<p class="MsoNormal"><span style="font-size: 11.0pt; line-height: 115%; font-family: 'Arial',sans-serif;">This document provides guidance on how to apply the risk management process of ISO 14971:2019 to ML-enabled medical devices (MLMD). This document is intended to be used in conjunction with ISO 14971 and does not alter the risk management requirements specified in ISO 14971. </span></p>
<p class="MsoNormal"><span style="font-size: 11.0pt; line-height: 115%; font-family: 'Arial',sans-serif;">This document addresses risks specific to machine learning (ML). Those risks can be related to topics such as data management, feature extraction, unwanted bias, information security, training the ML model by an ML algorithm, evaluation and testing of the trained ML model. See Figure 1 for an overview of the relevant terms and their relationship. See Annex A for an explanation of bias.</span></p>
<p class="MsoNormal"><span style="font-size: 11.0pt; line-height: 115%; font-family: 'Arial',sans-serif;">It is recognized that the ML model can require retraining after a period of use to redefine its parameters. An ML model can learn continuously from patient data and modify their parameters accordingly. The description “continuous(ly) learning” is used throughout this document; the term “adaptive” is sometimes used in other documents. This document also provides examples and suggests strategies for eliminating or controlling these ML-related risks.</span></p>
Registration number (WIID)
87600
Scope
<p class="MsoNormal"><span style="font-size: 11.0pt; line-height: 115%; font-family: 'Arial',sans-serif;">This document provides guidance on how to apply the risk management process of ISO 14971:2019 to ML-enabled medical devices (MLMD). This document is intended to be used in conjunction with ISO 14971 and does not alter the risk management requirements specified in ISO 14971. </span></p>
<p class="MsoNormal"><span style="font-size: 11.0pt; line-height: 115%; font-family: 'Arial',sans-serif;">This document addresses risks specific to machine learning (ML). Those risks can be related to topics such as data management, feature extraction, unwanted bias, information security, training the ML model by an ML algorithm, evaluation and testing of the trained ML model. See Figure 1 for an overview of the relevant terms and their relationship. See Annex A for an explanation of bias.</span></p>
<p class="MsoNormal"><span style="font-size: 11.0pt; line-height: 115%; font-family: 'Arial',sans-serif;">It is recognized that the ML model can require retraining after a period of use to redefine its parameters. An ML model can learn continuously from patient data and modify their parameters accordingly. The description “continuous(ly) learning” is used throughout this document; the term “adaptive” is sometimes used in other documents. This document also provides examples and suggests strategies for eliminating or controlling these ML-related risks.</span></p>
