Ing. Denis Baručić

Metagenomics provides a tool to assess the functional potential of environmental and host-associated microbiomes based on the analysis of environmental DNA: assembly, gene prediction and annotation. While gene prediction is straightforward for most bacterial and archaeal taxa, it has limited applicability in the majority of eukaryotic organisms, including fungi that contain introns in gene coding sequences. As a consequence, eukaryotic genes are underrepresented in metagenomics datasets and our understanding of the contribution of fungi and other eukaryotes to microbiome functioning is limited. Here, we developed a machine intelligence-based algorithm that predicts fungal introns in environmental DNA with reasonable precision and used it to improve the annotation of environmental metagenomes. Intron removal increased the number of predicted genes by up to 9.1% and improved the annotation of several others. The proportion of newly predicted genes increased with the share of eukaryotic genes in the metagenome and—within fungal taxa—increased with the number of introns per gene. Our approach provides a tool named SVMmycointron for improved metagenome annotation, especially of microbiomes with a high proportion of eukaryotes. The scripts described in the paper are made publicly available and can be readily utilized by microbiome researchers analysing metagenomics data.

Measuring object size is fast and a standard part of many radiological evaluation procedures. We describe a deep learning segmentation method that can be trained on a small number of pixel-wise reference segmentation and then fine-tuned from the weak annotations of the object thickness. The difficulty is in the non-differentiability of the thickness function defined using the pixel-wise distance transform. We overcome it by optimizing the expected value of the loss function after the injection of a virtual random noise. Further speedup is possible using the properties of the distance transform. We demonstrate the benefit of the proposed method on ultrasound images of the carotid artery. The fine-tuning improves the performance by about 10% IoU.

In this work, we classify chemotherapeutic agents (topoisomerase inhibitors) based on their effect on U-2 OS cells. We use phase-contrast microscopy images, which are faster and easier to obtain than fluorescence images and support live cell imaging. We use a convolutional neural network (CNN) trained end-to-end directly on the input images without requiring for manual segmentations or any other auxiliary data. Our method can distinguish between tested cytotoxic drugs with an accuracy of 98%, provided that their mechanism of action differs, outperforming previous work. The results are even better when substance-specific concentrations are used. We show the benefit of sharing the extracted features over all classes (drugs). Finally, a 2D visualization of these features reveals clusters, which correspond well to known class labels, suggesting the possible use of our methodology for drug discovery application in analyzing new, unseen drugs.

In learning from label proportions (LLP), the instances are grouped into bags, and the task is to learn an instance classifier given relative class proportions in training bags. LLP is useful when obtaining individual instance labels is impossible or costly. In this work, we focus on the case of small bags, which allows to design an algorithm that explicitly considers all consistent instance label combinations. In particular, we propose an EM algorithm alternating between optimizing a general neural network instance classifier and incorporating bag-level annotations. Using two different image datasets, we experimentally compare this method with an approach based on normal approximation and two existing LLP methods. The results show that our approach converges faster to a comparable or better solution.

Deep learning has proved particularly useful for semantic segmentation, a fundamental image analysis task. However, the standard deep learning methods need many training images with ground-truth pixel-wise annotations, which are usually laborious to obtain and, in some cases (e.g., medical images), require domain expertise. Therefore, instead of pixel-wise annotations, we focus on image annotations that are significantly easier to acquire but still informative, namely the size of foreground objects. We define the object size as the maximum Chebyshev distance between a foreground and the nearest background pixel. We propose an algorithm for training a deep segmentation network from a dataset of a few pixel-wise annotated images and many images with known object sizes. The algorithm minimizes a discrete (non-differentiable) loss function defined over the object sizes by sampling the gradient and then using the standard back-propagation algorithm. Experiments show that the new approach improves the segmentation performance.

Primary aldosteronism (PA) is the most frequent cause of secondary hypertension. Early diagnoses of PA are essential to avoid the long-term negative effects of elevated aldosterone concentration on the cardiovascular and renal system. In this work, we study the texture of the carotid artery vessel wall from longitudinal ultrasound images in order to automatically distinguish between PA and essential hypertension (EH). The texture is characterized using 140 Haralick and 10 wavelet features evaluated in a region of interest in the vessel wall, followed by the XGBoost classifier. Carotid ultrasound studies were carried out on 33 patients aged 42–72 years with PA, 52 patients with EH, and 33 normotensive controls. For the most clinically relevant task of distinguishing PA and EH classes, we achieved a classification accuracy of 73% as assessed by a leave-one-out procedure. This result is promising even compared to the 57% prediction accuracy using clinical characteristics alone or 63% accuracy using a combination of clinical characteristics and intima-media thickness (IMT) parameters. If the accuracy is improved and the method incorporated into standard clinical procedures, this could eventually lead to an improvement in the early diagnosis of PA and consequently improve the clinical outcome for these patients in future.

Infertility is becoming an issue for an increasing number of couples. The most common solution, in vitro fertilization, requires embryologists to carefully examine light microscopy images of human oocytes to determine their developmental potential. We propose an automatic system to improve the speed, repeatability, and accuracy of this process. We first localize individual oocytes and identify their principal components using CNN (U-Net) segmentation. Next, we calculate several descriptors based on geometry and texture. The final step is an SVM classifier. Both the segmentation and classification training is based on expert annotations. The presented approach leads to a classification accuracy of 70%.