Author Correction: Analysis of the Human Protein Atlas Image Classification competition.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Publisher Correction: Analysis of the Human Protein Atlas Image Classification competition.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Publisher Correction: Analysis of the Human Protein Atlas Image Classification competition.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

A fast and scalable method for quality assurance of deformable image registration on lung CT scans using convolutional neural networks.

PURPOSE: To develop and evaluate a method to automatically identify and quantify deformable image registration (DIR) errors between lung computed tomography (CT) scans for quality assurance (QA) purposes. METHODS: We propose a deep learning method to flag registration errors. The method involves preparation of a dataset for machine learning model training and testing, design of a three-dimensional (3D) convolutional neural network architecture that classifies registrations into good or poor classes, and evaluation of a metric called registration error index (REI) which provides a quantitative measure of registration error. RESULTS: Our study shows that, despite having limited number of training images available (10 CT scan pairs for training and 17 CT scan pairs for testing), the method achieves 0.882 AUC-ROC on the test dataset. Furthermore, the combined standard uncertainty of the estimated REI by our model lies within ± 0.11 (± 11% of true REI value), with a confidence level of approximately 68%. CONCLUSIONS: We have developed and evaluated our method using original clinical registrations without generating any synthetic/simulated data. Moreover, test data were acquired from a different environment than that of training data, so that the method was validated robustly. The results of this study showed that our algorithm performs reasonably well in challenging scenarios.

Analysis of the Human Protein Atlas Image Classification competition.

Pinpointing subcellular protein localizations from microscopy images is easy to the trained eye, but challenging to automate. Based on the Human Protein Atlas image collection, we held a competition to identify deep learning solutions to solve this task. Challenges included training on highly imbalanced classes and predicting multiple labels per image. Over 3 months, 2,172 teams participated. Despite convergence on popular networks and training techniques, there was considerable variety among the solutions. Participants applied strategies for modifying neural networks and loss functions, augmenting data and using pretrained networks. The winning models far outperformed our previous effort at multi-label classification of protein localization patterns by ~20%. These models can be used as classifiers to annotate new images, feature extractors to measure pattern similarity or pretrained networks for a wide range of biological applications.