having any COVID was significantly higher. This allowed them to segregate high-risk patients from low-risk patients. In turn this allowed them to manage to triage patients efficiently. Many people have used our app while they awaited the RT-PCR test results to know whether they contracted the COVID and it has turned out 100% accurate all the time till now,” shared Dr Jyothi.
Currently, MYAIRA is used by 138 users and 5 institutional users including a Govt Hospital, Multispeciality Hospitals, and Diagnostic Labs.
MYAIRA Web Page: https://myaira.aibharata.com/
MYAIRA Android APP (Beta Version): https://play.google.com/store/apps/details?id=com.aibharata.myaira
Videos about MYAIRA: https://www.youtube.com/watch?v=Y_g01kLqsi0
Aikenist – Aiken QuickDiag COVID19 CT
Aikenist (http://www.aikenist.com) is another Bengaluru-based startup that is building technologies to assist in solving numerous problems of humanity, improve the quality of life, and save lives. The startup has joined hands to fight the ongoing pandemic by developing useful AI’s that can assist in the early detection of COVID-19 infections.
Aikenist has developed a solution that detects the extent of damage inside the lungs. Explaining the working of the AI solution, Ashwin Amarapur, co-Founder and CEO, Aikenist, says that chest CT scan is an important tool for COVID 19 lung infection detection, disease tracking, and treatment. Chest CT scan includes both Coronal and Axial scans. The quantitative metrics are available to measure the severity of damage to the lung. Among them, Co-Rads and CTSS (CT Severity Score) are commonly used. Co-Rads score goes from 1 to 6 and shows the level of suspicion of lung infection due to COVID-19. CTSS gives the quantitative extension of damage to the lung due to ground-glass opacity.”
Explaining further, Ashwin says, “Today most Radiology centers for COVID-19 CT scans provide either or both scoring methods. It is normally done by Radiologists after going through CT Axial scans slice by slice from top to bottom. This takes a few mins for Radiologists to score. AI comes to the aid of Radiologists to reduce scoring time. The AI technology developed by Aikenist can provide an automatic preliminary report with CTSS to the Radiologists. Radiologists can modify the report if required and sign it. Aiken QuickDiag COVID19 CT provides this feature. The product can be easily plugged into existing infrastructure. The cloud solution is developed in such a way that it can be scaled to millions of scans in a short duration.”
The CTSS is based on auto detection of lung lesions due to COVID-19. It is currently being used by the Teleradiology centers. This helps them to serve a greater number of patients each day and helps Radiologists save time. Since AI uses accurate quantification methods, this helps in better scoring. As the technology helps to address a greater number of patients, the needy patients get the report on time which is very crucial to do faster diagnosis, treatment and saving life.
Ashwin indicated that the AI tool can help detect bilobular anomaly, opacities, and consolidation in a patient who might be infected with Coronavirus. According to him, the cloud software analyses the images and shares the results with healthcare professionals to help them decide the next course of action. He also claims the results are available within a minute after the scans are uploaded in the server.
Qure.ai – qXR
Health tech firm Qure.ai has developed qXR, which is a chest X-ray screening tool built with deep learning technology. qXR is one of Qure.ai’s AI-driven and machine-powered healthcare tools. It automatically generates chest X-ray interpretation reports and is claimed to be faster and more efficient than humans in detecting, tuberculosis, chronic obstructive pulmonary diseases, lung malignancies, and medical emergencies and in evaluating COVID-19 X-rays. It is built with the ability to classify chest X-rays as normal or abnormal, to identify abnormal findings, and to highlight them on the X-ray.
It can be used to separate normal from abnormal X-rays, for pre-read assistance, or as a radiology audit tool. qXR includes patented algorithms that can detect a total of 29 findings on the chest X-ray. The CE-certified algorithms have been trained and tested using a growing database (over 3.7 Million) of X-rays from diverse sources.
Recently, a clinical study successfully evaluated Qure.ai algorithm’s ability to predict adverse clinical outcomes in patients with COVID-19.
Qure.ai’s diagnostic tool for COVID-19 has been used as part of a clinical study conducted in Milan by IRCCS Ospedale San Raffaele. The study focused on the early predictors of clinical outcomes of COVID-19 and showed that the performance of Qure.ai’s chest X-ray interpretation technology, qXR, was comparable to radiologists’ assessment in predicting severity in initial chest x-rays in patients with COVID-19.
Published in European Radiology, the article stated that a high qXR chest X-ray quantitative score, based on percent of lung involvement, at admission was highly predictive of mortality due to COVID-19. This was comparable to what was obtained with the RALE score, previously validated to evaluate Acute Respiratory Distress Syndrome (ARDS).
Qure.ai, in its website, states that IRCCS Ospedale San Raffaele has been using its qXR technology since early April to aid the efforts of their physicians in pandemic management.
Prashant Warier, CEO and co-Founder of Qure.ai, was quoted saying that the company’s AI solution for chest X-rays qXR has proven to be useful in two areas in pandemic management. Firstly, to screen for individuals exhibiting signs of COVID-19 on a chest X-ray. Secondly, to monitor confirmed cases of COVID-19 to track the progression of the disease.
In a latest LinkedIn Post, Prashant Warier revealed that qXR is now just a text message away. In partnership with PATH, Qure.ai launched a qXR-bot on Telegram Messenger. A physician or health worker can send a DICOM or a JPEG of a Chest X-ray to the bot on Telegram and it gets back with the AI interpretation results in seconds. Warier believes that at a time when India is reeling from the effects of COVID-19, and frontline healthcare workers are overwhelmed, this technology can provide a second opinion to physicians and health workers or serve as an assistant in diagnosing COVID-19. Notably, PATH is one of the earliest adopters of Qure.ai’s qXR AI solution.
Qure.ai also recently announced an on-device integration with Fujifilm, which will take the company’s X-ray AI product qXR to the Fujifilm X-ray device, Xair.
Well, I am sure there are a lot many more medical companies and young AI entrepreneurs who are working to assist the entire medical frontline force in bringing about resolutions that are the need of the hour, saving time and lives and more so preventing physician burnout.
* Reference: “An artificial intelligence system for predicting the deterioration of COVID-19 patients in the emergency department” by Farah E. Shamout, Yiqiu Shen, Nan Wu, Aakash Kaku, Jungkyu Park, Taro Makino, Stanisław Jastrzębski, Jan Witowski, Duo Wang, Ben Zhang, Siddhant Dogra, Meng Cao, Narges Razavian, David Kudlowitz, Lea Azour, William Moore, Yvonne W. Lui, Yindalon Aphinyanaphongs, Carlos Fernandez-Granda and Krzysztof J. Geras, 12 May 2021, npj Digital Medicine.
You must log in to post a comment.