Deep Learning and Hand-Crafted Feature Based Approaches for Polyp Detection in Medical Videos

Pogorelov, Konstantin; Ostroukhova, Olga; Jeppsson, Mattis; Espeland, Håvard; Griwodz, Carsten; Lange, Thomas de; Johansen, Dag; Riegler, Michael; Halvorsen, Pål

doi:10.1109/cbms.2018.00073

Cited by 48 publications

(38 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…For the FDSI sub-task, we have also opted for a CNN and a transfer learning-based classification approach, previously validated in a different application domain in our work [41]. In fact, we initially tried to apply the well-performing GAN approach introduced in our previous works for the flood detection satellite imagery [2] and medical imagery [42,43].…”

Section: Methodsology For Fdsi Taskmentioning

confidence: 99%

Automatic detection of passable roads after floods in remote sensed and social media data

Ahmad

Pogorelov

Riegler

et al. 2019

Signal Processing: Image Communication

Self Cite

View full text Add to dashboard Cite

This paper addresses the problem of floods classification and floods aftermath detection utilizing both social media and satellite imagery. Automatic detection of disasters such as floods is still a very challenging task. The focus lies on identifying passable routes or roads during floods. Two novel solutions are presented, which were developed for two corresponding tasks at the MediaEval 2018 benchmarking challenge. The tasks are (i) identification of images providing evidence for road passability and (ii) differentiation and detection of passable and non-passable roads in images from two complementary sources of information. For the first challenge, we mainly rely on object and scene-level features extracted through multiple deep models pre-trained on the ImageNet and Places datasets. The object and scene-level features are then combined using early, late and double fusion techniques. To identify whether or not it is possible for a vehicle to pass a road in satellite images, we rely on Convolutional Neural Networks and a transfer learning-based classification approach. The evaluation of the proposed methods are carried out on the largescale datasets provided for the benchmark competition. The results demonstrate significant improvement in the performance over the recent state-of-art approaches.

show abstract

Section: Methodsology For Fdsi Taskmentioning

confidence: 99%

Automatic detection of passable roads after floods in remote sensed and social media data

Ahmad

Pogorelov

Riegler

et al. 2019

Signal Processing: Image Communication

Self Cite

View full text Add to dashboard Cite

show abstract

“…Device Lending is a mechanism for decoupling devices from the hosts they physically reside in. Using hardware memory mappings, we facilitate the use of remote hardware resources without adding any software overhead [1,2,45]. We have extended our original Device Lending with KVM support for peer-to-peer transfers between multiple devices passed through to a VM.…”

Section: Discussionmentioning

confidence: 99%

“…For both steps, we use the stochastic gradient descent optimizer available in Keras. We ran the training on an 8-classes image dataset of the gastrointestinal tract called Kvasir [43][44][45].…”

Section: Image Classification Workloadmentioning

confidence: 99%

Flexible device compositions and dynamic resource sharing in PCIe interconnected clusters using Device Lending

et al. 2019

Self Cite

View full text Add to dashboard Cite

Modern workloads often exceed the processing and I/O capabilities provided by resource virtualization, requiring direct access to the physical hardware in order to reduce latency and computing overhead. For computers interconnected in a cluser, access to remote hardware resources often requires facilitation both in hardware and specialized drivers with virtualization support. This limits the availability of resources to specific devices and drivers that are supported by the virtualization technology being used, as well as what the interconnection technology supports. For PCI Express (PCIe) clusters, we have previously proposed Device Lending as a solution for enabling direct low latency access to remote devices. The method has extremely low computing overhead, and does not require any application- or device-specific distribution mechanisms. Any PCIe device, such as network cards disks, and GPUs, can easily be shared among the connected hosts. In this work, we have extended our solution with support for a virtual machine (VM) hypervisor. Physical remote devices can be “passed through” to VM guests, enabling direct access to physical resources while still retaining the flexibility of virtualization. Additionally, we have also implemented multi-device support, enabling shortest-path peer-to-peer transfers between remote devices residing in different hosts.Our experimental results prove that multiple remote devices can be used, achieving bandwidth and latency close to native PCIe, and without requiring any additional support in device drivers. I/O intensive workloads run seamlessly using both local and remote resources. With our added VM and multi-device support, Device Lending offers highly customizable configurations of remote devices that can be dynamically reassigned and shared to optimize resource utilization, thus enabling a flexible composable I/O infrastructure for VMs as well as bare-metal machines.

show abstract

“…In addition to the above contributions, the author also supervised several master students, organized workshops and was part of program committees for conferences. One of the latest papers describing author's GAN-based detection and localization approach (that was developed for the DeepEIR system) called "Deep Learning and Hand-crafted Feature Based Approaches for Polyp Detection in Medical Videos" won a Best Paper Award at the 2018 IEEE 31st International Symposium on Computer-Based Medical Systems [92] (Paper XV).…”

Section: This Thesis Author's Independent Contributionsmentioning

confidence: 99%

“…In the validation experiments performed using different datasets, we evaluated the activation thresholds from one pixel to a quarter of the frame. The best detection results were achieved with a threshold value of 50 pixels [92], which has been used for the detection experiments.…”

Section: Gan-based Segmentation Localization and Detectionmentioning

confidence: 99%

Lire

Lux

Riegler

Halvorsen

et al. 2016

Proceedings of the 7th International Conference on Multimedia Systems

Self Cite

View full text Add to dashboard Cite

Advanced and automated medical systems have been in the research focus for a long time. Together with the rapid development of sensing devices, the modern information analysis methods allow the new wave of computer-assisted systems to improve health care, quality of life, and patient survival rate. Together with the traditional computer vision and medical imaging, core competencies of the multimedia community such as integration and analysis of data from several sources, real-time processing and the assessment of usefulness for end-users play an essential role for the successful improvement of health care systems addressing challenges and open problems in the field of medicine. Our work explores different fields in multimedia research, starting from collection and annotation of multimedia data through automatic analysis of content and efficient processing of workloads to visualization and results representation. We have researched and developed a holistic medical multimedia system addressing a use case with an important medical and societal impact. We target lesions and findings detection and localization in the gastrointestinal (GI) tract of the human body in order to be able to support medical experts in their daily routine work. The early and precise detection of abnormalities in the GI tract greatly increases the chance of successful treatment if the initial observation of disease indicators occurs before the patient notices any symptoms, it is a non-trivial task that can be, however, efficiently automated. We investigated the GI tract visual analysis from a multimedia research point of view via several steps of research and development. First, we looked into the problem of medical data acquisition. We collected, annotated, and published several datasets and data annotation tools as open source. Then, we designed and developed a set of lesion and findings detection and localization approaches based on hand-crafted methods as well as on global-, local-and deepfeature-based methods, which serves as the algorithmic basis of our system. Next, we created a holistic medical multimedia system called DeepEIR. We researched and developed different subsystems for our DeepEIR system, namely (i) the data exploration and annotation subsystem, which makes it possible to collect and annotate data and transfer knowledge from medical experts into our system; (ii) the detection and localization subsystem, which perform medical data analysis in order to detect and localize lesions and findings; and (iii) the visualization and results representation subsystem that provides the information to medical personnel. Furthermore, the focus of the DeepEIR system lies on the accurate and time-efficient processing of multimedia data. We investigated, therefore, parallel and distributed processing, GPU-based acceleration and different classification and segmentation approaches that are evaluated and compared with state-of-the-art methods, algorithms, and systems. We demonstrated that the DeepEIR system could outperform state-of-the-art app...

show abstract

Deep Learning and Hand-Crafted Feature Based Approaches for Polyp Detection in Medical Videos

Cited by 48 publications

References 25 publications

Automatic detection of passable roads after floods in remote sensed and social media data

Automatic detection of passable roads after floods in remote sensed and social media data

Flexible device compositions and dynamic resource sharing in PCIe interconnected clusters using Device Lending

Lire

Contact Info

Product

Resources

About