ORCID Profile
0000-0002-5382-2097
Current Organisation
The University of Auckland
Does something not look right? The information on this page has been harvested from data sources that may not be up to date. We continue to work with information providers to improve coverage and quality. To report an issue, use the Feedback Form.
Publisher: IEEE
Date: 08-2014
Publisher: Wiley
Date: 17-02-2023
DOI: 10.1002/JMRI.28643
Abstract: Recently, deep learning via convolutional neural networks (CNNs) has largely superseded conventional methods for proton ( 1 H)‐MRI lung segmentation. However, previous deep learning studies have utilized single‐center data and limited acquisition parameters. Develop a generalizable CNN for lung segmentation in 1 H‐MRI, robust to pathology, acquisition protocol, vendor, and center. Retrospective. A total of 809 1 H‐MRI scans from 258 participants with various pulmonary pathologies (median age (range): 57 (6–85) 42% females) and 31 healthy participants (median age (range): 34 (23–76) 34% females) that were split into training (593 scans (74%) 157 participants (55%)), testing (50 scans (6%) 50 participants (17%)) and external validation (164 scans (20%) 82 participants (28%)) sets. 1.5‐T and 3‐T / 3D spoiled‐gradient recalled and ultrashort echo‐time 1 H‐MRI . 2D and 3D CNNs, trained on single‐center, multi‐sequence data, and the conventional spatial fuzzy c‐means (SFCM) method were compared to manually delineated expert segmentations. Each method was validated on external data originating from several centers. Dice similarity coefficient (DSC), average boundary Hausdorff distance (Average HD), and relative error (XOR) metrics to assess segmentation performance. Kruskal–Wallis tests assessed significances of differences between acquisitions in the testing set. Friedman tests with post hoc multiple comparisons assessed differences between the 2D CNN, 3D CNN, and SFCM. Bland–Altman analyses assessed agreement with manually derived lung volumes. A P value of .05 was considered statistically significant. The 3D CNN significantly outperformed its 2D analog and SFCM, yielding a median (range) DSC of 0.961 (0.880–0.987), Average HD of 1.63 mm (0.65–5.45) and XOR of 0.079 (0.025–0.240) on the testing set and a DSC of 0.973 (0.866–0.987), Average HD of 1.11 mm (0.47–8.13) and XOR of 0.054 (0.026–0.255) on external validation data. The 3D CNN generated accurate 1 H‐MRI lung segmentations on a heterogenous dataset, demonstrating robustness to disease pathology, sequence, vendor, and center. 4. Stage 1.
Publisher: Elsevier BV
Date: 12-2010
DOI: 10.1016/J.EJMECH.2010.09.018
Abstract: Known drug space (KDS) was analysed for the occurrence of natural products and their derivatives. A database of 1000 marketed drugs was compiled. It was found that 10% of the drugs on the market are unaltered natural products, 29% are their derivatives (semi-synthetics) and the rest (61%) have a synthetic origin. Of the natural products, and their derivatives, polycyclic drugs were the most abundant at 21% followed by simple drugs (16%) and steroids (15%). In regard to the molecular descriptors the natural products had larger statistical means and standard deviations than their synthetic counterparts. It was found that KDS occupies a larger volume in chemical space with respect to drug-like chemicals, i.e., KDS fully encompasses drug-like chemical space with the parameters of molecular weight≤800 g mol(-1), log P≤6.5, hydrogen bond acceptors≤15, hydrogen bond donors≤7, polar surface area≤180 Å2, and rotatable bonds≤17. Only 13% of the drugs analysed are outside one or more of these parameters. The definition of KDS gives drug designers a larger volume to work in compared to drug-like chemical space. However, the bulk of known drugs are found within the volume of drug-like chemical space.
Publisher: ASME International
Date: 18-01-2050
DOI: 10.1115/1.4029919
Abstract: Previous studies of the ex vivo lung have suggested significant intersubject variability in lung lobe geometry. A quantitative description of normal lung lobe shape would therefore have value in improving the discrimination between normal population variability in shape and pathology. To quantify normal human lobe shape variability, a principal component analysis (PCA) was performed on high resolution computed tomography (HRCT) imaging of the lung at full inspiration. Volumetric imaging from 22 never-smoking subjects (10 female and 12 male) with normal lung function was included in the analysis. For each subject, an initial finite element mesh geometry was generated from a group of manually selected nodes that were placed at distinct anatomical locations on the lung surface. Each mesh used cubic shape functions to describe the surface curvilinearity, and the mesh was fitted to surface data for each lobe. A PCA was performed on the surface meshes for each lobe. Nine principal components (PCs) were sufficient to capture % of the normal variation in each of the five lobes. The analysis shows that lobe size can explain between 20% and 50% of intersubject variability, depending on the lobe considered. Diaphragm shape was the next most significant intersubject difference. When the influence of lung size difference is removed, the angle of the fissures becomes the most significant shape difference, and the variability in relative lobe size becomes important. We also show how a lobe from an independent subject can be projected onto the study population’s PCs, demonstrating potential for abnormalities in lobar geometry to be defined in a quantitative manner.
Location: United Kingdom of Great Britain and Northern Ireland
No related grants have been discovered for Ho-Fung Chan.