Ultrasound Image Synthesis Using Generative AI for Lung Ultrasound Detection
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913644068995072 |
|---|---|
| author | Chou, Yu-Cheng Li, Gary Y. Chen, Li Zahiri, Mohsen Balaraju, Naveen Patil, Shubham Hicks, Bryson Schnittke, Nikolai Kessler, David O. Shupp, Jeffrey Parker, Maria Baloescu, Cristiana Moore, Christopher Gregory, Cynthia Gregory, Kenton Raju, Balasundar Kruecker, Jochen Chen, Alvin |
| author_facet | Chou, Yu-Cheng Li, Gary Y. Chen, Li Zahiri, Mohsen Balaraju, Naveen Patil, Shubham Hicks, Bryson Schnittke, Nikolai Kessler, David O. Shupp, Jeffrey Parker, Maria Baloescu, Cristiana Moore, Christopher Gregory, Cynthia Gregory, Kenton Raju, Balasundar Kruecker, Jochen Chen, Alvin |
| contents | Developing reliable healthcare AI models requires training with representative and diverse data. In imbalanced datasets, model performance tends to plateau on the more prevalent classes while remaining low on less common cases. To overcome this limitation, we propose DiffUltra, the first generative AI technique capable of synthesizing realistic Lung Ultrasound (LUS) images with extensive lesion variability. Specifically, we condition the generative AI by the introduced Lesion-anatomy Bank, which captures the lesion's structural and positional properties from real patient data to guide the image synthesis.We demonstrate that DiffUltra improves consolidation detection by 5.6% in AP compared to the models trained solely on real patient data. More importantly, DiffUltra increases data diversity and prevalence of rare cases, leading to a 25% AP improvement in detecting rare instances such as large lung consolidations, which make up only 10% of the dataset. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_06356 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Ultrasound Image Synthesis Using Generative AI for Lung Ultrasound Detection Chou, Yu-Cheng Li, Gary Y. Chen, Li Zahiri, Mohsen Balaraju, Naveen Patil, Shubham Hicks, Bryson Schnittke, Nikolai Kessler, David O. Shupp, Jeffrey Parker, Maria Baloescu, Cristiana Moore, Christopher Gregory, Cynthia Gregory, Kenton Raju, Balasundar Kruecker, Jochen Chen, Alvin Image and Video Processing Artificial Intelligence Computer Vision and Pattern Recognition Developing reliable healthcare AI models requires training with representative and diverse data. In imbalanced datasets, model performance tends to plateau on the more prevalent classes while remaining low on less common cases. To overcome this limitation, we propose DiffUltra, the first generative AI technique capable of synthesizing realistic Lung Ultrasound (LUS) images with extensive lesion variability. Specifically, we condition the generative AI by the introduced Lesion-anatomy Bank, which captures the lesion's structural and positional properties from real patient data to guide the image synthesis.We demonstrate that DiffUltra improves consolidation detection by 5.6% in AP compared to the models trained solely on real patient data. More importantly, DiffUltra increases data diversity and prevalence of rare cases, leading to a 25% AP improvement in detecting rare instances such as large lung consolidations, which make up only 10% of the dataset. |
| title | Ultrasound Image Synthesis Using Generative AI for Lung Ultrasound Detection |
| topic | Image and Video Processing Artificial Intelligence Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2501.06356 |