Automatic deep learning-based pleural effusion segmentation in lung ultrasound images.

BACKGROUND: Point-of-care lung ultrasound (LUS) allows real-time patient scanning to help diagnose pleural effusion (PE) and plan further investigation and treatment. LUS typically requires training and experience from the clinician to accurately interpret the images. To address this limitation, we previously demonstrated a deep-learning model capable of detecting the presence of PE on LUS at an accuracy greater than 90%, when compared to an experienced LUS operator. METHODS: This follow-up study aimed to develop a deep-learning model to provide segmentations for PE in LUS. Three thousand and forty-one LUS images from twenty-four patients diagnosed with PE were selected for this study. Two LUS experts provided the ground truth for training by reviewing and segmenting the images. The algorithm was then trained using ten-fold cross-validation. Once training was completed, the algorithm segmented a separate subset of patients. RESULTS: Comparing the segmentations, we demonstrated an average Dice Similarity Coefficient (DSC) of 0.70 between the algorithm and experts. In contrast, an average DSC of 0.61 was observed between the experts. CONCLUSION: In summary, we showed that the trained algorithm achieved a comparable average DSC at PE segmentation. This represents a promising step toward developing a computational tool for accurately augmenting PE diagnosis and treatment.

Ultrasound-contrast-agent dispersion and velocity imaging for prostate cancer localization.

Prostate cancer (PCa) is the second-leading cause of cancer death in men; however, reliable tools for detection and localization are still lacking. Dynamic Contrast Enhanced UltraSound (DCE-US) is a diagnostic tool that is suitable for analysis of vascularization, by imaging an intravenously injected microbubble bolus. The localization of angiogenic vascularization associated with the development of tumors is of particular interest. Recently, methods for the analysis of the bolus convective dispersion process have shown promise to localize angiogenesis. However, independent estimation of dispersion was not possible due to the ambiguity between convection and dispersion. Therefore, in this study we propose a new method that considers the vascular network as a dynamic linear system, whose impulse response can be locally identified. To this end, model-based parameter estimation is employed, that permits extraction of the apparent dispersion coefficient (D), velocity (v), and Péclet number (Pe) of the system. Clinical evaluation using data recorded from 25 patients shows that the proposed method can be applied effectively to DCE-US, and is able to locally characterize the hemodynamics, yielding promising results (receiver-operating-characteristic curve area of 0.84) for prostate cancer localization.