GRUU-Net: Integrated convolutional and gated recurrent neural network for cell segmentation

Cell segmentation in microscopy images is a common and challenging task. In recent years, deep neural networks achieved remarkable improvements in the field of computer vision. The dominant paradigm in segmentation is using convolutional neural networks, less common are recurrent neural networks. In this work, we propose a new deep learning method for cell segmentation, which integrates convolutional neural networks and gated recurrent neural networks over multiple image scales to exploit the strength of both types of networks. To increase the robustness of the training and improve segmentation, we introduce a novel focal loss function. We also present a distributed scheme for optimized training of the integrated neural network. We applied our proposed method to challenging data of glioblastoma cell nuclei and performed a quantitative comparison with state-of-the-art methods. Insights on how our extensions affect training and inference are also provided. Moreover, we benchmarked our method using a wide spectrum of all 22 real microscopy datasets of the Cell Tracking Challenge.

Field campaign on sediment transport behaviour in a pressure main from pumping station to wastewater treatment plant in Berlin

As part of the project KURAS, the Berliner Wasserbetriebe realized a field campaign in 2015 in order to increase the process knowledge regarding the behaviour of transported sediment in the pressure main leading from the pumpstation to the wastewater treatment plant. The field campaign was conducted because of a lack of knowledge about the general condition of the pressure main due to its bad accessibility and the suspicion of deposits caused by hydraulic underload. The practical evidence of the sediment transport performance of this part of the sewer system, dependent on different load cases, should present a basis for further analysis, for example regarding flushing measures. A positive side-effect of the investigation was the description of the amount of pollutants caused by different weather conditions in combined sewer systems and the alterations of the sewage composition due to biogenic processes during transport. The concept included the parallel sampling of the inflow at the pumpstation and the outflow at the end of the pressure main during different weather conditions. By calculating the inflow to the pressure main, as well as its outflow at different flow conditions, it was possible to draw conclusions in regard to the transport behaviour of sediment and the bioprocesses within an 8.5 km section of the pressure main. The results show clearly that the effects of sedimentation and remobilization depend on the flow conditions. The balance of the total suspended solids (TSS) load during daily variations in dry weather shows that the remobilization effect during the run-off peak is not able to compensate for the period of sedimentation happening during the low flow at night. Based on the data for dry weather, an average of 238 kg of TSS deposits in the pressure main remains per day. The remobilization of sediment occurs only due to the abruptly increased delivery rates caused by precipitation events. These high pollution loads lead to a sudden strain at the wastewater treatment plant. It was found that the sediment transport behaviour is characterized by sedimentation up to a flow velocity of 0.35 m/s, while remobilization effects occur above 0.5 m/s. The assumption of bad sediment transport performance in the pressure main was confirmed. Therefore, the results can be used as a basis for further analysis, for example regarding periodical flushing as a means of cleaning the pressure main. The findings, especially regarding the methods and processes, are transferable and can be applied to other pressure mains in combined sewer systems. Besides the outlined evaluation of the sediment transport behaviour of the pressure main, the collected data were used in the project to calibrate a sewer system model, including a water quality model for the catchment area, and as a contribution towards an early physically based sediment transport modelling in InfoWorks CS.

Functional nuclear organization of transcription and DNA replication: a topographical marriage between chromatin domains and the interchromatin compartment

We studied the nuclear topography of RNA transcription and DNA replication in mammalian cell types with super-resolution fluorescence microscopy, which offers a resolution beyond the classical Abbe/Raleigh limit. Three-dimensional structured illumination microscopy (3D-SIM) demonstrated a network of channels and wider lacunas, called the interchromatin compartment (IC). The IC starts at nuclear pores and expands throughout the nuclear space. It is demarcated from the compact interior of higher-order chromatin domains (CDs) by a 100-200-nm thick layer of decondensed chromatin, termed the perichromatin region (PR). Nascent DNA, nascent RNA, RNA polymerase II (RNA Pol II), as well as histone modifications for transcriptionally competent/active chromatin, are highly enriched in the PR, whereas splicing speckles are observed in the interior of the IC. In line with previous electron microscopic evidence, spectral precision distance/position determination microscopy (SPDM) confirmed the presence of RNA Pol II clusters indicative of transcription factories. Still, a substantial part of transcription apparently takes place outside of such factories. Previous electron microscopic evidence has suggested that the functional nuclear organization of DNA replication depends on brownian movements of chromatin between the CD interior and the PR. As an incentive for future studies, we hypothesize that such movements also take place during transcription, i.e., only the actually transcribed part of a gene may be located within the PR, whereas its major part, including previously or later transcribed sequences, is embedded in a higher-order chromatin configuration in the interior of the CD.

Drug monitoring of low-dose PEG-asparaginase (Oncaspar) in children with relapsed acute lymphoblastic leukaemia

Use of asparaginase (ASNase) in the treatment of relapsed childhood acute lymphoblastic leukaemia (ALL) is associated with a high rate of hypersensitive reactions. 'Silent' inactivation may additionally reduce treatment intensity. Therefore, PEG-ASNase (Oncaspar), a polyethylene glycol conjugate of the native Escherichia coli-ASNase, was introduced into the Berlin-Frankfurt-Münster (BFM) 96 treatment protocol for relapsed ALL under drug monitoring conditions. A single i.v. dose of 500 IU/m2 PEG-ASNase, substituted for the native ASNases, was administered to supply a plasma activity of 100 IU/l for 1 week. From November 1997 to March 2000, 35 patients from 23 BFM-associated hospitals, with or without a previous allergic reaction to one or both native preparations, underwent monitoring. After 82 applications, a total of 270 samples were submitted to be tested for ASNase activity. The ASNase activity on the day of the administration and the following day ranged between < 20 and 693 IU/l, with a median of 413 IU/l (53 samples). The median on d 7 +/- 1 was 199 IU/l (range <20--421 IU/l; 41 samples) and on d 14 +/- 1, 105 IU/l (range <20--188 IU/l; 19 samples). An ASNase activity of > 100 IU/l was seen on d 7 in 36 activity time courses of 52 interpretable applications (69%). Intraindividual variability of activity time courses was low. However, a rapid decrease in ASNase activity after repeated applications was observed in 4 out of 20 children. Previously experienced allergic reactions to native ASNases did not influence PEG-ASNase pharmacokinetics. PEG-ASNase is a useful alternative to the native ASNases in children with relapsed ALL. Whenever possible, drug monitoring should be performed to identify patients with 'silent' inactivation.

Pegylated asparaginase (Oncaspar) in children with ALL: drug monitoring in reinduction according to the ALL/NHL-BFM 95 protocols

Hypersensitivity reactions are relevant adverse effects of asparaginase therapy. Therefore, children treated with native Escherichia coli asparaginase in induction therapy of acute lymphoblastic leukaemia (ALL) or non-Hodgkin's lymphoma (NHL) were switched to the pegylated enzyme for reinduction under drug monitoring. Seventy children, including four patients with allergic reactions during induction, were given one dose of Oncaspar 1,000 U/m2 intravenously. Activity was determined every third or fourth day until it dropped below the limit of quantification. In current reinduction protocols [ALL/NHL-Berlin-Frankfurt-Münster (BFM) 95 trials], four doses of 10,000 U/m2 E. coli asparaginase deplete asparagine for about 2-3 weeks, therefore activities of >/= 100 U/l up to day 14 and >/= 50 U/l up to day 21 were targeted. In 66 patients without an allergic reaction during induction, the mean activity was 606 +/- 313 U/l, 232 +/- 211 U/l and 44 +/- 50 U/l after 1, 2 and 3 weeks respectively. In 44/66 patients, activity was >/= 100 U/l after 14 d. A rapid decline in activity was seen in the remaining 22 patients, including 8/22 patients who showed no activity after 1 week. Toxicity was low and comparable to the native enzymes but, in contrast to about 30% of hypersensitivity reactions with conventional reinduction therapy, no allergic reaction was seen. Substituting 4 x 10,000 U/m2 asparaginase medac for one dose of 1,000 U/m2 Oncaspar was safe and well tolerated. Comparable pharmacokinetic treatment intensity was achieved in about two-thirds of patients.