Register for free to
view the full innovation profile.
IVFvision.ai is a digital platform based on AI-computer vision, which predicts which embryos are most likely to lead to pregnancy.
About
IVFvision.ai: AI-powered system for the prediction of pregnancy in fertility treatment based on single-image embryo classification Umnet need: Infertility affects one in six couples worldwide. In Vitro Fertilization (IVF) is one of the most common treatments for infertility. In IVF, selection of the most viable embryo is key to achieving pregnancy. Conventional embryo evaluation involves visual morphology assessment and manual grading of human embryos at the blastocyst stage (embryo on day 5) by skilled embryologists. Although universally used in clinical practice, this is a crude, subjective evaluation of embryo quality, is incomplete and time-consuming. The assessment produces different results between embryologists and different grading systems are used among medical centres. Attempts to establish a universal grading and selection system have failed. In some countries, embryologists lack the training and experience to make accurate selections. As a result of all these factors: IVF failure rates remain high - average 79% failure per embryo transferred (all ages) patients often undergo multiple cycles to achieve pregnancy Some patients abandon treatment (high drop-out rates). to overcome uncertainties in embryo quality, multiple embryos are frequently transferred, resulting in undesired multiple pregnancies and complications Patients suffer from: Physical and psychological burden resulting from failed cycles or undergoing multiple cycles to achieve pregnancy High costs (private patients) Low levels of satisfaction Complications linked to multiple births e.g. perinatal mortality, cerebral palsy, hypertension, preeclampsia, gestational diabetes, caesarean birth, miscarriage In some countries, social stigma associated with childlessness UK tax payers are impacted because £87m of their taxes are funding failed NHS cycles and avoidable births. 33,445 cycles are funded by the NHS annually. Proposed Solution We have developed IVFvision.ai, a digital platform based on AI-computer vision, which predicts which embryos are most likely to lead to pregnancy. The system can be used by Clinical Embryologists as a decision-support tool for selecting the best embryo for transfer to the womb, in order to increase IVF success rates. The cloud-based platform is fully operational and can be accessed at www.IVFvision.ai. Advantages of our system versus current practice: Advanced-technology convolutional neural networks identify patterns invisible to the human eye Objective Consistent DHT classification: IVFvision.ai is classified as digital health technology (DHT) Tier 3b. Level of development of proposed solution IVFvision.ai is at Technology Readiness Level (TRL) 7. “System prototype demonstration in operational environment”. Training of the AI model Convolutional Neural Networks were used to develop an algorithm using 1500 anonymised digital embryo images, from both standard microscopy and time-lapse technology. Proof of concept Preliminary clinical validation of IVFvision.ai was performed at Cambridge IVF Clinic in the UK. IVFvision.ai was better in predicting implantation than KIDScore and three experienced embryologists. The reliability of IVFvision.ai was perfect (ICC=1.00), consistently returning the same classification after a triple-reading process. The reliability between embryologists was moderate (ICC=0.744). Study presented at 2020 Annual Meeting of American Society of Reproductive Medicine. Adaptability of the system IVFvision.ai is a versatile AI system based on a modual architecture. We have successfully adaped the AI algorithm to diagnose brain tumours from MRI images with AUC 99%. The system may be easily adapted to be used in - histopathology - MRI images - sperm selection in IVF treatment
Key Benefits
Benefits for Patients Higher chances of pregnancy with single embryo transfer Reduced incidence of multiple births Reduction in complications linked to multiple births Reduction of physical/psychological burden Reduced costs Fewer mental health issues Greater choice of clinics closer to patient’s home resulting in less travel abroad Reduced drop-out rates i.e. patients abandoning treatment Benefits for the NHS £13m per annum could be spent on successes rather than failures. Of the total 81,596 IVF cycles per year in the UK, 21% lead to live birth (20,000 IVF babies) 41% of all UK IVF cycles are NHS funded i.e. 33,455 Assuming a 79% failure rate, NHS funded 26,429 failed cycles Average cost per NHS-funded cycle is £3000 NHS is spending £79.3 million per annum on failed cycles Difference in cost for the NHS between a single and a twin birth is £9067 Approximately 10% of IVF NHS births are multiple births i.e. 821 multiples Extra cost to the NHS to deliver 821 twin births resulting from NHS funded IVF cycles is £7.5 million So, the total cost of failed NHS funded IVF treatments and avoidable multiple births is over £86.8 million per annum Impact on IVF clinics increased productivity improved laboratory workflow better quality control faster staff training improved consistency in embryo assessments among embryologists Other impacts Reduction in need to travel, reducing carbon footprint UK/NHS seen as an early adopter of innovation and promoter of best international practice A different staffing model could be deployed with a reduction in the need for highly trained staff resulting in more jobs and cost savings per capita Equality - making treatment available to all people irrespective of background or wealth Choice - giving couples a wider choice of high quality providers
Applications
- In vitro fertilisation (automated assessment of oocytes, sparmatozoa and embryos) - Histopathology (automated diagnossi by reading tissue sections on slides) - Brain tumour diagnosis using MRI images