A5054576402- COVID-19 diagnosis using AI
- Advanced Neural Network Applications
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Generative Adversarial Networks and Image Synthesis
- Anomaly Detection Techniques and Applications
- Cell Image Analysis Techniques
- Single-cell and spatial transcriptomics
- Prostate Cancer Diagnosis and Treatment
- Image Retrieval and Classification Techniques
- Adversarial Robustness in Machine Learning
- Prostate Cancer Treatment and Research
- Machine Learning and Data Classification
- Protein Structure and Dynamics
- Mathematical Biology Tumor Growth
- Medical Image Segmentation Techniques
- Machine Learning in Bioinformatics
- Brain Tumor Detection and Classification
- Retinal Imaging and Analysis
- Enzyme Structure and Function
- Network Security and Intrusion Detection
- Fault Detection and Control Systems
- Foot and Ankle Surgery
- X-ray Diffraction in Crystallography
- Access Control and Trust
DeepMind (United Kingdom)
2021
German Cancer Research Center
2018-2021
Heidelberg University
2019-2021
Google (United States)
2019
DKFZ-ZMBH Alliance
2018
Karlsruhe Institute of Technology
2015
EIA University
2015
Universidad de Antioquia
2015
Abstract Proteins are essential to life, and understanding their structure can facilitate a mechanistic of function. Through an enormous experimental effort 1–4 , the structures around 100,000 unique proteins have been determined 5 but this represents small fraction billions known protein sequences 6,7 . Structural coverage is bottlenecked by months years painstaking required determine single structure. Accurate computational approaches needed address gap enable large-scale structural...
Abstract Protein structures can provide invaluable information, both for reasoning about biological processes and enabling interventions such as structure-based drug development or targeted mutagenesis. After decades of effort, 17% the total residues in human protein sequences are covered by an experimentally determined structure 1 . Here we markedly expand structural coverage proteome applying state-of-the-art machine learning method, AlphaFold 2 , at a scale that covers almost entire...
In this work, we report the set-up and results of Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with IEEE International Symposium on Biomedical Imaging (ISBI) 2017 Conferences Medical Image Computing Computer-Assisted Intervention (MICCAI) 2018. The image dataset is diverse contains primary secondary tumors varied sizes appearances various lesion-to-background levels (hyper-/hypo-dense), created collaboration seven hospitals research institutions. Seventy-five...
The U-Net was presented in 2015. With its straight-forward and successful architecture it quickly evolved to a commonly used benchmark medical image segmentation. adaptation of the novel problems, however, comprises several degrees freedom regarding exact architecture, preprocessing, training inference. These choices are not independent each other substantially impact overall performance. present paper introduces nnU-Net ('no-new-Net'), which refers robust self-adapting framework on basis 2D...
We describe the operation and improvement of AlphaFold, system that was entered by team AlphaFold2 to "human" category in 14th Critical Assessment Protein Structure Prediction (CASP14). The AlphaFold CASP14 is entirely different one CASP13. It used a novel end-to-end deep neural network trained produce protein structures from amino acid sequence, multiple sequence alignments, homologous proteins. In assessors' ranking summed z scores (>2.0), scored 244.0 compared 90.8 next best group....
Background Men suspected of having clinically significant prostate cancer (sPC) increasingly undergo MRI. The potential deep learning to provide diagnostic support for human interpretation requires further evaluation. Purpose To compare the performance clinical assessment a system optimized segmentation trained with T2-weighted and diffusion MRI in task detection lesions suspicious sPC. Materials Methods In this retrospective study, sequences from consecutive men examined single 3.0-T...
In this work, we report the set-up and results of Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with IEEE International Symposium on Biomedical Imaging (ISBI) 2017 Conferences Medical Image Computing Computer-Assisted Intervention (MICCAI) 2018. The image dataset is diverse contains primary secondary tumors varied sizes appearances various lesion-to-background levels (hyper-/hypo-dense), created collaboration seven hospitals research institutions. Seventy-five...
Many real-world vision problems suffer from inherent ambiguities. In clinical applications for example, it might not be clear a CT scan alone which particular region is cancer tissue. Therefore group of graders typically produces set diverse but plausible segmentations. We consider the task learning distribution over segmentations given an input. To this end we propose generative segmentation model based on combination U-Net with conditional variational autoencoder that capable efficiently...
Purpose To compare biparametric contrast-free radiomic machine learning (RML), mean apparent diffusion coefficient (ADC), and radiologist assessment for characterization of prostate lesions detected during prospective MRI interpretation. Materials Methods This single-institution study included 316 men (mean age ± standard deviation, 64.0 years 7.8) with an indication MRI-transrectal US fusion biopsy between May 2015 September 2016 (training cohort, 183 patients; test 133 patients). Lesions...
Semantic segmentation constitutes an integral part of medical image analyses for which breakthroughs in the field deep learning were high relevance. The large number trainable parameters neural networks however renders them inherently data hungry, a characteristic that heavily challenges imaging community. Though interestingly, with de facto standard training fully convolutional (FCNs) semantic being agnostic towards `structure' predicted label maps, valuable complementary information about...
Reliable detection of out-of-distribution (OOD) inputs is increasingly understood to be a precondition for deployment machine learning systems. This paper proposes and investigates the use contrastive training boost OOD performance. Unlike leading methods detection, our approach does not require access examples labeled explicitly as OOD, which can difficult collect in practice. We show extensive experiments that significantly helps performance on number common benchmarks. By introducing...
Computational design of protein-binding proteins is a fundamental capability with broad utility in biomedical research and biotechnology. Recent methods have made strides against some target proteins, but on-demand creation high-affinity binders without multiple rounds experimental testing remains an unsolved challenge. This technical report introduces AlphaProteo, family machine learning models for protein design, details its performance on the de novo binder problem. With we achieve 3- to...
Unsupervised learning can leverage large-scale data sources without the need for annotations. In this context, deep learning-based auto encoders have shown great potential in detecting anomalies medical images. However, state-of-the-art anomaly scores are still based on reconstruction error, which lacks two essential parts: it ignores model-internal representation employed reconstruction, and formal assertions comparability between samples. We address these shortcomings by proposing...
The task of localizing and categorizing objects in medical images often remains formulated as a semantic segmentation problem. This approach, however, only indirectly solves the coarse localization by predicting pixel-level scores, requiring ad-hoc heuristics when mapping back to object-level scores. State-of-the-art object detectors on other hand, allow for individual scoring an end-to-end fashion, while ironically trading ability exploit full pixel-wise supervision signal. can be...
Medical imaging only indirectly measures the molecular identity of tissue within each voxel, which often produces ambiguous image evidence for target interest, like semantic segmentation. This diversity and variations plausible interpretations are specific to given regions may thus manifest on various scales, spanning all way from pixel level. In order learn a flexible distribution that can account multiple scales variations, we propose Hierarchical Probabilistic U-Net, segmentation network...