Manual Segmentation Approach Research Articles

A multiparametric and multiscale approach to automated segmentation of brain veins.

Cerebral vein analysis provides a fundamental tool to study brain diseases such as neurodegenerative disorders or traumatic brain injuries. In order to assess the vascular anatomy, manual segmentation approaches can be used but are observer-dependent and time-consuming. In the present work, a fully automated cerebral vein segmentation method is proposed, based on a multiscale and multiparametric approach. The combined investigation of the R2(*)- and a Vesselness probability-map was used to obtain a fast and highly reliable classification of venous voxels. A semiquantitative analysis showed that our approach outperformed the previous state-of-the-art algorithm both in sensitivity and specificity. Inclusion of this tool within a parametric brain framework may therefore pave the way for a quantitative study of the intracranial venous system.

Spatially varying accuracy and reproducibility of prostate segmentation in magnetic resonance images using manual and semiautomated methods.

Three-dimensional (3D) prostate image segmentation is useful for cancer diagnosis and therapy guidance, but can be time-consuming to perform manually and involves varying levels of difficulty and interoperator variability within the prostatic base, midgland (MG), and apex. In this study, the authors measured accuracy and interobserver variability in the segmentation of the prostate on T2-weighted endorectal magnetic resonance (MR) imaging within the whole gland (WG), and separately within the apex, midgland, and base regions. The authors collected MR images from 42 prostate cancer patients. Prostate border delineation was performed manually by one observer on all images and by two other observers on a subset of ten images. The authors used complementary boundary-, region-, and volume-based metrics [mean absolute distance (MAD), Dice similarity coefficient (DSC), recall rate, precision rate, and volume difference (ΔV)] to elucidate the different types of segmentation errors that they observed. Evaluation for expert manual and semiautomatic segmentation approaches was carried out. Compared to manual segmentation, the authors' semiautomatic approach reduces the necessary user interaction by only requiring an indication of the anteroposterior orientation of the prostate and the selection of prostate center points on the apex, base, and midgland slices. Based on these inputs, the algorithm identifies candidate prostate boundary points using learned boundary appearance characteristics and performs regularization based on learned prostate shape information. The semiautomated algorithm required an average of 30 s of user interaction time (measured for nine operators) for each 3D prostate segmentation. The authors compared the segmentations from this method to manual segmentations in a single-operator (mean whole gland MAD = 2.0 mm, DSC = 82%, recall = 77%, precision = 88%, and ΔV = - 4.6 cm(3)) and multioperator study (mean whole gland MAD = 2.2 mm, DSC = 77%, recall = 72%, precision = 86%, and ΔV = - 4.0 cm(3)). These results compared favorably with observed differences between manual segmentations and a simultaneous truth and performance level estimation reference for this data set (whole gland differences as high as MAD = 3.1 mm, DSC = 78%, recall = 66%, precision = 77%, and ΔV = 15.5 cm(3)). The authors found that overall, midgland segmentation was more accurate and repeatable than the segmentation of the apex and base, with the base posing the greatest challenge. The main conclusions of this study were that (1) the semiautomated approach reduced interobserver segmentation variability; (2) the segmentation accuracy of the semiautomated approach, as well as the accuracies of recently published methods from other groups, were within the range of observed expert variability in manual prostate segmentation; and (3) further efforts in the development of computer-assisted segmentation would be most productive if focused on improvement of segmentation accuracy and reduction of variability within the prostatic apex and base.

Comparison of Point Estimates and Average Thicknesses of Retinal Layers Measured Using Manual Optical Coherence Tomography Segmentation for Quantification of Retinal Neurodegeneration in Multiple Sclerosis

Purpose: The advent of macular optical coherence tomography (OCT) segmentation has enabled the in vivo quantitative assessment of retinal axonal and neuronal subpopulations. Recent studies employing OCT in multiple sclerosis (MS) have utilized various manual macular segmentation approaches to quantify retinal layer thicknesses. We investigated whether measurements of retinal layers solely at the points of maximal macular thickness (point estimates) within the central macular B-scan are representative of the corresponding average layer thicknesses for the ganglion cell + inner plexiform (GCIP) layers, inner nuclear layer (INL), outer plexiform layer (OPL) and outer nuclear layer (ONL) in MS and healthy controls. Additionally, we examined the correlation of manual segmentation-derived measures of composite layers with average thickness measures derived from automated 3-D segmentation of the macular cube.Materials and Methods: Spectral-domain OCT central macular B-scans of 52 MS patients and 30 healthy controls (HCs) were manually segmented. Average layer thicknesses and layer thicknesses at the points of maximal macular thickness were calculated. Macular cube scans were also segmented utilizing a fully automated 3-D segmentation algorithm.Results: GCIP, INL and OPL maximal thicknesses derived from point estimates correlated well with the average thicknesses of these layers within the central macular B-scan, whereas the ONL maximal thickness did not correlate as strongly. Manual segmentation-derived point estimates and average thickness measures of the GCIP correlated excellently with corresponding automated segmentation-derived measures. MS patients had significantly decreased GCIP maximal and average thicknesses relative to HCs. ONL average thickness was significantly decreased in MS compared to HCs, but this was not true of the ONL maximal thickness.Conclusions: GCIP, INL and OPL maximal layer thicknesses may be used as surrogates to assess the gross structural integrity of these layers in MS, in a time-conservative fashion.

Edged watershed segmentation: A semi-interactive algorithm for segmentation of low-resolution maps from electron cryomicroscopy

Electron cryomicroscopy (cryo-EM) allows for the structural analysis of large protein complexes that may be difficult to study by other means. Frequently, maps of complexes from cryo-EM are obtained at resolutions between 10 and 25Å. To aid in the interpretation of these medium- to low-resolution maps, they may be subdivided into three-dimensional segments representing subunits or subcomplexes. This division is often accomplished using a manual segmentation approach. While extremely useful, manual segmentation is subjective. We have developed a novel semi-interactive segmentation algorithm that can incorporate prior knowledge of subunit composition or structure without biasing the boundaries between subunits or subcomplexes. This algorithm has been characterized with experimental and simulated cryo-EM density maps at resolutions between 10 and 25Å.

Semiautomated Segmentation of Pleural Effusions in MDCT Datasets

To develop and evaluate a novel algorithm for semiautomated segmentation and volumetry of pleural effusions in multidetector computed tomography (MDCT) datasets. A seven-step algorithm for semiautomated segmentation of pleural effusions in MDCT datasets was developed, mainly using algorithms from the ITK image processing library. Semiautomated segmentation of pleural effusions was performed in 40 MDCT datasets of the chest (males = 22, females = 18, mean age: 56.7 +/- 19.3 years). The accuracy of the semiautomated segmentation as compared with a manual segmentation approach was quantified based on the differences of the segmented volumes, the degree of over-/undersegmentation, and the Hausdorff distance. The time needed for the semiautomated and the manual segmentation process were recorded and compared. The mean volume of the pleural effusions was 557.30 mL (+/- 477.27 mL) for the semiautomated and 553.19 (+/- 473.49 mL) for the manual segmentation. The difference was not statistically significant (Student t-test, P = .133). Regression analysis confirmed a strong relationship between the semiautomated algorithm and the gold standard (r(2) = 0.998). Mean overlap of the segmented areas was 79% (+/- 9.3%) over all datasets with moderate oversegmentation (22% +/- 9.3%) and undersegmentation (21% +/- 9.7%). The mean Hausdorff distance was 17.2 mm (+/- 8.35 mm). The mean duration of the semiautomated segmentation process with user interaction was 8.4 minutes (+/- 2.6 minutes) as compared to 32.9 minutes (+/- 17.4 minutes) for manual segmentation. The semiautomated algorithm for segmentation and volumetry of pleural effusions in MDCT datasets shows a high diagnostic accuracy when compared with manual segmentation.

TU‐C‐332‐02: Concomitant Segmentation and Registration of Liver Anatomy Using SPECT‐CT Imaging

Purpose: To develop an automatic and accurate technique for concomitant segmentations and registration of liver anatomy using SPECT and CT images for unsealed source radiotherapy. Method and Materials: The link between segmentation and registration is given by the using the level set of a liver segmentation into the registration process. In the combined approach, the liver is automatically segmented from the CT image by evolving an initial seed with a level set until it locks to the liver's border as observed in the CT images. The time‐crossing map of the level set is then used to match gradients in the SPECT image to the level set by using a data structure containing the signed distance values at a small band of neighboring pixels. Results: The technique was applied to three cases of metastatic liver disease treated with unsealed source therapy. Results indicated that the speed map of the level set plays an importance role in obtainng an accurate registration and produce a segmentation that is superior in registration time and accuracy over manual segmentation or the standard registration approach using mutual information. Accuracy measured with the convergence analysis method was of less than 0.5 mm rotation and 1 degree rotation. Conclusion: With the proposed combined segmentation‐registration technique, the uncertainty of soft‐tissue target localization could be greatly reduced ensuring accurate therapy assessment to be precisely delivered as planned. The combined all‐in‐one approach is automated and provides excellent accuracy over manual segmentation and mutual information approaches.

Evaluation of two 3D virtual computer reconstructions for comparison of cleft lip and palate to normal fetal microanatomy

Cleft lip and palate reconstructive surgery requires thorough knowledge of normal and pathological labial, palatal, and velopharyngeal anatomy. This study compared two software algorithms and their 3D virtual anatomical reconstruction because exact 3D micromorphological reconstruction may improve learning, reveal spatial relationships, and provide data for mathematical modeling. Transverse and frontal serial sections of the midface of 18 fetal specimens (11th to 32nd gestational week) were used for two manual segmentation approaches. The first manual segmentation approach used bitmap images and either Windows-based or Mac-based SURFdriver commercial software that allowed manual contour matching, surface generation with average slice thickness, 3D triangulation, and real-time interactive virtual 3D reconstruction viewing. The second manual segmentation approach used tagged image format and platform-independent prototypical SeViSe software developed by one of the authors (F.W.). Distended or compressed structures were dynamically transformed. Registration was automatic but allowed manual correction, such as individual section thickness, surface generation, and interactive virtual 3D real-time viewing. SURFdriver permitted intuitive segmentation, easy manual offset correction, and the reconstruction showed complex spatial relationships in real time. However, frequent software crashes and erroneous landmarks appearing "out of the blue," requiring manual correction, were tedious. Individual section thickness, defined smoothing, and unlimited structure number could not be integrated. The reconstruction remained underdimensioned and not sufficiently accurate for this study's reconstruction problem. SeViSe permitted unlimited structure number, late addition of extra sections, and quantified smoothing and individual slice thickness; however, SeViSe required more elaborate work-up compared to SURFdriver, yet detailed and exact 3D reconstructions were created.

Chinese word segmentation and its effect on information retrieval

A set of IR experiments was carried out to study the impact of Chinese word segmentation and its effect on information retrieval (IR) at the Division of Information Studies, Nanyang Technological University, Singapore. A total of four automatic character-based segmentation approaches and a manual word segmentation approach was first carried out to obtain the word segments for indexing and to evaluate the segmentation accuracy of these automatic approaches. The IR experiments study both the influence of different document segmentation approaches on IR effectiveness and the methods used for query segmentation. Traditional data recall and precision measures were used to gauge IR effectiveness. A number of queries were selected and subjected to further detailed analysis to further explore the influence of word segmentation on IR.The findings reveal that the segmentation approach has an effect on IR effectiveness. Better IR results are obtained by using the same method for query and document processing as this increase the probability of the query-document match. The recognition of a higher number of 2-character words generally contributes to the improvement of IR effectiveness. However, manual segmentation does not always work better than character-based segmentation as a result of the existence of longer words with more than two characters. No evidence is found that ambiguous words resulting from the segmentation process significantly affect IR.