When most people think about tele-health they think of video conferencing with a doctor – but tele-health is much more than that! Tele-health, increasingly referred to as Virtual Care, refers to a range of modes of delivering healthcare to people “where they are”, rather than having to meet in person. And increasingly it is so called “asynchronous” modes of delivery – where information is collected virtually and collated for review. This review step is increasingly occurring using AI technologies – allowing a much more timely response.
Our Perth based tele-health teams develop a range of tele-health technologies that enable information to be collected remotely and then sent to the medical team for review. An initial AI based review can enable prioritisation of the review by physicians, and increasingly might be relied upon to inform decision making.
Today in the second part of our AI in Healthcare Video series, we have three videos from our Perth based tele-health team:
- Mr Jana Vignarajan – giving an overview of the technology we build for our asynchronous tele-health healthcare
- Dr Sajib Saha – describing how we use deep learning to diagnose Age Related Macular degeneration
- Ms Maryam Mehdizadeh – describing our collaboration with the Lions Eye Institute to diagnose Stargardt disease
Our Perth based scientists, Jana, Sajib and Maryam will be attending the AIDH Digital Health Summit in Perth on November 13 – they’d love to see you there!
And don’t forget you can download the our AI Report: Exemplars of Artificial Intelligence and Machine Learning in Healthcare (PDF) report
Title:[Start of recorded material at 00:00:00] [CSIRO Team Leader Mr Jana Vignarajan appears prominent on the screen]
How do you deploy great artificial intelligence technologies for health care in a challenging environment such as rural and remote settings? Introducing our stack of Telehealth Platforms?
Hi, my name is Janardhan Vignarajan and I lead the Telehealth Solutions team at CSIRO AEHRC. Our team specialises in deploying world class Telehealth Solutions across various health care scenarios.
In a remote healthcare service delivery model, you will have all sorts of data stored across multiple systems.
The data include patient medical images, patient information, diagnosis and management plans.
Streamlining such data storage and utilization and providing excellent patient care is always been a challenge. Data is generally stored across multiple systems with various providers.
Introduction of AI and decision support tools using these data sets will provide efficient patient care delivery on the spot.
However, such AI systems need to be delivered in a controlled platform to reach out to end users.
Over the years, we have developed excellent platform technologies to assist in delivering health care in many remote settings, including rural and metropolitan area.
These platform technologies act as host for the AI algorithm and decision support tools. By engaging with multiple clinical partners and adopting agile development methodology, we have built the platforms co-developing with the clinicians which provide the best results with fit for purpose clinical applications.
Our work in teleophthalmology has achieved great outcomes and we have developed and deployed a store and forward system for retinal imaging in the rural and remote settings called Remote-I. This platform is highly adaptable and works across multiple settings and hosts our diabetic retinopathy automatic grading system such that the patient retinal photographs are screened on the spot by the AI for diabetic retinopathy. The platform is now actively being used internationally and in rural Australia.
We have also built various mobile technology solutions for health and one such platform is MICE short for Medical Image Communication and Exchange platform. In collaboration with South Metropolitan Health Service in Western Australia, we have developed two mobile apps to transmit images and patient consent securely and effectively while sending the records to the hospital systems.
We are also building innovative AI solutions to detect various dental conditions, using our tele-dental platform using dental x-rays which assist in the detection of caries and other clinical pathologies.
Our work in Tele-health space has generated many real world deployment in the health care industry. The AI system for diabetic retinopathy detection is actively being used by the community.
Our mobile apps are now being used in Western Australian hospitals for patient medical imaging and consent capture process. We have also now come up with an efficient annotation tool to train the AI, so that our researchers can now perform the data collection quickly compared to other traditional methods. Due to the changing nature of the technology and advancements of cloud deployment methodologies, it is paramount that we keep rediscovering Tele-health and deploy new technologies and platforms efficiently and effectively.
Download the report today for more insights into using artificial intelligence and machine learning for health applications. Read exciting case studies from Australia’s largest digital health initiative, the Australian e-Health Research Centre, and get in touch with us to discuss collaboration’s.
[End of recorded material 00:05:04]
Title:[Start of recorded material at 00:00:00] [CSIRO Software Engineer Mrs Maryam Mehdizadeh appears prominent on the screen]
Stargardt disease is a genetic eye disease and the most common childhood macular condition affecting about one in ten thousand people. It affects the function of the macular, central part of the retina, leading to a progressive loss of central vision of both eyes. It first appears, between the ages of 10 to 20, although visual impairment may not be apparent until 40 to 50 years of age. The detection and monitoring of the disease rely mainly on manual quantification of flecks in retinal images.
Find out how the Australian eHealth Research Centre uses deep machine learning to speed up the monitoring and disease progression.
[Animation of logo of Australian e-Health Research Centre]
[Maryam Mehdizadeh appears prominent on the screen. Her name and title briefly appear on screen]
My name is Maryam Mehdizadeh, working as a software engineer at the AEHRC, CSIRO. We developed and tested a deep learning model for automated segmentation of retinal flecks associated with Stargardt disease. We worked in conjunction with Lions Eye Institute. This study was supported by Retina Australia, Telethon Perth Children’s Hospital and Macular Disease Foundation, Australia.
Until now, there is no effective treatment for Stargardt disease. Fundis autofluorescence imaging has been primarily used for detecting and quantifying the area of flecks in the retina. These retinal flecks usually result from the deaths of retinal pigment epithelial cells. Manual segmentation of retinal flecks allows monitoring the progress and severity of the disease. However, this approach is subjective and time consuming because it relies on manual interpretation by clinicians. At the AEHRC we developed and validated a deep learning model for automated segmentation of retinal flecks.
[Maryam Mehdizadeh appears prominent on the screen]
The newly developed software initially aligns the images that were obtained over different time points. These images were then used to feed a trained deep learning segmentation model to automatically detect the flecks. A range of fleck parameters including the number and accumulative area of flecks can also be obtained. This study demonstrated the feasibility of utilizing deep learning to segment and quantify retinal lesions associated with Stargardt disease. This lays the foundation for future clinical studies. The central idea behind the study was to enable clinicians to have an objective and timely quantification of number and total area of flecks over a period of time.
This is important in assessing the progression and severity of Stargardt disease, as well as preventing permanent vision loss. This software can shorten the amount of time a clinician spends to assess the progress of the disease. Such approach will free up the time for ophthalmologists and allow them to focus on treatment of the disease. Currently, there is no effective treatment for Stargardt disease, although various possible treatments are currently under investigation. Improving the detection and quantification of flecks over time, can result in greater information that will be critical in evaluating the effectiveness of various treatment options.
Deep learning can facilitate creating digital archives, which is fundamental in assessing the progression and severity of Stargardt disease over time.
Download the report today for more insights into using artificial intelligence and machine learning for health applications, read exciting case studies from Australia’s largest digital health initiative, the Australian eHealth Research Centre, and get in touch with us to discuss collaborations.
[End of recorded material 00:04:35]
Title:[Start of recorded material at 00:00:00] [CSIRO Research Scientist Dr Sajib Saha appears prominent on the screen]
Age related macular degeneration is responsible for half of all cases of blindness in Australia. Early detection can help reduce the risk of blindness. Find out how the Australian e-Health Research Centre uses artificial intelligence to detect age related macular degeneration at an early stage.
[Animation of logo of Australian e-Health Research Centre]
[Sajib Saha appears prominent on the screen. His name and title briefly appear on screen]
I’m Dr Sajib Saha, a scientist in the Artificial Intelligence and TeleHealth Team in CSIRO, together with the Doheny Eye Institute of the University of California, Los Angeles we have developed an artificial intelligence method for detection and classification of early biomarkers for age related macular degeneration.
Age related macular degeneration, also known as macular degeneration, is the name given to a group of chronic degenerative diseases that affect the retina at the back of the eye and cause progressive loss of central vision. Macular degeneration affects millions of people and is a leading cause of blindness throughout the world. Though treatments are available, early detection is crucial to reduce the risk of blindness. The newly developed optical coherence tomography, or in short, OCT, has become a clinical practice in recent days to diagnose macular degeneration.
In this work we develop artificial intelligence method for detection and classification of early macular degeneration biomarkers in OCT so that macular degeneration can be diagnosed at an early stage and better treatment plan can be facilitated. We have developed deep learning models for automated detection and classification of hyperreflective foci, hyporeflective foci within drusen, and subretinal drusenoid deposits from OCT B-scans.
For each of the pathology types we trained a different convolutional neural network, with varying depths. Pre segmentation of the retinal layers was performed using another independently trained confolutional neural network.
[Sajib Saha appears prominent on the screen.]
Eleven different training approaches were tested, each fine tuning different portions of the network layers. A total of nineteen thousand five hundred and eighty four spectral domain, (SD)-OCT B-scans were used for this study and these OCT scans were collected from one hundred and fifty three patients who are diagnosed with early or intermediate macular degeneration in at least one eye between 2010 and 2014.
We achieved an overall accuracy of 87 percent for identifying the presence of early macular degeneration pathology.
Over the last years, our collaborative research on macular degeneration has had direct impact, both in theoretical and applied research, including multiple publications.
Given the increasing burden of macular degeneration on healthcare system, the proposed automated system is highly likely to perform a vital role in decision support systems for patient management and in population and primary care based screening approaches for macular degeneration.
[Image of report cover appears on black background with a voiceover]
Download the report today for more insights into using artificial intelligence and machine learning for health applications. Read exciting case studies from Australia’s largest digital health initiative, the Australian e-Health Research Centre, and get in touch with us to discuss collaborations.
[End of recorded material 00:05:04]