Philipp Kainz

Body composition in sport: a comparison of a novel ultrasound imaging technique to measure subcutaneous fat tissue compared with skinfold measurement

Background Extremely low weight and rapid changes in weight and body composition have become major concerns in many sports, but sufficiently accurate field methods for body composition assessment in athletes are missing. This study aimed to explore the use of ultrasound methods for assessment of body fat content in athletes. Methods 19 female athletes (stature: 1.67(±0.06) m, weight: 59.6(±7.6) kg; age: 19.5(±3.3) years) were investigated by three observers using a novel ultrasound method for thickness measurement of uncompressed subcutaneous adipose tissue and of embedded structures. Two observers also measured skinfold thickness at eight International Society for the Advancement of Kinanthrometry (ISAK) sites; mean skinfold values were compared to mean subcutaneous adipose tissue thicknesses measured by ultrasound. Interobserver reliability of imaging and evaluation obtained by this ultrasound technique: intraclass correlation coefficient ICC=0.968 (95% CI 0.957 to 0.977); evaluation of given images: ICC=0.997 (0.993 to 0.999). Results Skinfold compared to ultrasound thickness showed that compressibility of subcutaneous adipose tissue depends largely on the site and the person: regression slopes ranged from 0.61 (biceps) to 1.59 (thigh) and CIs were large. Limits of agreement ranged from 2.6 to 8.6 mm. Regression lines did not intercept the skinfold axis at zero because of the skin thickness being included in the skinfold. The four ISAK trunk sites caused ultrasound imaging problems in 13 of 152 sites (8 ISAK sites, 19 athletes). Conclusions The ultrasound method allows measurement of uncompressed subcutaneous adipose tissue thickness with an accuracy of 0.1–0.5 mm, depending on the probe frequency. Compressibility of the skinfold depends on the anatomical site, and skin thickness varies by a factor of two. This inevitably limits the skinfold methods for body fat estimation. Ultrasound accuracy for subcutaneous adipose tissue measurement is limited by the plasticity of fat and furrowed tissue borders. Comparative US measurements show that skinfold measurements do not allow accurate assessment of subcutaneous adipose tissue thickness.

Collaborative analysis of multi-gigapixel imaging data using Cytomine

Motivation: Collaborative analysis of massive imaging datasets is essential to enable scientific discoveries. Results: We developed Cytomine to foster active and distributed collaboration of multidisciplinary teams for large-scale image-based studies. It uses web development methodologies and machine learning in order to readily organize, explore, share and analyze (semantically and quantitatively) multi-gigapixel imaging data over the internet. We illustrate how it has been used in several biomedical applications. Availability and implementation: Cytomine (http://www.cytomine.be/) is freely available under an open-source license from http://github.com/cytomine/. A documentation wiki (http://doc.cytomine.be) and a demo server (http://demo.cytomine.be) are also available. Contact: info@cytomine.be Supplementary information: Supplementary data are available at Bioinformatics online.

You Should Use Regression to Detect Cells

Automated cell detection in histopathology images is a hard problem due to the large variance of cell shape and appearance. We show that cells can be detected reliably in images by predicting, for each pixel location, a monotonous function of the distance to the center of the closest cell. Cell centers can then be identified by extracting local extremums of the predicted values. This approach results in a very simple method, which is easy to implement. We show on two challenging microscopy image datasets that our approach outperforms state-of-the-art methods in terms of accuracy, reliability, and speed. We also introduce a new dataset that we will make publicly available.

Body composition in sport: interobserver reliability of a novel ultrasound measure of subcutaneous fat tissue

Background Very low body mass, extreme mass changes, and extremely low per cent body fat are becoming increasingly common in many sports, but sufficiently reliable and accurate field methods for body composition assessment in athletes are missing. Methods Nineteen female athletes were investigated (mean (SD) age: 19.5 (±3.3) years; body mass: 59.6 (±7.6) kg; height: 1.674 (±0.056) m; BMI: 21.3 (±2.3) kg/m2). Three observers applied diagnostic B-mode-ultrasound (US) combined with the evaluation software for subcutaneous adipose tissue measurements at eight ISAK sites (International Society for the Advancement of Kinanthrometry). Regression and reliability analyses are presented. Results US measurements and evaluation of subcutaneous adipose tissue (SAT) thicknesses (including fibrous structures: Dincluded; n=378) resulted in an SE of estimate SEE=0.60 mm, R2=0.98 (p<0.001), limit of agreement LOA=1.18, ICC=0.968 (0.957–0.977). Similar values were found for Dexcluded: SEE=0.68 mm, R2=0.97 (p<0.001). Dincluded at individual ISAK sites: at biceps, R2=0.87 and intraclass-correlation coefficient ICC=0.811 were lowest and SEE=0.79 mm was highest. Values at all other sites ranged from R2: 0.94–0.99, SEE: 0.42–0.65 mm, and ICC: 0.917–0.985. Interobserver coefficients ranged from 0.92 to 0.99, except for biceps (0.74, 0.83 and 0.87). Evaluations of 20 randomly selected US images by three observers (Dincluded) resulted in: SEE=0.15 mm, R2=0.998(p<0.001), ICC=0.997 (0.993, 0999). Conclusions Subject to optimal choice of sites and certain standardisations, US can offer a highly reliable field method for measurement of uncompressed thickness of the SAT. High accuracy and high reliability of measurement, as obtained with this US approach, are essential for protection of the athlete’s health and also for optimising performance.

Subcutaneous fat patterning in athletes: selection of appropriate sites and standardisation of a novel ultrasound measurement technique: ad hoc working group on body composition, health and performance, under the auspices of the IOC Medical Commission

Background Precise and accurate field methods for body composition analyses in athletes are needed urgently. Aim Standardisation of a novel ultrasound (US) technique for accurate and reliable measurement of subcutaneous adipose tissue (SAT). Methods Three observers captured US images of uncompressed SAT in 12 athletes and applied a semiautomatic evaluation algorithm for multiple SAT measurements. Results Eight new sites are recommended: upper abdomen, lower abdomen, erector spinae, distal triceps, brachioradialis, lateral thigh, front thigh, medial calf. Obtainable accuracy was 0.2 mm (18 MHz probe; speed of sound: 1450 m/s). Reliability of SAT thickness sums (N=36): R2=0.998, SEE=0.55 mm, ICC (95% CI) 0.998 (0.994 to 0.999); observer differences from their mean: 95% of the SAT thickness sums were within ±1 mm (sums of SAT thicknesses ranged from 10 to 50 mm). Embedded fibrous tissues were also measured. Conclusions A minimum of eight sites is suggested to accommodate inter-individual differences in SAT patterning. All sites overlie muscle with a clearly visible fascia, which eases the acquisition of clear images and the marking of these sites takes only a few minutes. This US method reaches the fundamental accuracy and precision limits for SAT measurements given by tissue plasticity and furrowed borders, provided the measurers are trained appropriately.

IQM: An Extensible and Portable Open Source Application for Image and Signal Analysis in Java

Image and signal analysis applications are substantial in scientific research. Both open source and commercial packages provide a wide range of functions for image and signal analysis, which are sometimes supported very well by the communities in the corresponding fields. Commercial software packages have the major drawback of being expensive and having undisclosed source code, which hampers extending the functionality if there is no plugin interface or similar option available. However, both variants cannot cover all possible use cases and sometimes custom developments are unavoidable, requiring open source applications. In this paper we describe IQM, a completely free, portable and open source (GNU GPLv3) image and signal analysis application written in pure Java. IQM does not depend on any natively installed libraries and is therefore runnable out-of-the-box. Currently, a continuously growing repertoire of 50 image and 16 signal analysis algorithms is provided. The modular functional architecture based on the three-tier model is described along the most important functionality. Extensibility is achieved using operator plugins, and the development of more complex workflows is provided by a Groovy script interface to the JVM. We demonstrate IQM’s image and signal processing capabilities in a proof-of-principle analysis and provide example implementations to illustrate the plugin framework and the scripting interface. IQM integrates with the popular ImageJ image processing software and is aiming at complementing functionality rather than competing with existing open source software. Machine learning can be integrated into more complex algorithms via the WEKA software package as well, enabling the development of transparent and robust methods for image and signal analysis.

Multi-factorial Age Estimation from Skeletal and Dental MRI Volumes

Age estimation from radiologic data is an important topic in forensic medicine to assess chronological age or to discriminate minors from adults, e.g. asylum seekers lacking valid identification documents. In this work we propose automatic multi-factorial age estimation methods based on MRI data to extend the maximal age range from 19 years, as commonly used for age assessment based on hand bones, up to 25 years, when combined with wisdom teeth and clavicles. Mimicking how radiologists perform age estimation, our proposed method based on deep convolutional neural networks achieves a result of \(1.14 \pm 0.96\) years of mean absolute error in predicting chronological age. Further, when fine-tuning the same network for majority age classification, we show an improvement in sensitivity of the multi-factorial system compared to solely relying on the hand.

Segmentation and classification of colon glands with deep convolutional neural networks and total variation regularization

Segmentation of histopathology sections is a necessary preprocessing step for digital pathology. Due to the large variability of biological tissue, machine learning techniques have shown superior performance over conventional image processing methods. Here we present our deep neural network-based approach for segmentation and classification of glands in tissue of benign and malignant colorectal cancer, which was developed to participate in the GlaS@MICCAI2015 colon gland segmentation challenge. We use two distinct deep convolutional neural networks (CNN) for pixel-wise classification of Hematoxylin-Eosin stained images. While the first classifier separates glands from background, the second classifier identifies gland-separating structures. In a subsequent step, a figure-ground segmentation based on weighted total variation produces the final segmentation result by regularizing the CNN predictions. We present both quantitative and qualitative segmentation results on the recently released and publicly available Warwick-QU colon adenocarcinoma dataset associated with the GlaS@MICCAI2015 challenge and compare our approach to the simultaneously developed other approaches that participated in the same challenge. On two test sets, we demonstrate our segmentation performance and show that we achieve a tissue classification accuracy of 98% and 95%, making use of the inherent capability of our system to distinguish between benign and malignant tissue. Our results show that deep learning approaches can yield highly accurate and reproducible results for biomedical image analysis, with the potential to significantly improve the quality and speed of medical diagnoses.

Robust Bone Marrow Cell Discrimination by Rotation-Invariant Training of Multi-class Echo State Networks

Classification of cell types in context of the architecture in tissue specimen is the basis of diagnostic pathology and decisions for comprehensive investigations rely on a valid interpretation of tissue morphology. Especially visual examination of bone marrow cells takes a considerable amount of time and inter-observer variability can be remarkable. In this work, we propose a novel rotation-invariant learning scheme for multi-class Echo State Networks (ESNs), which achieves very high performance in automated bone marrow cell classification. Based on representing static images as temporal sequence of rotations, we show how ESNs robustly recognize cells of arbitrary rotations by taking advantage of their short-term memory capacity.

Training echo state networks for rotation-invariant bone marrow cell classification

The main principle of diagnostic pathology is the reliable interpretation of individual cells in context of the tissue architecture. Especially a confident examination of bone marrow specimen is dependent on a valid classification of myeloid cells. In this work, we propose a novel rotation-invariant learning scheme for multi-class echo state networks (ESNs), which achieves very high performance in automated bone marrow cell classification. Based on representing static images as temporal sequence of rotations, we show how ESNs robustly recognize cells of arbitrary rotations by taking advantage of their short-term memory capacity. The performance of our approach is compared to a classification random forest that learns rotation-invariance in a conventional way by exhaustively training on multiple rotations of individual samples. The methods were evaluated on a human bone marrow image database consisting of granulopoietic and erythropoietic cells in different maturation stages. Our ESN approach to cell classification does not rely on segmentation of cells or manual feature extraction and can therefore directly be applied to image data.