How an AI triage radiology service gained FDA clearance
Envisionit Deep AI® became the first ever African-founded company to receive FDA clearance for an AI device. With Hardian Health’s support, Envisionit’s assisted radiology solution, RADIFY® AI, received FDA 501(k) clearance to triage Pneumothorax and Pleural Effusion – two conditions that result in challenges for teams working in emergency rooms and intensive care units.
Drawing on her extensive experience in paediatric radiology, Dr. Jaishree Naidoo identified a significant deficiency in diagnostic services throughout Africa, particularly in regions with limited access to specialised care. Motivated by her firsthand experiences in public hospitals and outreach programs, Envisionit was founded with a mission to leverage the power of AI to democratise access to diagnostic healthcare globally, beginning with RADIFY®.
RADIFY®’s goal is to enhance the efficiency and effectiveness of medical professionals in emergency rooms and intensive care units. By aiding in the triage process and delivering rapid, accurate prioritisation of radiology images, it facilitates the fast and accurate treatment of various diseases. Promoting a shift towards a more proactive rather than reactive solution, empowering healthcare teams to preemptively address medical challenges in time-critical settings and improving overall patient outcomes.
In 2019, Dr. Naidoo took the bold step of leaving her clinical career to fully commit to bringing Envisionit to life with the help of her co-founders, Terence Naidu and Andrei Migatchev. The onset of the COVID-19 pandemic further highlighted the global nature of the diagnostic challenges she aimed to address.
RADIFY® AI platform
What makes RADIFY® stand out in radiology AI?
With AI becoming increasingly talked about in healthcare settings, numerous companies are well on their way to developing and receiving approval for AI as a medical device (AIaMD) products. However, this is by no means an easy task. Between biases and the need for substantial evidence, the obstacles are plenty. So, what sets Envisionit apart, and how were they able to overcome the common challenges faced by many AIaMD developers?
Incorporating diverse data sets has been instrumental for Envisionit in mitigating bias commonly experienced with the use of AI. Doing so has enabled RADIFY® to operate effectively across diverse demographics, providing reliable diagnostics to a global patient population – enhancing the tool's accuracy and broadening its applicability.
While diverse datasets are undoubtedly crucial, what is rich data without accurate validation? This is a question the FDA is certainly always asking…
The rigorous process to achieve FDA clearance for RADIFY® involved an extensive validation framework, leveraging the company's proprietary platform, RATify. This facilitated continual refinement of RADIFY® through real-time feedback from radiologists. By incorporating their expert consensus and extensive data analysis, this iterative process ensured that the AI's diagnostic recommendations were grounded in accuracy.
The path to FDA 501(k) clearance
Gaining FDA 501(k) clearance for RADIFY®, was no small feat. The process was challenging due to recent updates to the FDA cybersecurity guidelines, which became effective only in September 2023, requiring far more detailed and meticulous attention to security risk management including threat modelling, cybersecurity risk assessment, and implementation and documentation of cybersecurity controls.
Envisionit came to Hardian Health for support in navigating the regulatory landscape and guidance on the FDA's stringent requirements for an AI product like RADIFY®. Through close collaboration with the team, we gained a comprehensive understanding of their needs and goals. Leveraging our blend of clinical and technical expertise, and close understanding of the regulatory landscape, enabled us to effectively support Envisionit throughout their journey to achieving FDA clearance.
We worked with Envisionit to first establish an Intended Use Statement, then define their predicate devices. Once a regulatory strategy was in place, we worked with them to help them build their own Quality Management System (QMS). This proactive approach empowered them to identify and address any existing operating gaps, facilitating continuous improvement. Building this QMS not only supported the FDA clearance process but also enabled the team to construct a strong regulatory foundation to achieve and maintain regulatory compliance and quality assurance. By leveraging these insights, the Envisionit team were able to tailor their approach to meet the regulatory criteria more efficiently.
“The regulatory expertise of the Hardian Health team guided us in understanding the intricate details and requirements, as well as the compliance standards that were necessary for our product’s approval” – Dr. Jaishree Naidoo
Dr. Jaishree Naidoo – CEO, Envisionit Deep AI
A critical step was also leveraging Envisionit’s proprietary validation platform, RATify, to collect and analyse data, facilitate expert feedback, and enable the continuous refinement of RADIFY®. It ensured that the AI's diagnostic recommendations were robust, accurate, and met the FDA's rigorous standards.
Despite changes to the FDA’s requirements during the approval process, as a team, we were able to adapt quickly to ensure there were no delays in Envisionit’s journey to submission. This meant the team had to enhance RADIFY®’s capabilities and incorporate more detailed subgroup analyses and robust cybersecurity measures. In fact from the date of first submission to final clearance, 182 days elapsed - this compares favourably with an analysis that we had performed on the real length of time taken for FDA 510(k) clearance, which shows that for the FDA’s product code QFM (the one used for RADIFY®) the time to clearance of the last 20 submissions on the FDA database had a maximum of 262 days, a median of 154 days and a mean of 144 days - the need to provide the more detailed clinical data and analysis took the bulk of the additional time. We revised the cybersecurity aspects in parallel with augmenting the clinical data and analysis; we received no findings at all from the FDA on the approach to cybersecurity that we took under the changed requirements.
The way forward for AI diagnostics
"There are constant hordes of bees that contribute to building the infrastructure of a bee hive. But if there was one bee or a series of bees that dropped the bar, that brought a lower quality into the hive, the whole hive is affected.
We all contribute to the symbiotic relationship of artificial intelligence. So by incorporating monitoring and validation in an AI tool, this is a way to actually bridge a gap between artificial intelligence and radiologists by getting their feedback, by improving their adoption to it.
And this is really the foundation of Envisionit Deep AI." – Dr. Jaishree Naidoo
Envisionit’s core philosophy likens the collaborative effort required to enhance AI tools for healthcare to the collective work of bees in a hive. Each contribution is vital, and the integrity of the entire system depends on maintaining high standards across all inputs.
The development of RADIFY® reflected this ethos by integrating a continuous feedback loop from radiologists into its AI model. This approach served in refining RADIFY®'s accuracy and enhanced its relevance and usability in clinical settings. The active involvement of radiologists in the tool’s development and improvement processes implied that RADIFY® remained attuned to the realities of medical practice, thereby facilitating smoother adoption and more effective application in patient care.
Moreover, the emphasis on monitoring and validation within RADIFY®’s design was reflective of Envisionit's broader mission: to bridge the gap between cutting-edge AI technology and the day-to-day demands of radiological work.
With a successful 510(k) clearance under their belt, we look forward to continuing to witness the positive impact that the team will have on the sector and on both radiologists and patients alike.
Want to know more about upcoming AI standards like ISO 42001? Our partners at URM Consulting are hosting a webinar in this topic on Wednesday 24th April.
Hardian Health is a clinical digital consultancy focused on leveraging technology into healthcare markets through clinical strategy, scientific validation, regulation, health economics and intellectual property.
