Biomedical subjects

Bram van Ginneken

Publications and source records attributed to Bram van Ginneken.

18 recordsLinked to original sources

PubMed2026

Artificial intelligence-based tumour infiltrating lymphocyte quantification in patients with triple-negative breast cancer: an independent validation study.

BACKGROUND: Tumour-infiltrating lymphocytes (TILs) are a robust prognostic marker in patients with triple-negative breast cancer. Artificial intelligence (AI)-derived computational tools assessing TILs could improve efficiency, but require independent validation against clinical outcomes. We aimed to compare the prognostic performance of AI-derived TIL scores with pathologist-scored TILs in a large, prospectively collected dataset pooled from randomised controlled trials. METHODS: CATALINA was an independent, external validation study using prospectively collected long-term clinical outcome data pooled from seven randomised clinical trials conducted at multiple sites. We independently evaluated two previously validated AI pipelines that generate five computationally assessed tumour-infiltrating lymphocyte (cTIL) scores by masked, independent deployment of locked models. cTIL scores were correlated with the mean of the pathologist-scored stromal TILs (sTILs) in 220 digitised haematoxylin and eosin whole slide images in a cohort of patients with early-stage triple-negative or HER-2 positive breast cancer, previously scored by trained pathologists in a TIL-reproducibility study. Prognostic performance was assessed in a separate cohort of patients with early triple-negative breast cancer pooled from seven prospective, randomised adjuvant trials. Multivariable Cox regression models adjusted for clinicopathological factors and study heterogeneity assessed associations of cTIL score and sTIL score with invasive disease-free survival, distant disease-free survival, and overall survival. 5-year discrimination was estimated using time-dependent area under the receiver operating characteristic curve (AUC). FINDINGS: Individual data were collated from 1759 patients, of whom 1356 had complete clinicopathological data, pathologist sTIL scores, and cTIL scores available. Modest correlation (r 0·375-0·473) was observed between cTIL scores and the mean pathologist sTIL score. Both sTIL and cTIL were independently associated with 5-year invasive disease-free survival, distant disease-free survival, and overall survival after adjustment for clinicopathological factors (hazard ratio for invasive disease-free survival was 0·73 [95% CI 0·66-0·82]; q<0·0001, distant disease-free survival was 0·70 [0·61-0·79]; q<0·0001, and overall survival was 0·72 [0·63-0·82]; q<0·0001 for sTIL scores and 0·80 [0·73-0·89]; q<0·0001, 0·77 [0·69-0·86]; q<0·0001, and 0·79 [0·70-0·88]; q=0·0002, respectively, for percentage_lymphocyte scores). In models adjusted for clinicopathological variables and sTIL score, cTIL score did not maintain a statistically significant prognostic association. Both sTIL and cTIL scores improved the 5-year AUC over clinicopathological variables alone, while cTIL score did not significantly further improve AUC when combined with clinicopathological variables and sTIL score. INTERPRETATION: Two cTIL models deployed entirely without retraining or modification provided statistically significant prognostic information and improved risk discrimination compared with clinicopathological variables alone in this large, platform-based, independent validation study. Although cTIL score did not incrementally improve prognostication compared with models combining clinicopathological variables with sTIL score, these findings support the application of cTILs as a reproducible prognostic biomarker, particularly in settings where routine or widespread pathologist assessment is unavailable. FUNDING: Breast Cancer Research Foundation (USA).

Humans↗

PubMed2006

Automatic rib segmentation and labeling in computed tomography scans using a general framework for detection, recognition and segmentation of objects in volumetric data.

A system for automatic segmentation and labeling of the complete rib cage in chest CT scans is presented. The method uses a general framework for automatic detection, recognition and segmentation of objects in three-dimensional medical images. The framework consists of five stages: (1) detection of relevant image structures, (2) construction of image primitives, (3) classification of the primitives, (4) grouping and recognition of classified primitives and (5) full segmentation based on the obtained groups. For this application, first 1D ridges are extracted in 3D data. Then, primitives in the form of line elements are constructed from the ridge voxels. Next a classifier is trained to classify the primitives in foreground (ribs) and background. In the grouping stage centerlines are formed from the foreground primitives and rib numbers are assigned to the centerlines. In the final segmentation stage, the centerlines act as initialization for a seeded region growing algorithm. The method is tested on 20 CT-scans. Of the primitives, 97.5% is classified correctly (sensitivity is 96.8%, specificity is 97.8%). After grouping, 98.4% of the ribs are recognized. The final segmentation is qualitatively evaluated and is very accurate for over 80% of all ribs, with slight errors otherwise.

Algorithms↗

PubMed2006

Segmentation of anatomical structures in chest radiographs using supervised methods: a comparative study on a public database.

The task of segmenting the lung fields, the heart, and the clavicles in standard posterior-anterior chest radiographs is considered. Three supervised segmentation methods are compared: active shape models, active appearance models and a multi-resolution pixel classification method that employs a multi-scale filter bank of Gaussian derivatives and a k-nearest-neighbors classifier. The methods have been tested on a publicly available database of 247 chest radiographs, in which all objects have been manually segmented by two human observers. A parameter optimization for active shape models is presented, and it is shown that this optimization improves performance significantly. It is demonstrated that the standard active appearance model scheme performs poorly, but large improvements can be obtained by including areas outside the objects into the model. For lung field segmentation, all methods perform well, with pixel classification giving the best results: a paired t-test showed no significant performance difference between pixel classification and an independent human observer. For heart segmentation, all methods perform comparably, but significantly worse than a human observer. Clavicle segmentation is a hard problem for all methods; best results are obtained with active shape models, but human performance is substantially better. In addition, several hybrid systems are investigated. For heart segmentation, where the separate systems perform comparably, significantly better performance can be obtained by combining the results with majority voting. As an application, the cardio-thoracic ratio is computed automatically from the segmentation results. Bland and Altman plots indicate that all methods perform well when compared to the gold standard, with confidence intervals from pixel classification and active appearance modeling very close to those of a human observer. All results, including the manual segmentations, have been made publicly available to facilitate future comparative studies.

Algorithms↗

PubMed2006

Image structure clustering for image quality verification of color retina images in diabetic retinopathy screening.

Reliable verification of image quality of retinal screening images is a prerequisite for the development of automatic screening systems for diabetic retinopathy. A system is presented that can automatically determine whether the quality of a retinal screening image is sufficient for automatic analysis. The system is based on the assumption that an image of sufficient quality should contain particular image structures according to a certain pre-defined distribution. We cluster filterbank response vectors to obtain a compact representation of the image structures found within an image. Using this compact representation together with raw histograms of the R, G, and B color planes, a statistical classifier is trained to distinguish normal from low quality images. The presented system does not require any previous segmentation of the image in contrast with previous work. The system was evaluated on a large, representative set of 1000 images obtained in a screening program. The proposed method, using different feature sets and classifiers, was compared with the ratings of a second human observer. The best system, based on a Support Vector Machine, has performance close to optimal with an area under the ROC curve of 0.9968.

Color↗

PubMed2006

Computer analysis of computed tomography scans of the lung: a survey.

Current computed tomography (CT) technology allows for near isotropic, submillimeter resolution acquisition of the complete chest in a single breath hold. These thin-slice chest scans have become indispensable in thoracic radiology, but have also substantially increased the data load for radiologists. Automating the analysis of such data is, therefore, a necessity and this has created a rapidly developing research area in medical imaging. This paper presents a review of the literature on computer analysis of the lungs in CT scans and addresses segmentation of various pulmonary structures, registration of chest scans, and applications aimed at detection, classification and quantification of chest abnormalities. In addition, research trends and challenges are identified and directions for future research are discussed.

Algorithms↗

PubMed2006

Local noise weighted filtering for emphysema scoring of low-dose CT images.

Computed tomography (CT) has become the new reference standard for quantification of emphysema. The most popular measure of emphysema derived from CT is the pixel index (PI), which expresses the fraction of the lung volume with abnormally low intensity values. As PI is calculated from a single, fixed threshold on intensity, this measure is strongly influenced by noise. This effect shows up clearly when comparing the PI score of a high-dose scan to the PI score of a low-dose (i.e., noisy) scan of the same subject. In this paper, the noise variance (NOVA) filter is presented: a general framework for (iterative) nonlinear filtering, which uses an estimate of the spatially dependent noise variance in an image. The NOVA filter iteratively estimates the local image noise and filters the image. For the specific purpose of emphysema quantification of low-dose CT images, a dedicated, noniterative NOVA filter is constructed by using prior knowledge of the data to obtain a good estimate of the spatially dependent noise in an image. The performance of the NOVA filter is assessed by comparing characteristics of pairs of high-dose and low-dose scans. The compared characteristics are the PI scores for different thresholds and the size distributions of emphysema bullae. After filtering, the PI scores of high-dose and low-dose images agree to within 2%-3% points. The reproducibility of the high-dose bullae size distribution is also strongly improved. NOVA filtering of a CT image of typically 400 x 512 x 512 voxels takes only a couple of minutes which makes it suitable for routine use in clinical practice.

Algorithms↗

PubMed2006

Segmentation of the posterior ribs in chest radiographs using iterated contextual pixel classification.

The task of segmenting the posterior ribs within the lung fields of standard posteroanterior chest radiographs is considered. To this end, an iterative, pixel-based, supervised, statistical classification method is used, which is called iterated contextual pixel classification (ICPC). Starting from an initial rib segmentation obtained from pixel classification, ICPC updates it by reclassifying every pixel, based on the original features and, additionally, class label information of pixels in the neighborhood of the pixel to be reclassified. The method is evaluated on 30 radiographs taken from the JSRT (Japanese Society of Radiological Technology) database. All posterior ribs within the lung fields in these images have been traced manually by two observers. The first observer's segmentations are set as the gold standard; ICPC is trained using these segmentations. In a sixfold cross-validation experiment, ICPC achieves a classification accuracy of 0.86 +/- 0.06, as compared to 0.94 +/- 0.02 for the second human observer.

Algorithms↗

PubMed2006

Automated classification of hyperlucency, fibrosis, ground glass, solid, and focal lesions in high-resolution CT of the lung.

An automatic method for textural analysis of complete HRCT lung slices is presented. The system performs classification of regions of interest (ROIs) into one of six classes: normal, hyperlucency, fibrosis, ground glass, solid, and focal. We propose a novel method of automatically generating ROIs that contain homogeneous texture. The use of such regions rather than square regions is shown to improve performance of the automated system. Furthermore, the use of two different, previously published, feature sets is investigated. Both feature sets are shown to yield similar results. Classification performance of the complete system is characterized by ROC curves for each of the classes of abnormality and compared to a total of three expert readings by two experienced radiologists. The different types of abnormality can be automatically distinguished with areas under the ROC curve that range from 0.74 (focal) to 0.95 (solid). The kappa statistics for intraobserver agreement, interobserver agreement, and computer versus observer agreement were 0.70, 0.53+/-0.02, and 0.40+/-0.03, respectively. The question whether or not a class of abnormality was present in a slice could be answered by the computer system with an accuracy comparable to that of radiologists.

Automation↗

PubMed2005

A computer-aided diagnosis system for detection of lung nodules in chest radiographs with an evaluation on a public database.

A computer algorithm for nodule detection in chest radiographs is presented. The algorithm consists of four main steps: (i) image preprocessing; (ii) nodule candidate detection; (iii) feature extraction; (iv) candidate classification. Two optional extensions to this scheme are tested: candidate selection and candidate segmentation. The output of step (ii) is a list of circles, which can be transformed into more detailed contours by the extra candidate segmentation step. In addition, the candidate selection step (which is a classification step using a small number of features) can be used to reduce the list of nodule candidates before step (iii). The algorithm uses multi-scale techniques in several stages of the scheme: Candidates are found by looking for local intensity maxima in Gaussian scale space; nodule boundaries are detected by tracing edge points found at large scales down to pixel scale; some of the features used for classification are taken from a multi-scale Gaussian filterbank. Experiments with this scheme (with and without the segmentation and selection steps) are carried out on a previously characterized, publicly available database, that contains a large number of very subtle nodules. For this database, counting as detections only those nodules that were indicated with a confidence level of 50% or more, radiologists previously detected 70% of the nodules. For our algorithm, it turns out that the selection step does have an added value for the system, while segmentation does not lead to a clear improvement. With the scheme with the best performance, accepting on average two false positives per image results in the identification of 51% of all nodules. For four false positives, this increases to 67%. This is close to the previously reported 70% detection rate of the radiologists.

Algorithms↗

PubMed2005

Automatic detection of red lesions in digital color fundus photographs.

The robust detection of red lesions in digital color fundus photographs is a critical step in the development of automated screening systems for diabetic retinopathy. In this paper, a novel red lesion detection method is presented based on a hybrid approach, combining prior works by Spencer et al. (1996) and Frame et al. (1998) with two important new contributions. The first contribution is a new red lesion candidate detection system based on pixel classification. Using this technique, vasculature and red lesions are separated from the background of the image. After removal of the connected vasculature the remaining objects are considered possible red lesions. Second, an extensive number of new features are added to those proposed by Spencer-Frame. The detected candidate objects are classified using all features and a k-nearest neighbor classifier. An extensive evaluation was performed on a test set composed of images representative of those normally found in a screening set. When determining whether an image contains red lesions the system achieves a sensitivity of 100% at a specificity of 87%. The method is compared with several different automatic systems and is shown to outperform them all. Performance is close to that of a human expert examining the images for the presence of red lesions.

Algorithms↗

PubMed2005

Toward automated segmentation of the pathological lung in CT.

Conventional methods of lung segmentation rely on a large gray value contrast between lung fields and surrounding tissues. These methods fail on scans with lungs that contain dense pathologies, and such scans occur frequently in clinical practice. We propose a segmentation-by-registration scheme in which a scan with normal lungs is elastically registered to a scan containing pathology. When the resulting transformation is applied to a mask of the normal lungs, a segmentation is found for the pathological lungs. As a mask of the normal lungs, a probabilistic segmentation built up out of the segmentations of 15 registered normal scans is used. To refine the segmentation, voxel classification is applied to a certain volume around the borders of the transformed probabilistic mask. Performance of this scheme is compared to that of three other algorithms: a conventional, a user-interactive and a voxel classification method. The algorithms are tested on 10 three-dimensional thin-slice computed tomography volumes containing high-density pathology. The resulting segmentations are evaluated by comparing them to manual segmentations in terms of volumetric overlap and border positioning measures. The conventional and user-interactive methods that start off with thresholding techniques fail to segment the pathologies and are outperformed by both voxel classification and the refined segmentation-by-registration. The refined registration scheme enjoys the additional benefit that it does not require pathological (hand-segmented) training data.

Algorithms↗

PubMed2004

Interactive segmentation of abdominal aortic aneurysms in CTA images.

A model-based approach to interactive segmentation of abdominal aortic aneurysms from CTA data is presented. After manual delineation of the aneurysm sac in the first slice, the method automatically detects the contour in subsequent slices, using the result from the previous slice as a reference. If an obtained contour is not sufficiently accurate, the user can intervene and provide an additional manual reference contour. The method is inspired by the active shape model (ASM) segmentation scheme (), in which a statistical shape model, derived from corresponding landmark points in manually labeled training images, is fitted to the image in an iterative manner. In our method, a shape model of the contours in two adjacent image slices is progressively fitted to the entire volume. The contour obtained in one slice thus constrains the possible shapes in the next slice. The optimal fit is determined on the basis of multi-resolution gray level models constructed from gray value patches sampled around each landmark. We propose to use the similarity of adjacent image slices for this gray level model, and compare these to single-slice features that are more generally used with ASM. The performance of various image features is evaluated in leave-one-out experiments on 23 data sets. Features that use the similarity of adjacent image slices outperform measures based on single-slice features in all cases. The average number of slices in our datasets is 51, while on average eight manual initializations are required, which decreases operator segmentation time by a factor of 6.

Algorithms↗

PubMed2004

Automatic detection of calcifications in the aorta from CT scans of the abdomen. 3D computer-aided diagnosis.

RATIONALE AND OBJECTIVES: Automated detection and quantification of arterial calcifications can facilitate epidemiologic research and, eventually, the use of full-body calcium scoring in clinical practice. An automatic computerized method to detect calcifications in CT scans is presented. MATERIALS AND METHODS: Forty abdominal CT scans have been randomly selected from clinical practice. They all contained contrast material and belonged to one of four categories: containing "no," "small," "moderate," or "large" amounts of arterial calcification. There were ten scans in each category. The experiments were restricted to the vertical range from the point where the superior mesenteric artery branches off of the descending aorta until the first bifurcation of the iliac arteries. The automatic method starts by extracting all connected objects above 220 Hounsfield units (HU) from the scan. These objects include all calcifications, as well as bony structures and contrast material. To distinguish calcifications from non-calcifications, a number of features are calculated for each object. These features are based on the object's size, location, shape characteristics, and surrounding structures. Subsequently a classification of each object is performed in two stages. First the probability that an object represents a calcification is computed assuming a multivariate Gaussian distribution for the calcifications. Objects with low probability are discarded. The remaining objects are then classified into calcifications and non-calcifications using a 5-nearest-neighbor classifier and sequential forward feature selection. Based on the total volume of calcifications determined by the system, the scan is assigned to one of the four categories mentioned above. RESULTS: The 40 scans contained a total of 249 calcifications as determined by a human observer. The method detected 209 calcifications (sensitivity 83.9%) at the expense of on average 1.0 false-positive object per scan. The correct category label was assigned to 30 scans and only 2 scans were off by more than one category. Most incorrect classifications can be attributed to the presence of contrast material in the scans. CONCLUSION: It is possible to identify the majority of arterial calcifications in abdominal CT scans in a completely automatic fashion with few false positive objects, even if the scans contain contrast material.

Aorta, Abdominal↗

PubMed2004

Ridge-based vessel segmentation in color images of the retina.

A method is presented for automated segmentation of vessels in two-dimensional color images of the retina. This method can be used in computer analyses of retinal images, e.g., in automated screening for diabetic retinopathy. The system is based on extraction of image ridges, which coincide approximately with vessel centerlines. The ridges are used to compose primitives in the form of line elements. With the line elements an image is partitioned into patches by assigning each image pixel to the closest line element. Every line element constitutes a local coordinate frame for its corresponding patch. For every pixel, feature vectors are computed that make use of properties of the patches and the line elements. The feature vectors are classified using a kappaNN-classifier and sequential forward feature selection. The algorithm was tested on a database consisting of 40 manually labeled images. The method achieves an area under the receiver operating characteristic curve of 0.952. The method is compared with two recently published rule-based methods of Hoover et al. and Jiang et al. The results show that our method is significantly better than the two rule-based methods (p < 0.01). The accuracy of our method is 0.944 versus 0.947 for a second observer.

Algorithms↗

PubMed2003

Adapting Active Shape Models for 3D segmentation of tubular structures in medical images.

Active Shape Models (ASM) have proven to be an effective approach for image segmentation. In some applications, however, the linear model of gray level appearance around a contour that is used in ASM is not sufficient for accurate boundary localization. Furthermore, the statistical shape model may be too restricted if the training set is limited. This paper describes modifications to both the shape and the appearance model of the original ASM formulation. Shape model flexibility is increased, for tubular objects, by modeling the axis deformation independent of the cross-sectional deformation, and by adding supplementary cylindrical deformation modes. Furthermore, a novel appearance modeling scheme that effectively deals with a highly varying background is developed. In contrast with the conventional ASM approach, the new appearance model is trained on both boundary and non-boundary points, and the probability that a given point belongs to the boundary is estimated non-parametrically. The methods are evaluated on the complex task of segmenting thrombus in abdominal aortic aneurysms (AAA). Shape approximation errors were successfully reduced using the two shape model extensions. Segmentation using the new appearance model significantly outperformed the original ASM scheme; average volume errors are 5.1% and 45% respectively.

Algorithms↗

PubMed2003

Computer-aided diagnosis in high resolution CT of the lungs.

A computer-aided diagnosis (CAD) system is presented to automatically distinguish normal from abnormal tissue in high-resolution CT chest scans acquired during daily clinical practice. From high-resolution computed tomography scans of 116 patients, 657 regions of interest are extracted that are to be classified as displaying either normal or abnormal lung tissue. A principled texture analysis approach is used, extracting features to describe local image structure by means of a multi-scale filter bank. The use of various classifiers and feature subsets is compared and results are evaluated with ROC analysis. Performance of the system is shown to approach that of two expert radiologists in diagnosing the local regions of interest, with an area under the ROC curve of 0.862 for the CAD scheme versus 0.877 and 0.893 for the radiologists.

Algorithms↗

PubMed2002

Automatic detection of abnormalities in chest radiographs using local texture analysis.

A fully automatic method is presented to detect abnormalities in frontal chest radiographs which are aggregated into an overall abnormality score. The method is aimed at finding abnormal signs of a diffuse textural nature, such as they are encountered in mass chest screening against tuberculosis (TB). The scheme starts with automatic segmentation of the lung fields, using active shape models. The segmentation is used to subdivide the lung fields into overlapping regions of various sizes. Texture features are extracted from each region, using the moments of responses to a multiscale filter bank. Additional "difference features" are obtained by subtracting feature vectors from corresponding regions in the left and right lung fields. A separate training set is constructed for each region. All regions are classified by voting among the k nearest neighbors, with leave-one-out. Next, the classification results of each region are combined, using a weighted multiplier in which regions with higher classification reliability weigh more heavily. This produces an abnormality score for each image. The method is evaluated on two databases. The first database was collected from a TB mass chest screening program, from which 147 images with textural abnormalities and 241 normal images were selected. Although this database contains many subtle abnormalities, the classification has a sensitivity of 0.86 at a specificity of 0.50 and an area under the receiver operating characteristic (ROC) curve of 0.820. The second database consist of 100 normal images and 100 abnormal images with interstitial disease. For this database, the results were a sensitivity of 0.97 at a specificity of 0.90 and an area under the ROC curve of 0.986.

Algorithms↗

PubMed2002

Active shape model segmentation with optimal features.

An active shape model segmentation scheme is presented that is steered by optimal local features, contrary to normalized first order derivative profiles, as in the original formulation [Cootes and Taylor, 1995, 1999, and 2001]. A nonlinear kNN-classifier is used, instead of the linear Mahalanobis distance, to find optimal displacements for landmarks. For each of the landmarks that describe the shape, at each resolution level taken into account during the segmentation optimization procedure, a distinct set of optimal features is determined. The selection of features is automatic, using the training images and sequential feature forward and backward selection. The new approach is tested on synthetic data and in four medical segmentation tasks: segmenting the right and left lung fields in a database of 230 chest radiographs, and segmenting the cerebellum and corpus callosum in a database of 90 slices from MRI brain images. In all cases, the new method produces significantly better results in terms of an overlap error measure (p < 0.001 using a paired T-test) than the original active shape model scheme.

Adolescent↗