Tomographic spectroscopic imaging; an experimental proof of concept.

Recording the electron energy loss spectroscopy data cube with a series of energy filtered images is a dose inefficient process because the energy slit blocks most of the electrons. When recording the data cube by scanning an electron probe over the sample, perfect dose efficiency is attained; but due to the low current in nanoprobes, this often is slower, with a smaller field of view. In W. Van den Broek et al. [Ultramicroscopy, 106 (2006) 269], we proposed a new method to record the data cube, which is more dose efficient than an energy filtered series. It produces a set of projections of the data cube and then tomographically reconstructs it. In this article, we demonstrate these projections in practice, we present a simple geometrical model that allows for quantification of the projection angles and we present the first successful experimental reconstruction, all on a standard post-column instrument.

Fusion and merging of multispectral images with use of multiscale fundamental forms.

A new multispectral image wavelet representation is introduced, based on multiscale fundamental forms. This representation describes gradient information of multispectral images in a multiresolution framework. The representation is, in particular, extremely suited for fusion and merging of multispectral images. For fusion as well as for merging, a strategy is described. Experiments are performed on multispectral images, where Landsat Thematic Mapper images are fused and merged with SPOT Panchromatic images. The proposed techniques are compared with wavelet-based techniques described in the literature.

Validation of nuclear texture, density, morphometry and tissue syntactic structure analysis as prognosticators of cervical carcinoma.

OBJECTIVE: To evaluate the performance of karyometry and histometry in the prediction of survival, recurrence and response of early-stage invasive cervical carcinoma. STUDY DESIGN: Nuclear morphometry, chromatin texture and tissue architecture (characterized by syntactic structure analysis) were measured using a semiautomated image analysis system on 46 cases of Feulgen-stained tissue sections. The performance of the features was compared to that of clinical features, reported to be the best prognosticators until now, such as age, lympho-vascular permeation, histologic type, stage and grade. A K nearest neighbor classifier was used for classification. RESULTS: In the prediction of three-year survival, recurrence and response, syntactic structure analysis proved to be the best performer. Classification rates were, respectively, 100%, 94.4% and 94.5%. In all classifications, karyometric and histometric features outperformed clinical features. In general, the best performing features described differences in second-order population statistics (standard deviations). CONCLUSION: The results show that a quantitative analysis based on nuclear morphology, chromatin texture and histology can be considered an excellent aid in the prognosis of invasive cervical carcinoma. The measurements are not hampered by the need to undertake complete resections and are suited to daily practice when implemented in a semiautomated image analysis system.

Wavelets as chromatin texture descriptors for the automated identification of neoplastic nuclei.

Chromatin distribution reflects the organization of the DNA of a nucleus and contains important cellular diagnostic and prognostic information. Feulgen staining of breast tissue enables the chromatin distribution of the nucleus to be visualized in the form of texture. Describing texture in an objective and quantitative way by means of a set of texture parameters, combined with the study of the relationship of such parameters to the pathobiological cell properties, is useful both for reduction of the subjectivity inherently coupled to visual observation and for more accurate prognosis or diagnosis. We have presented an automated classification scheme for the diagnosis and grading of invasive breast cancer. The input to this scheme was a digitized microscopical image, from which nuclei were segmented. Chromatin texture was described using a set of textural parameters that include first- and second-order statistics of the image grey levels. The more recently developed wavelet energy parameters were also included in our study. Classification was performed by a Knn-classifier, which is a versatile multivariate statistical classification technique. We investigated the role of the tissue preparation technique and found that parameters derived from cytospins were better texture descriptors than those from sections. A 100% correct classification was achieved in a patient diagnosis experiment and 82% in a nuclear grading experiment.

Computer-assisted differential diagnosis of malignant mesothelioma based on syntactic structure analysis.

BACKGROUND: Malignant mesothelioma, a mesoderm-derived tumor, is related to asbestos exposure and remains a diagnostic challenge because none of the genetic or immunohistochemical markers have yet been proven to be specific. To assist in the identification of mesothelioma and to differentiate it from other common lesions at the same location, we have tested the performance of syntactic structure analysis (SSA) in an automated classification procedure. MATERIALS AND METHODS: Light-microscopic images of tissue sections of malignant mesothelioma, hyperplastic mesothelium, and adenocarcinoma were analyzed using parameters selected from the Voronoi diagram, Gabriel's graph, and the minimum spanning tree which were classified with a K-nearest-neighbor algorithm. RESULTS: Results showed that mesotheliomas were diagnosed correctly in 74% of the cases; 76% of the adenocarcinomas were correctly graded, and 88% of the mesotheliomas were correctly typed. The performance of the parameters was dependent on the obtained classification (i.e., tumor-tumor versus tumor-benign). CONCLUSIONS: Our results suggest that SSA is valuable in the differential classification of mesothelioma and that it supplements a visually appraised diagnosis. The recognition scores may be increased by a combination of SSA with, for example, cellular or nuclear parameters, measured at higher magnifications to form a solid base for fully automated expert systems.

Value of morphometry, texture analysis, densitometry, and histometry in the differential diagnosis and prognosis of malignant mesothelioma.

Malignant mesothelioma is a tumour with increasing incidence due to widespread use of its causative agent, asbestos, in the past decades. The poor survival necessitates a correct differentiation from other lesions at the same site, such as hyperplastic mesothelium and carcinomas metastatic to pleura or peritoneum. Since genetic and immunohistochemical markers are not absolutely differentiating, the diagnosis is based on the histology complemented with (immuno)histochemistry. However, as the tumour presents itself in numerous heterogeneous histological forms, visual evaluation is extremely difficult. In order to evaluate the prognostic and diagnostic performance of syntactic structure analysis (SSA), chromatin texture analysis, densitometry, and morphometry, an automated KNN-classification system has been used to compare Feulgen-stained tissue sections of hyperplastic mesothelium, malignant mesothelioma, and pulmonary adenocarcinoma. In addition, we also studied most discriminative aspects in the differentiation, typing, and prediction of survival. The results indicate that for the diagnosis of malignant mesothelioma, chromatin texture parameters outperform SSA, densitometry, and morphometry (recognition score=96.8 per cent). Most discriminative parameters highlight spatial patterns of the chromatin distribution that are hard to appraise visually and directly show the benefits of a quantitative approach. Typing of the tumour is best described by SSA parameters, relating to the spatial arrangement of the cells in the tissue (recognition score=94.9 per cent). In survival time classifications, chromatin texture yields the highest recognition score (82.9 per cent), although accurate estimations are unreliable due to a large degree of misclassification.

Statistical texture characterization from discrete wavelet representations.

We conjecture that texture can be characterized by the statistics of the wavelet detail coefficients and therefore introduce two feature sets: (1) the wavelet histogram signatures which capture all first order statistics using a model based approach and (2) the wavelet co-occurrence signatures, which reflect the coefficients' second-order statistics. The introduced feature sets outperform the traditionally used energy. Best performance is achieved by combining histogram and co-occurrence signatures.

Automated breast tumor diagnosis and grading based on wavelet chromatin texture description.

In this paper, wavelets were employed for multi-scale image analysis to extract parameters for the description of chromatin texture in the cytological diagnosis and grading of invasive breast cancer. Their value was estimated by comparing the performance of co-occurrence, densitometric, and morphometric parameters in an automated K-nearest neighbor (Knn) classification scheme based on light microscopic images of isolated nuclei of paraffin-embedded tissue. This design allowed a multifaceted cytological retrospective study of which the practical value can be judged easily. Results show that wavelets perform excellently with classification scores comparable with densitometric and co-occurrence features. Moreover, because wavelets showed a high additive value with the other textural groups, this panel allowed a very profound description with higher recognition scores than previously reported (76% for individual nuclei, 100% for cases). Morphometric parameters performed less well and only slightly increased correct classification. The major drawback, besides image segmentation errors demanding operator supervision, emanated to be the few false-negative cases, which restrict the immediate practical use. However, an enlargement of the parameter set may avoid this misclassification, resulting in an applicable expert system of practical use.

Maximum-likelihood estimation of Rician distribution parameters.

The problem of parameter estimation from Rician distributed data (e.g., magnitude magnetic resonance images) is addressed. The properties of conventional estimation methods are discussed and compared to maximum-likelihood (ML) estimation which is known to yield optimal results asymptotically. In contrast to previously proposed methods, ML estimation is demonstrated to be unbiased for high signal-to-noise ratio (SNR) and to yield physical relevant results for low SNR.

Computer analysis of two-dimensional electrophoresis gels: a new segmentation and modeling algorithm.

The complexity of the spot patterns of two-dimensional electrophoresis gels made it necessary to use image processing techniques to analyze the gels. An important issue in the analysis is the detection and quantification of the protein spots. In this paper we describe a new technique to segment and model the different spots. For the segmentation of a gel into its different spot regions we apply a watershed technique, which is robust and efficient. For the quantification of the spots, a new spot model is constructed, based on diffusion principles. Besides the advantage of having a physical interpretation, the model is demonstrated to be superior to the commonly used Gaussian models.

Watershed-based segmentation of 3D MR data for volume quantization.

The aim of this work is the development of a semiautomatic segmentation technique for efficient and accurate volume quantization of Magnetic Resonance (MR) data. The proposed technique uses a 3D variant of Vincent and Soilles immersion-based watershed algorithm that is applied to the gradient magnitude of the MR data and that produces small volume primitives. The known drawback of the watershed algorithm, oversegmentation, is strongly reduced by a priori application of a 3D adaptive anisotropic diffusion filter to the MR data. Furthermore, oversegmentation is a posteriori reduced by properly merging small volume primitives that have similar gray level distributions. The outcome of the proceeding image processing steps is presented to the user for manual segmentation. Through selection of volume primitives, the user quickly segments of first slice, which contains the object of interest. Afterwards, the subsequent slices are automatically segmented by extrapolation. Segmentation results are contingently manually corrected. The proposed segmentation technique is tested on phantom objects, where segmentation errors less than 2% are observed. In addition, the technique is demonstrated on 3D MR data of the mouse head from which the cerebellum is extracted. Volumes of the mouse cerebellum and the mouse brains in toto are calculated.

Quantification and improvement of the signal-to-noise ratio in a magnetic resonance image acquisition procedure.

A procedure is developed to quantify and improve the signal-to-noise ratio (SNR) of magnetic resonance images. The image SNR is quantified using the correlation function of two independent acquisitions of an image. To test the performance of the quantification, SNR measurement data are fitted to theoretically expected curves. The proposed correlation technique is also used to improve the SNR by estimating the amplitude of the signal spectrum. The technique is applied to a set of MR images, and its performance in terms of gain in SNR, contrast-to-noise ratio (CNR), and resolution loss is compared to that of classical noise filters. The SNR as well as the CNR is improved significantly with minor loss of resolution. Finally, it is shown that the correlation technique can be implemented in a highly efficient way in almost any acquisition procedure of a magnetic resonance imaging system.