Continuous Learning

A better understanding of how to verify and validate continuous learning algorithms is needed. At this time there are none authorized for Digital Pathology by FDA. There is currently too much uncertainty around the least burdensome approach for verification and validation as well as testing for continuous learning algorithms. We are looking for initial input from stakeholders on general principles for verification and validation testing for increased efficiency and access. One outcome of this workgroup is a whitepaper outlining how to validate, verify, and/or achieve interoperability.

Key Elements, Next Steps, Timeline

 
  1. Use cases, start with QC/aid of pathologist, already cleared IVDs and describe clinical benefits and possible user concerns (timeline)

  2. Use current two guidances for changes, i.e. general modification guidance, and software modification guidance, FDA draft action plan for regulatory framework and timeline

  3. Use CADe and CADx + use RWD + timeline

  4. Use differences between radiology & digital pathology

  5. Use Pre-submission and mock submission paths

  6. Use of MDDT to create: a validated reference dataset (including clinical outcome data)

 

Concerns & Problems

 
  1. Pathologist will not review the image, HCP adoption, R&D and Class III

  2. Feature picked by AI is different versus pathologist clinical (need RWD and clinical outcome data), R&D and regulatory

  3. Reimburse concerns for patient, clinical, R&D and regulatory

 

Value Proposition

 
  1. Global access

  2. Reduce time to market

  3. Reduce costs

  4. Reduce submission risk

  5. Increase accuracy and precision in Dx

  6. Increasing precision medicine (right Dx and Rx for patient)

 

Implications & Efforts

 
  1. Clarity on regulatory pathway - retrain and how often to release

  2. Continuous learning implementing in RWD

  3. Retraining versus self-learning - using different data, need quality data - pre-specify changes

Current Projects

News & Updates



Relevant Publications

 

Coming soon

Group Leader

 
Esther Abels.png

Esther Abels, MSc

Esther Abels has a background in bridging R& D, proof of concept, socio economics and pivotal clinical validation studies used for registration purposes in different geographies, for both pharma and biotech products. She brings to Visiopharm a wealth of regulatory and clinical experience specializing in bringing products to clinical utility. She played a crucial role in getting WSI devices reclassified in USA. Esther currently also leads the Digital Pathology Association (DPA) Regulatory and Standards Taskforce and FDA collaborations to drive regulatory and standard clarifications for interoperability and computational pathology in the field of digital pathology. She is also a co-founder of the Alliance for Digital Pathology. Esther holds a MSc in Biomedical Health Science from Radboud University Nijmegen.