Classification of diffraction patterns in single particle imaging experiments performed at X-ray free-electron lasers using a convolutional neural network

TL;DR

This paper demonstrates that convolutional neural networks, specifically YOLOv2 and YOLOv3, can effectively classify diffraction patterns in single particle imaging experiments, achieving high accuracy and aiding automated data analysis at X-ray free-electron lasers.

Contribution

The study applies transfer learning with YOLO networks to classify SPI diffraction patterns, showing CNNs outperform traditional methods and can be trained with moderate data.

Findings

YOLOv2 achieved about 97% accuracy in classification.

Color images on linear scale improved classification performance.

CNN-based classification is comparable to manual data classification.

Abstract

Single particle imaging (SPI) is a promising method for native structure determination which has undergone a fast progress with the development of X-ray Free-Electron Lasers. Large amounts of data are collected during SPI experiments, driving the need for automated data analysis. The necessary data analysis pipeline has a number of steps including binary object classification (single versus multiple hits). Classification and object detection are areas where deep neural networks currently outperform other approaches. In this work, we use the fast object detector networks YOLOv2 and YOLOv3. By exploiting transfer learning, a moderate amount of data is sufficient for training of the neural network. We demonstrate here that a convolutional neural network (CNN) can be successfully used to classify data from SPI experiments. We compare the results of classification for the two different networks, with different depth and architecture, by applying them to the same SPI data with different data representation. The best results are obtained for YOLOv2 color images linear scale classification, which shows an accuracy of about 97% with the precision and recall of about 52% and 61%, respectively, which is in comparison to manual data classification.