美國東北大學: CUDA加速疾病光譜判讀 -- 戲劇的縮短7小時的運算, 到眨眼快的0.65秒就可以完成!!

美國東北大學: CUDA加速疾病光譜判讀 -- 戲劇性的縮短7小時的運算, 到眨眼快的0.65秒就可以完成!!

Northeastern University: Reducing Disease Diagnostics Time

美國東北大學的光譜分析實驗室的領導教授, Max Diem, 致力於疾病光譜判讀研究與開發. 這個實驗室量測物體, 並以數理統計方法, 判讀/標定疾病對人類細胞/組織/體液所造成的改變.

他們使用我們3月底提過的, AccelerEyes公司所銷售, 神奇的”外套”程式 -- 不需人類幫忙, 電腦可以即時智能改寫MATLAB程式碼, 成為具有CUDA加速的MATLAB程式碼.

以這個智能高速CUDA加速工具, 該實驗室大幅加速疾病高光譜影像分析流程, 戲劇性的縮短7小時的運算到0.65秒就可以完成!!

The Laboratory for Spectral Diagnosis at Northeastern University, under the leadership of Prof. Max Diem, is focused on the spectral diagnosis of disease. The lab uses objective measurements and mathematical-statistical methods to diagnose diseases affecting human cells, tissues or body fluids. Using Jacket from AccelerEyes to accelerate the hyperspectral image analysis workflow, the lab has dramatically reduced the time for diagnoses with CUDA.
- Background: In hyperspectral imaging, sensors collect a set of images, each of which
represents a range of the electromagnetic spectrum. The images are then combined to form
a 3D hyperspectral cube for processing and analysis.
- Read more and watch the videos here: http://www.accelereyes.com/resources/spectroscopy.
- Note: In the second video, hear how a 7-hour non-parallelized project now takes only
65 milliseconds with the GPU-based system.

請參閱本影片http://www.accelereyes.com/resources/spectroscopy.它提到縮短7小時的運算到0.65秒.