A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
April 21st, 2023
Here, a new model for thyroid nodule detection in ultrasound images is proposed, which uses Swin Transformer as the backbone to perform long-range context modeling. Experiments prove that it performs well in terms of sensitivity and accuracy.
This program implements an automatic and efficient measure for detecting thyroid nodules, which may reduce the workload of doctors in the future. This technique applies a Swin Transformer to a nodule detection network. Swin Transformers are effective in capturing long-range content information, which is very helpful in improving the sensitivity of nodule detention.
To begin, open a terminal on the machine and install Python 3.8 by typing sudo apt-get install python3.8 in the command line. To install PyTorch 1.7, go to the official website of Miniconda and install it. Then create a Conda environment by typing the command shown on the screen in the command line, and activate it by typing conda activate SwinFasterRCNN.
Install PyTorch by writing the shown command in the command line. For the MM detection installation, clone from the official GitHub repository, and then install MM Detection by writing the command pip install v e in the command line. Open the schedule_1x.
py file in the Vim text editor and edit the file to set the default training-related parameters, including the optimizer, learning rate, and epic. Begin the training by typing the command shown on the screen. Wait for the network to begin the training for 48 epics, and for the resulting trained weights of the SwinFasterRCNN network to be generated in the output folder.
Save the model weights with the highest accuracy on the validation set. After training, select the model with the best performance on the validation set for thyroid nodule detection in the new images. Run the test script to resize the image to 512 pixels by 512 pixels and normalize it.
Wait for the script to automatically load the pre-trained model parameters to the SwinFasterRCNN, and feed the pre-processed image into the SwinFasterRCNN for inference. Wait for the SwinFasterRCNN to output the prediction box for each image. Finally, allow the script to automatically perform NMS post-processing on each image to remove the duplicate detection boxes.
The detection results for the same image are shown in a given row. The columns are the detection results for SwinFasterRCNN, FasterRCNN, YOLOv3, SSD, RetinaNet, and DETR. The ground truths of the regions are marked with green rectangular boxes.
The detection results are framed by the red rectangular boxes. It is very important to select the appropriate hyper parameter to ensure those converges which needs to be determined by many experiments.
View the full transcript and gain access to thousands of scientific videos
Overview
This study presents a novel approach for detecting thyroid nodules in ultrasound images using a Swin Transformer model. The proposed method demonstrates improved sensitivity and accuracy in nodule detection.
Key Study Components
Area of Science
- Medical Imaging
- Machine Learning
- Thyroid Health
Background
- Thyroid nodules are common and require accurate detection for proper diagnosis.
- Traditional methods can be time-consuming and subjective.
- Machine learning techniques, particularly deep learning, have shown promise in medical image analysis.
- Swin Transformers are effective in capturing long-range contextual information in images.
Purpose of Study
- To develop an automated system for thyroid nodule detection.
- To enhance the sensitivity and accuracy of nodule detection in ultrasound images.
- To reduce the workload of healthcare professionals.
Methods Used
- Implementation of a Swin Transformer-based nodule detection network.
- Training the model with a dataset of ultrasound images.
- Utilization of Python and PyTorch for model development.
- Post-processing techniques to refine detection results.
Main Results
- The Swin Transformer model achieved high sensitivity and accuracy in detecting thyroid nodules.
- Comparison with other models like FasterRCNN and YOLOv3 showed superior performance.
- Effective hyperparameter tuning was crucial for optimal model performance.
- Results were visually represented with detection boxes highlighting true positives and false positives.
Conclusions
- The proposed model significantly improves the detection of thyroid nodules.
- Automation in nodule detection can alleviate the burden on healthcare providers.
- Future work may focus on further refining the model and expanding its applicability.
Frequently Asked Questions
Executive Industry Relevance
Automated detection of thyroid nodules in ultrasound images addresses a critical bottleneck in early disease identification and risk stratification. Integrating Swin Transformer-based models into discovery-stage imaging workflows enhances predictive confidence and supports scalable, reproducible analysis across large datasets. This capability is directly relevant for biopharma teams seeking to accelerate biomarker discovery and translational research in endocrine and oncology portfolios.
Strategic Applications in Biopharma R&D
Early Discovery & Target Validation
- Enables high-throughput, quantitative identification of disease-relevant features in imaging datasets.
- Supports objective, reproducible target validation by reducing operator subjectivity in nodule detection.
- Improves mechanistic de-risking by capturing long-range contextual dependencies in tissue morphology.
- Facilitates portfolio triage by providing robust, scalable image analysis outputs.
Screening & Assay Development
- Prepares validated imaging datasets for downstream phenotypic screening and biomarker analysis.
- Standardizes detection outputs, supporting reproducibility and cross-study comparability.
- Enables scalable screening of large patient cohorts for translational biomarker discovery.
- Provides quantitative bounding box and confidence score outputs for reliable evaluation.
Translational & Preclinical Research
- Aligns imaging outputs with disease-relevant endpoints for translational continuity.
- Supports risk-adjusted advancement decisions by improving sensitivity and specificity in detection tasks.
- Enables integration of imaging biomarkers into preclinical and clinical research pipelines.
- Reduces late-stage biological risk by improving early detection accuracy.
Pipeline & Workflow Integration
This Swin Transformer-based detection model fits within the imaging analysis continuum from early discovery through translational research, supporting both hypothesis testing and biomarker validation.
- Discovery Biology: Enhances hypothesis testing by providing objective, quantitative imaging readouts.
- Screening: Delivers reproducible, scalable detection outputs suitable for high-throughput screening workflows.
- Analytics: Outputs quantitative bounding boxes and confidence scores, enabling robust statistical comparisons.
- Translational Research: Bridges discovery and preclinical phases by aligning imaging outputs with disease endpoints.
- Enterprise Reuse: Provides a reusable, adaptable imaging analysis capability for diverse disease models.
Operational & Enterprise Impact
- Scientific Value: Increases predictive confidence and reduces mechanistic ambiguity in imaging-based studies.
- Operational Value: Standardizes and automates image analysis, improving reproducibility and scalability.
- Strategic Value: Enables better go/no-go decisions and capital efficiency by reducing manual review burden.
- Portfolio Impact: Supports risk-adjusted prioritization and advancement of imaging-driven discovery programs.
Implementation Considerations
- Requires expertise in machine learning, medical imaging, and Python-based frameworks.
- Needs access to GPU-enabled computational infrastructure for model training and inference.
- Demands cross-team standardization of imaging protocols and annotation practices.
- Adaptation across different imaging modalities or disease models may require additional validation.
- Hyperparameter tuning and model selection are critical for optimal performance and convergence.
