Behind the Paper
The real stories behind the latest research papers, from conception to publication, the highs and the lows
Rummagene brings to the surface a massive collection of published biomedical datasets that are currently buried and inaccessible
Rummagene is a biomedical research search engine providing access to hundreds of thousands of mammalian gene sets mined from the supporting materials of research publications listed on PubMed Central. Rummagene is freely and openly available at: https://rummagene.com.
Using CRFasRNNs in Brain Extraction
We study a unique way of using CRFs in brain extraction. To our knowledge, this is the first time successfully using CRFs within the Deep Learning network for 3D medical image segmentation.
Right Atrium Segmentation (RAS), how cardiac imaging data resources contribute to personalized diagnosis and treatment of atrial fibrillation patients through digital heart twins?
RAS, a data set for atrial structural analysis and mining, focuses on the structural remodeling of the atrium of patients with atrial fibrillation or personalized diagnosis and treatment based on digital twin hearts.
Speed of environmental change frames relative ecological risk in climate change and climate intervention scenarios
We show that considering the speed of temperature change helps place different scenarios of climate change and climate intervention in context to relative ecological risk.
Studying animal seed dispersal and forest regeneration: where cutting-edge modelling techniques meet intensive field-based research
Behind every large-scale ecological model is the indispensable fieldwork of numerous experts on the ground. Here, we dive into the collective research effort behind our latest paper – highlighting four datasets on frugivory, animal behaviour, gut passage time of seeds, and forest restoration.
Fast and effective molecular property prediction with transferability map
Transfer learning improves molecular property prediction in limited datasets, yet suffers from negative transfer due to insufficient relatedness. We develop a principal gradient-based measurement to evaluate transferability before applying transfer learning, significantly improving the performance.
