Periodic Report Summary - UVOLUMES (Novel ultrasound-based tridimensional tools for structural and functional evaluation of volumes of interest in human organs)
The aim of the project was to develop tools to improve ultrasound imaging analysis using fetal medicine as a model, with a strong focus on the development of tools that would be directly or indirectly applicable to three-dimensional imaging, and with a large transversal interest for ultrasound in other medical specialties. The specific aims of the project work packages (WPs) were:
1. WP1: semi-automated area-volume selection
2. WP2: automated quantitative assessment of tissue structure, and
3. WP3: improved methods for vascular and perfusion assessment.
In general, the development of the project was most remarkable in WP2 and WP3, which were more ambitious, and was rather slower for WP1.
Work on automated segmentation was successfully performed as part of WP1 and Siemens fully developed a tool for automated selection of areas and volumes. Region merging segmentation and binary partition tree were implemented to improve current segmentation techniques. Neonatal templates were successfully implemented and fetus templates were almost ready by the time of this report. We worked as with registration of magnetic resonance imaging (MRI) and ultrasonography (US) imaging modalities to pave the way for future work on automated three-dimensional fetal brain parcellation. The work concerning three-dimensional transducers was only partially implemented due to a delay in the development of silicon three-dimensional transducers by Siemens, which appeared to be permanent and was a combination of technological difficulties with changes in the company.
On the other hand, major advances were achieved in WP2 and a trade secret was already deposited. An automated tool for the analysis of image and acoustic changes was developed and a series of successful experiments was performed. Some of the results were in press by the time of this report and others were in preparation for publication. Automated analysis of acoustic information provided a discriminative imaging biomarker of white matter damage (Se:100 %/Sp:97 %) in a real population of 100 patients, one third of which suffered subtle brain damage. Following this report we would exploit the development of this new tool and new engineering work would be dedicated to improve the predictive capabilities in various conditions and with different imaging modalities to show that automated acoustic information analysis could be robust in large sets of patients with different professionals. The algorithm was so strong that a new set of experiments would test the ability to correlate not only with structural changes but also with functional data. The theoretical framework was successfully developed to transfer the automated acoustic information analysis to three-dimensional volumes and extract invariant features; however the most outstanding advance was the successful development of a robust algorithm which was expected to have a strong impact on medical and other fields of imaging.
In WP3 we started from a previous algorithm developed by the clinical partner, namely the fractional moving blood volume (FMBV), trying to overcome the limitations that prevented this algorithm to be (almost) independent of the operator and machine settings and to set the basis in order to export the algorithm to make real time three-dimensional and four-dimensional exams. We developed a number of methodologies to overcome these limitations. We demonstrated that dimensionless measurements presented the most reliable and repeatable results. 'Dimensionless' provided the mathematical basis to directly transfer such measurements to future four-dimensional acquisitions, which would include the three dimensions plus time. We also planned to work towards a full development of a range of dimensionless perfusion indexes. Two-dimensional plus time movies allowed to clearly identify systole and diastole peaks and might produce new diagnostics and finer monitoring. We were also working for the improvement of current clinical applications and focussed on the automation of two-dimensional images to two-dimensional plus time movies, which increased precision in the diagnosis and opened new opportunities for the development of imaging biomarkers. These new perfusion methodologies achieved a high correlation with clinical markers when applied to a new clinical application, namely fetal anaemia, and a first paper on this subject was in preparation by the end of this reporting period.
1. WP1: semi-automated area-volume selection
2. WP2: automated quantitative assessment of tissue structure, and
3. WP3: improved methods for vascular and perfusion assessment.
In general, the development of the project was most remarkable in WP2 and WP3, which were more ambitious, and was rather slower for WP1.
Work on automated segmentation was successfully performed as part of WP1 and Siemens fully developed a tool for automated selection of areas and volumes. Region merging segmentation and binary partition tree were implemented to improve current segmentation techniques. Neonatal templates were successfully implemented and fetus templates were almost ready by the time of this report. We worked as with registration of magnetic resonance imaging (MRI) and ultrasonography (US) imaging modalities to pave the way for future work on automated three-dimensional fetal brain parcellation. The work concerning three-dimensional transducers was only partially implemented due to a delay in the development of silicon three-dimensional transducers by Siemens, which appeared to be permanent and was a combination of technological difficulties with changes in the company.
On the other hand, major advances were achieved in WP2 and a trade secret was already deposited. An automated tool for the analysis of image and acoustic changes was developed and a series of successful experiments was performed. Some of the results were in press by the time of this report and others were in preparation for publication. Automated analysis of acoustic information provided a discriminative imaging biomarker of white matter damage (Se:100 %/Sp:97 %) in a real population of 100 patients, one third of which suffered subtle brain damage. Following this report we would exploit the development of this new tool and new engineering work would be dedicated to improve the predictive capabilities in various conditions and with different imaging modalities to show that automated acoustic information analysis could be robust in large sets of patients with different professionals. The algorithm was so strong that a new set of experiments would test the ability to correlate not only with structural changes but also with functional data. The theoretical framework was successfully developed to transfer the automated acoustic information analysis to three-dimensional volumes and extract invariant features; however the most outstanding advance was the successful development of a robust algorithm which was expected to have a strong impact on medical and other fields of imaging.
In WP3 we started from a previous algorithm developed by the clinical partner, namely the fractional moving blood volume (FMBV), trying to overcome the limitations that prevented this algorithm to be (almost) independent of the operator and machine settings and to set the basis in order to export the algorithm to make real time three-dimensional and four-dimensional exams. We developed a number of methodologies to overcome these limitations. We demonstrated that dimensionless measurements presented the most reliable and repeatable results. 'Dimensionless' provided the mathematical basis to directly transfer such measurements to future four-dimensional acquisitions, which would include the three dimensions plus time. We also planned to work towards a full development of a range of dimensionless perfusion indexes. Two-dimensional plus time movies allowed to clearly identify systole and diastole peaks and might produce new diagnostics and finer monitoring. We were also working for the improvement of current clinical applications and focussed on the automation of two-dimensional images to two-dimensional plus time movies, which increased precision in the diagnosis and opened new opportunities for the development of imaging biomarkers. These new perfusion methodologies achieved a high correlation with clinical markers when applied to a new clinical application, namely fetal anaemia, and a first paper on this subject was in preparation by the end of this reporting period.