Development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy CT scan
Background: Dual-energy CT (DECT) is a non-invasive way to determine the presence of monosodium urate (MSU) crystals in the workup of gout. Color-coding distinguishes MSU from calcium following material decomposition and post-processing. Most software labels MSU as green and calcium as blue. There a...
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2024-03-01
|
| Series: | Research in Diagnostic and Interventional Imaging |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S277265252400005X |
| _version_ | 1827304244295761920 |
|---|---|
| author | Shahriar Faghani Rhodes G. Nicholas Soham Patel Francis I. Baffour Mana Moassefi Pouria Rouzrokh Bardia Khosravi Garret M. Powell Shuai Leng Katrina N. Glazebrook Bradley J. Erickson Christin A. Tiegs-Heiden |
| author_facet | Shahriar Faghani Rhodes G. Nicholas Soham Patel Francis I. Baffour Mana Moassefi Pouria Rouzrokh Bardia Khosravi Garret M. Powell Shuai Leng Katrina N. Glazebrook Bradley J. Erickson Christin A. Tiegs-Heiden |
| author_sort | Shahriar Faghani |
| collection | DOAJ |
| description | Background: Dual-energy CT (DECT) is a non-invasive way to determine the presence of monosodium urate (MSU) crystals in the workup of gout. Color-coding distinguishes MSU from calcium following material decomposition and post-processing. Most software labels MSU as green and calcium as blue. There are limitations in the current image processing methods of segmenting green-encoded pixels. Additionally, identifying green foci is tedious, and automated detection would improve workflow. This study aimed to determine the optimal deep learning (DL) algorithm for segmenting green-encoded pixels of MSU crystals on DECTs. Methods: DECT images of positive and negative gout cases were retrospectively collected. The dataset was split into train (N = 28) and held-out test (N = 30) sets. To perform cross-validation, the train set was split into seven folds. The images were presented to two musculoskeletal radiologists, who independently identified green-encoded voxels. Two 3D Unet-based DL models, Segresnet and SwinUNETR, were trained, and the Dice similarity coefficient (DSC), sensitivity, and specificity were reported as the segmentation metrics. Results: Segresnet showed superior performance, achieving a DSC of 0.9999 for the background pixels, 0.7868 for the green pixels, and an average DSC of 0.8934 for both types of pixels, respectively. According to the post-processed results, the Segresnet reached voxel-level sensitivity and specificity of 98.72 % and 99.98 %, respectively. Conclusion: In this study, we compared two DL-based segmentation approaches for detecting MSU deposits in a DECT dataset. The Segresnet resulted in superior performance metrics. The developed algorithm provides a potential fast, consistent, highly sensitive and specific computer-aided diagnosis tool. Ultimately, such an algorithm could be used by radiologists to streamline DECT workflow and improve accuracy in the detection of gout. |
| first_indexed | 2024-04-24T17:26:34Z |
| format | Article |
| id | doaj.art-0cb6cdc5dcba4d3da14462de6f09bdbb |
| institution | Directory Open Access Journal |
| issn | 2772-6525 |
| language | English |
| last_indexed | 2024-04-24T17:26:34Z |
| publishDate | 2024-03-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Research in Diagnostic and Interventional Imaging |
| spelling | doaj.art-0cb6cdc5dcba4d3da14462de6f09bdbb2024-03-28T06:39:40ZengElsevierResearch in Diagnostic and Interventional Imaging2772-65252024-03-019100044Development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy CT scanShahriar Faghani0Rhodes G. Nicholas1Soham Patel2Francis I. Baffour3Mana Moassefi4Pouria Rouzrokh5Bardia Khosravi6Garret M. Powell7Shuai Leng8Katrina N. Glazebrook9Bradley J. Erickson10Christin A. Tiegs-Heiden11Radiology Informatics Lab, Department of Radiology, Mayo Clinic, Rochester, MN, USADivision of Musculoskeletal Radiology, Department of Radiology, Mayo Clinic, Rochester, MN, USADivision of Musculoskeletal Radiology, Department of Radiology, Mayo Clinic, Rochester, MN, USADivision of Musculoskeletal Radiology, Department of Radiology, Mayo Clinic, Rochester, MN, USARadiology Informatics Lab, Department of Radiology, Mayo Clinic, Rochester, MN, USARadiology Informatics Lab, Department of Radiology, Mayo Clinic, Rochester, MN, USARadiology Informatics Lab, Department of Radiology, Mayo Clinic, Rochester, MN, USADivision of Musculoskeletal Radiology, Department of Radiology, Mayo Clinic, Rochester, MN, USADepartment of Radiology, Mayo Clinic, Rochester, MN, USADivision of Musculoskeletal Radiology, Department of Radiology, Mayo Clinic, Rochester, MN, USADepartment of Radiology, Mayo Clinic, Rochester, MN, USADivision of Musculoskeletal Radiology, Department of Radiology, Mayo Clinic, Rochester, MN, USA; Corresponding author at: Department of Radiology, Mayo Clinic, 200 1st St. SW, Rochester, MN 55905.Background: Dual-energy CT (DECT) is a non-invasive way to determine the presence of monosodium urate (MSU) crystals in the workup of gout. Color-coding distinguishes MSU from calcium following material decomposition and post-processing. Most software labels MSU as green and calcium as blue. There are limitations in the current image processing methods of segmenting green-encoded pixels. Additionally, identifying green foci is tedious, and automated detection would improve workflow. This study aimed to determine the optimal deep learning (DL) algorithm for segmenting green-encoded pixels of MSU crystals on DECTs. Methods: DECT images of positive and negative gout cases were retrospectively collected. The dataset was split into train (N = 28) and held-out test (N = 30) sets. To perform cross-validation, the train set was split into seven folds. The images were presented to two musculoskeletal radiologists, who independently identified green-encoded voxels. Two 3D Unet-based DL models, Segresnet and SwinUNETR, were trained, and the Dice similarity coefficient (DSC), sensitivity, and specificity were reported as the segmentation metrics. Results: Segresnet showed superior performance, achieving a DSC of 0.9999 for the background pixels, 0.7868 for the green pixels, and an average DSC of 0.8934 for both types of pixels, respectively. According to the post-processed results, the Segresnet reached voxel-level sensitivity and specificity of 98.72 % and 99.98 %, respectively. Conclusion: In this study, we compared two DL-based segmentation approaches for detecting MSU deposits in a DECT dataset. The Segresnet resulted in superior performance metrics. The developed algorithm provides a potential fast, consistent, highly sensitive and specific computer-aided diagnosis tool. Ultimately, such an algorithm could be used by radiologists to streamline DECT workflow and improve accuracy in the detection of gout.http://www.sciencedirect.com/science/article/pii/S277265252400005XGoutDual-energy CTDeep learningSegmentationConvolutional neural networkVision transformers |
| spellingShingle | Shahriar Faghani Rhodes G. Nicholas Soham Patel Francis I. Baffour Mana Moassefi Pouria Rouzrokh Bardia Khosravi Garret M. Powell Shuai Leng Katrina N. Glazebrook Bradley J. Erickson Christin A. Tiegs-Heiden Development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy CT scan Research in Diagnostic and Interventional Imaging Gout Dual-energy CT Deep learning Segmentation Convolutional neural network Vision transformers |
| title | Development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy CT scan |
| title_full | Development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy CT scan |
| title_fullStr | Development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy CT scan |
| title_full_unstemmed | Development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy CT scan |
| title_short | Development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy CT scan |
| title_sort | development of a deep learning model for the automated detection of green pixels indicative of gout on dual energy ct scan |
| topic | Gout Dual-energy CT Deep learning Segmentation Convolutional neural network Vision transformers |
| url | http://www.sciencedirect.com/science/article/pii/S277265252400005X |
| work_keys_str_mv | AT shahriarfaghani developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT rhodesgnicholas developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT sohampatel developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT francisibaffour developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT manamoassefi developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT pouriarouzrokh developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT bardiakhosravi developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT garretmpowell developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT shuaileng developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT katrinanglazebrook developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT bradleyjerickson developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan AT christinatiegsheiden developmentofadeeplearningmodelfortheautomateddetectionofgreenpixelsindicativeofgoutondualenergyctscan |