Assisting radiologists with transformer-based fracture detection for chest radiographs

Chest x-rays are a widely used diagnostic imaging technique in the medical field, but expert interpretation can be time-consuming and subjective, leading to potential diagnostic errors. To overcome these limitations, computer-aided diagnostic (CAD) systems using artificial intelligence (AI) have been developed, with transformer-based object detectors showing promising results. This study explores the rib fracture detection power of an out-of-the-box transformer detector and examines its adaptability to medical imaging. A private dataset of 30,000 chest radiographs was used to conduct experiments and compare the performance of transformer-based object detectors with a traditional CNN-based Faster-RCNN implementation. A retrospective analysis showed that transformer detectors had superior performance (i.e., 25.3mAP vs. 23.6mAP). Additionally, a reader study was conducted to evaluate the added diagnostic value of the AI system when assisting two radiologists in diagnosing chest x-rays. The results showed that the AI system provided valuable support improving the true positive rates from 73.6% to 83.4%. In conclusion, the study highlights the potential of transformer-based object detectors as a valuable tool in medical imaging, particularly in the diagnosis of chest x-ray rib fractures.

[1]  Jiangfen Wu,et al.  Convolutional neural network for detecting rib fractures on chest radiographs: a feasibility study , 2023, BMC Medical Imaging.

[2]  Zhonghao Zhang,et al.  Focal DETR: Target-Aware Token Design for Transformer-Based Object Detection , 2022, Sensors.

[3]  Y. Zhan,et al.  Assessment of automatic rib fracture detection on chest CT using a deep learning algorithm , 2022, European Radiology.

[4]  Rini Widyaningrum,et al.  Artificial intelligence based on Convolutional Neural Network for detecting dental caries on bitewing and periapical radiographs , 2022, Jurnal Radiologi Dentomaksilofasial Indonesia (JRDI).

[5]  Chencui Huang,et al.  Development and assessment of deep learning system for the location and classification of rib fractures via computed tomography. , 2022, European journal of radiology.

[6]  H. Shum,et al.  DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection , 2022, ICLR.

[7]  Sang Min Lee,et al.  Self-evolving vision transformer for chest X-ray diagnosis through knowledge distillation , 2022, Nature Communications.

[8]  Akira M. Murakami,et al.  Improving Radiographic Fracture Recognition Performance and Efficiency Using Artificial Intelligence. , 2021, Radiology.

[9]  Minming Zhang,et al.  Rib fracture detection system based on deep learning , 2021, Scientific Reports.

[10]  Khan Muhammad,et al.  COVID-Transformer: Interpretable COVID-19 Detection Using Vision Transformer for Healthcare , 2021, International journal of environmental research and public health.

[11]  Gang Zeng,et al.  Conditional DETR for Fast Training Convergence , 2021, 2021 IEEE/CVF International Conference on Computer Vision (ICCV).

[12]  Huangjing Lin,et al.  Development and Evaluation of a Deep Learning Algorithm for Rib Segmentation and Fracture Detection from Multicenter Chest CT Images. , 2021, Radiology. Artificial intelligence.

[13]  Bernhard Kainz,et al.  RATCHET: Medical Transformer for Chest X-ray Diagnosis and Reporting , 2021, MICCAI.

[14]  Thomas Muller,et al.  DoT: An efficient Double Transformer for NLP tasks with tables , 2021, FINDINGS.

[15]  G. Durhan,et al.  Errors, discrepancies and underlying bias in radiology with case examples: a pictorial review , 2021, Insights into Imaging.

[16]  Joon Beom Seo,et al.  Multi-task vision transformer using low-level chest X-ray feature corpus for COVID-19 diagnosis and severity quantification , 2021, Medical Image Analysis.

[17]  Ausif Mahmood,et al.  The NLP Cookbook: Modern Recipes for Transformer Based Deep Learning Architectures , 2021, IEEE Access.

[18]  Lei Chen,et al.  A fully automated rib fracture detection system on chest CT images and its impact on radiologist performance , 2021, Skeletal Radiology.

[19]  Binh T. Nguyen,et al.  VinDr-CXR: An open dataset of chest X-rays with radiologist’s annotations , 2020, Scientific Data.

[20]  Nicolas Usunier,et al.  End-to-End Object Detection with Transformers , 2020, ECCV.

[21]  I. Tetko,et al.  State-of-the-art augmented NLP transformer models for direct and single-step retrosynthesis , 2020, Nature Communications.

[22]  Michael A. Bruno.  Error and Uncertainty in Diagnostic Radiology , 2019 .

[23]  Yifan Yu,et al.  CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison , 2019, AAAI.

[24]  N. Nadeem,et al.  Diagnostic Accuracy of Computed Tomography Scout Film and Chest X-ray for Detection of Rib Fractures in Patients with Chest Trauma: A Cross-sectional Study , 2019, Cureus.

[25]  O. Arthurs,et al.  Chest radiographs versus CT for the detection of rib fractures in children (DRIFT): a diagnostic accuracy observational study , 2018, The Lancet. Child & adolescent health.

[26]  Le Lu,et al.  ChestX-Ray8: Hospital-Scale Chest X-Ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases , 2017, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[27]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[28]  Pietro Perona,et al.  Microsoft COCO: Common Objects in Context , 2014, ECCV.

[29]  C. Tulay,et al.  Comparison of Chest Computed Tomography and Chest X-Ray in the Diagnosis of Rib Fractures in Patients with Blunt Chest Trauma , 2012 .

[30]  Fei-Fei Li,et al.  ImageNet: A large-scale hierarchical image database , 2009, 2009 IEEE Conference on Computer Vision and Pattern Recognition.

[31]  M. Grgic,et al.  Overview of the DICOM standard , 2008, 2008 50th International Symposium ELMAR.

[32]  Mathias Prokop,et al.  Digital chest radiography: an update on modern technology, dose containment and control of image quality , 2008, European Radiology.

[33]  S. L. Miller,et al.  Clinical Rib Fractures: Are Follow-Up Chest X-Rays A Waste of Resources? , 2002, The American surgeon.

[34]  Stephen Lin,et al.  Swin Transformer: Hierarchical Vision Transformer using Shifted Windows , 2021, 2021 IEEE/CVF International Conference on Computer Vision (ICCV).

[35]  J. Crandall,et al.  Rib fracture patterns and radiologic detection--a restraint-based comparison. , 2000, Annual proceedings. Association for the Advancement of Automotive Medicine.