Events

Seminar Series

Presenter:

Huan Truong

Date:

10-07-2013

Time:

11:00AM-12:00PM

Location:

2206A Student Center

Large-Scale Pairwise Alignments on GPU Clusters: Exploring the Implementation Space

Several problems in computational biology require the all-against-all pairwise comparisons of tens of thousands of individual biological sequences. Each such comparison can be performed with the well-known Needleman-Wunsch alignment algorithm. However, with the rapid growth of biological databases, performing all possible comparisons with this algorithm in serial becomes extremely time-consuming. The massive computational power of graphics processing units (GPUs) makes them an appealing choice for accelerating these computations. As such, CPU-GPU clusters can enable all-against-all comparisons on large datasets.

In this work, we present four GPU implementations for large-scale pairwise sequence alignment: TiledDScan-mNW, DScan-mNW, RScan-mNW and LazyRScan-mNW. The proposed GPU kernels exhibit different parallelization patterns: we discuss how each parallelization strategy affects the memory accesses and the utilization of the underlying GPU hardware. We evaluate our implementations on a variety of low- and high-end GPUs with different compute capabilities. Our results show that all the proposed solutions outperform the existing open-source implementation from the Rodinia Benchmark Suite, and LazyRScan-mNW is the preferred solution for applications that require performing the trace-back operation only on a subset of the considered sequence pairs (for example, the pairs whose alignment score exceeds a predefined threshold). Finally, we discuss the integration of the proposed GPU kernels into a hybrid MPI-CUDA framework for deployment on CPU-GPU clusters. In particular, our proposed distributed design targets both homogeneous and heterogeneous clusters with nodes that differ amongst themselves in their hardware configuration.

Seminar Series

Presenter:

Mirna Becevic

Date:

09-30-2013

Time:

11:00AM-12:00PM

Location:

2206A Student Center

TeleMDID: Mobile Technology Applications for Interactive Diagnoses in Teledermatology

A web-based dermatology image management application, Missouri Dermatology Image Database (MDID), has been developed to facilitate dermatology practices. The digital images captured offsite are transferred to MDID’s secure server via encrypted connection and user authentication. Uploaded images can be organized by multiple criteria, and patients and images can be easily searched. Originally, the MDID database only applied to in-person patients.  Prior to designing the mobile application, we conducted informal observations of telehealth workflow. A typical Missouri Telehealth Network (MTN) dermatology session lasts 15-25 minutes, and is similar to a clinic visit. Images from offsite are transmitted by attaching a digital camera on a secondary input through hybrid store-and-forward. Photos are normally taken before the telehealth session to save interface time.

A number of limitations exist with using digital cameras on secondary inputs. One, dermatologists have minimal control over the images. There is overhead associated with switching to secondary input, and clinicians must ask offsite presenters to step through images manually. This interrupts communication with patients, since video feed is no longer received on the monitor. Images must be presented twice if residents are involved. The offsite staff must delete photos from the camera after each session to maintain patient confidentiality. Once images are deleted, they are not available for history, diagnostic, or training purposes, or meta-data analyses. We developed TeleMDID, and installed it on iPads in 10 offsite clinics and 2 MU clinics. Data were collected in May-June 2013. Two pre-launch and two post-launch internet-based surveys were distributed to offsite presenters and dermatology clinic providers. Post-launch surveys showed 100% of providers who used TeleMDID preferred using TeleMDID to digital cameras, while 75% of offsite presenters preferred using TeleMDID.Our aim was to adapt in-person clinical technologies to teledermatology clinics to compare usability and effectiveness with current teledermatology modalities. Our pilot showed experiences differed between provider and offsite presenters. Providers reported greater satisfaction with TeleMDID than offsite presenters. Providers viewed it as an improvement to clinic flow. All providers had previously used mobile technologies daily, compared to 30% offsite presenters, possibly influencing satisfaction levels.

We observed TeleMDID appointments were more efficient, since photos were ready prior to connecting with providers. This resulted in shorter patient visit time. Providers could spend more time on diagnosis, without waiting for offsite presenters to show photos and switch inputs. While costs were not examined, we observed that TeleMDID increased provider and presenter satisfactions by allowing more streamlined, better organized appointments, and were closer to in-person appointments.

Future work has both implementation and research goals. New functionalities are desirable, as is full-scale deployment to 150 MTN sites. Future improvements include interface to provide meta-data about the images and to annotate regions of an image, including ability to view highlighted regions offsite in real-time. Patient history and diagnosis need to be available to clinicians. For offsites, improved training/technical support is needed. Potential cost savings of this solution needs verification. The TeleMDID app did show clear improvement in work flow.

Seminar Series

Presenter:

Ginger Han

Date:

03-18-2013

Time:

11:00AM-12:00PM

Location:

2206A Student Center

1 19 20 21