Summary
UCLA researchers in the Department of Radiology have developed a semi-automated image processing technique that recognizes areas of open lung, the anatomical lobes of the lung, and the fissures (boundaries) between them in computed tomography (CT) images. This method serves to partially automate the radiological evaluation process, thereby improving its speed and volumetric capacity over the visual inspections performed in current clinical practice for the diagnosis and monitoring of lung disease.
Background
Computed tomography (CT) imaging is commonly used in clinical practice to visualize the location, extent, and progression of lung disease. However, the reading of CT images is typically performed via visual inspection by a radiologist, requiring time and skill, and remaining prone to error. An automated image processing algorithm that can quickly and reliably segment CT images into areas of open lung, the anatomical lobes of the lung, and the fissures separating them would improve the speed and accuracy of radiological evaluation. Researchers at other institutions have developed automated image processing systems that can reliably recognize areas of open lung, but automated lobe segmentation has not previously been achieved with adequate accuracy and robustness for broad clinical implementation.
Innovation
Dr. Matthew Brown has developed a semi-automated image processing method that rapidly and reliably identifies lung, lobe, and fissure voxels in CT images. The method is robust across large patient cohorts and enables quantitative lung assessment to be performed in clinical practice.
Applications
The reported technology has been developed specifically for image analysis and segmentation of CT lung scans, which are applied for the diagnosis and monitoring of the following lung conditions:
- Emphysema-quantitative assessment and treatment planning
- Lung tumors (cancer)-localization and surveillance
- Asthma-regional assessment of air trapping
- Pulmonary Fibrosis-regional quantitative assessment
- Tuberculosis-regional quantitative assessment
The image processing algorithms may be further adaptable for the diagnosis and/or evaluation of diseases in other organ systems (brain, heart, kidney, etc.), and/or via other imaging modalities (MRI, X-ray, fluoroscopy, etc.).
Advantages
- Enables quantitative analysis for the diagnosis, assessment, and treatment of lung disease
- Increases processing speed and capacity compared to visual inspection of CT scans
- Improves lobe segmentation accuracy compared to existing fully-automated image processing systems
- Provides reliable results across a wide range of individual anatomic variation, disease pathology, and CT imaging variables
State Of Development
A prototype system has been built, and preliminary investigations have confirmed the viability of the technological approach.