Mitigating Bias in Radiology Machine Learning: 2. Model Development.

There are increasing concerns about the bias and fairness of artificial intelligence (AI) models as they are put into clinical practice. Among the steps for implementing machine learning tools into clinical workflow, model development is an important stage where different types of biases can occur. This report focuses on four aspects of model development where such bias may arise: data augmentation, model and loss function, optimizers, and transfer learning. This report emphasizes appropriate considerations and practices that can mitigate biases in radiology AI studies. Keywords: Model, Bias, Machine Learning, Deep Learning, Radiology © RSNA, 2022.

Mitigating Bias in Radiology Machine Learning: 3. Performance Metrics.

The increasing use of machine learning (ML) algorithms in clinical settings raises concerns about bias in ML models. Bias can arise at any step of ML creation, including data handling, model development, and performance evaluation. Potential biases in the ML model can be minimized by implementing these steps correctly. This report focuses on performance evaluation and discusses model fitness, as well as a set of performance evaluation toolboxes: namely, performance metrics, performance interpretation maps, and uncertainty quantification. By discussing the strengths and limitations of each toolbox, our report highlights strategies and considerations to mitigate and detect biases during performance evaluations of radiology artificial intelligence models. Keywords: Segmentation, Diagnosis, Convolutional Neural Network (CNN) © RSNA, 2022.

Interatomic Potentials Transferability for Molecular Simulations: A Comparative Study for Platinum, Gold and Silver.

A perfectly transferable interatomic potential that works for different materials and systems of interest is lacking. This work considers the transferability of several existing interatomic potentials by evaluating their capability at various temperatures, to determine the range of accuracy of these potentials in atomistic simulations. A series of embedded-atom-method (EAM) based interatomic potentials has been examined for three precious and popular transition metals in nanoscale studies: platinum, gold and silver. The potentials have been obtained from various credible and trusted repositories and were evaluated in a wide temperature range to tackle the lack of a transferability comparison between multiple available force fields. The interatomic potentials designed for the single elements, binary, trinary and higher order compounds were tested for each species using molecular dynamics simulation. Validity of results arising from each potential was investigated against experimental values at different temperatures from 100 to 1000 K. The data covers accuracy of all studied potentials for prediction of the single crystals' elastic stiffness constants as well as the bulk, shear and Young's modulus of the polycrystalline specimens. Results of this paper increase users' assurance and lead them to the right model by a way to easily look up data.