arXiv subjects

Lars Ailo Bongo

Publications and source records attributed to Lars Ailo Bongo.

23 records · Page 2Linked to original sources

arXiv2018

Replication study: Development and validation of deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs

Replication studies are essential for validation of new methods, and are crucial to maintain the high standards of scientific publications, and to use the results in practice. We have attempted to replicate the main method in 'Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs' published in JAMA 2016; 316(22). We re-implemented the method since the source code is not available, and we used publicly available data sets. The original study used non-public fundus images from EyePACS and three hospitals in India for training. We used a different EyePACS data set from Kaggle. The original study used the benchmark data set Messidor-2 to evaluate the algorithm's performance. We used the same data set. In the original study, ophthalmologists re-graded all images for diabetic retinopathy, macular edema, and image gradability. There was one diabetic retinopathy grade per image for our data sets, and we assessed image gradability ourselves. Hyper-parameter settings were not described in the original study. But some of these were later published. We were not able to replicate the original study. Our algorithm's area under the receiver operating curve (AUC) of 0.94 on the Kaggle EyePACS test set and 0.80 on Messidor-2 did not come close to the reported AUC of 0.99 in the original study. This may be caused by the use of a single grade per image, different data, or different not described hyper-parameter settings. This study shows the challenges of replicating deep learning, and the need for more replication studies to validate deep learning methods, especially for medical image analysis. Our source code and instructions are available at: https://github.com/mikevoets/jama16-retina-replication

cs.CV↗

arXiv2017

Feature Extraction for Machine Learning Based Crackle Detection in Lung Sounds from a Health Survey

In recent years, many innovative solutions for recording and viewing sounds from a stethoscope have become available. However, to fully utilize such devices, there is a need for an automated approach for detecting abnormal lung sounds, which is better than the existing methods that typically have been developed and evaluated using a small and non-diverse dataset. We propose a machine learning based approach for detecting crackles in lung sounds recorded using a stethoscope in a large health survey. Our method is trained and evaluated using 209 files with crackles classified by expert listeners. Our analysis pipeline is based on features extracted from small windows in audio files. We evaluated several feature extraction methods and classifiers. We evaluated the pipeline using a training set of 175 crackle windows and 208 normal windows. We did 100 cycles of cross validation where we shuffled training sets between cycles. For all the division between training and evaluation was 70%-30%. We found and evaluated a 5-dimenstional vector with four features from the time domain and one from the spectrum domain. We evaluated several classifiers and found SVM with a Radial Basis Function Kernel to perform best. Our approach had a precision of 86% and recall of 84% for classifying a crackle in a window, which is more accurate than found in studies of health personnel. The low-dimensional feature vector makes the SVM very fast. The model can be trained on a regular computer in 1.44 seconds, and 319 crackles can be classified in 1.08 seconds. Our approach detects and visualizes individual crackles in recorded audio files. It is accurate, fast, and has low resource requirements. It can be used to train health personnel or as part of a smartphone application for Bluetooth stethoscopes.

cs.SD↗

arXiv2016

nsroot: Minimalist Process Isolation Tool Implemented With Linux Namespaces

Data analyses in the life sciences are moving from tools run on a personal computer to services run on large computing platforms. This creates a need to package tools and dependencies for easy installation, configuration and deployment on distributed platforms. In addition, for secure execution there is a need for process isolation on a shared platform. Existing virtual machine and container technologies are often more complex than traditional Unix utilities, like chroot, and often require root privileges in order to set up or use. This is especially challenging on HPC systems where users typically do not have root access. We therefore present nsroot, a lightweight Linux namespaces based process isolation tool. It allows restricting the runtime environment of data analysis tools that may not have been designed with security as a top priority, in order to reduce the risk and consequences of security breaches, without requiring any special privileges. The codebase of nsroot is small, and it provides a command line interface similar to chroot. It can be used on all Linux kernels that implement user namespaces. In addition, we propose combining nsroot with the AppImage format for secure execution of packaged applications. nsroot is open sourced and available at: https://github.com/uit-no/nsroot

cs.DC↗

arXiv2016

META-pipe - Pipeline Annotation, Analysis and Visualization of Marine Metagenomic Sequence Data

The marine environment is one of the most important sources for microbial biodiversity on the planet. These microbes are drivers for many biogeochemical processes, and their enormous genetic potential is still not fully explored or exploited. Marine metagenomics (DNA shotgun sequencing), not only offers opportunities for studying structure and function of microbial communities, but also identification of novel biocatalysts and bioactive compounds. However, data analysis, management, storage, processing and interpretation are significant challenges in marine metagenomics due to the high diversity in samples and the size of the marine flagship projects. We provide a new pipeline, META-pipe, for marine metagenomics analysis. It offers pre- processing, assembly, taxonomic classification and functional analysis. To reduce the effort to develop and deploy it, we have integrated existing biological analysis frameworks, and compute and storage infrastructure resources. Our current META-pipe web service provides integration with identity provider services, distributed storage, computation on a Supercomputer, Galaxy workflows, and interactive data visualizations. We have evaluated the scalability and performance of the analysis pipeline. Our results demonstrate how to develop and deploy a pipeline on distributed compute and storage resources, and discusses important challenges related to this process.

cs.DC↗

arXiv2016

Large-scale Biological Meta-database Management

Up-to-date meta-databases are vital for the analysis of biological data. However,the current exponential increase in biological data leads to exponentially increasing meta-database sizes. Large-scale meta-database management is therefore an important challenge for production platforms providing services for biological data analysis. In particular, there is often a need either to run an analysis with a particular version of a meta-database, or to rerun an analysis with an updated meta-database. We present our GeStore approach for biological meta-database management. It provides efficient storage and runtime generation of specific meta-database versions, and efficient incremental updates for biological data analysis tools. The approach is transparent to the tools, and we provide a framework that makes it easy to integrate GeStore with biological data analysis frameworks. We present the GeStore system, an evaluation of the performance characteristics of the system, and an evaluation of the benefits for a biological data analysis workflow.

cs.DC↗