In this paper, we review data mining approaches for health applications. Our focus is on hardware-centric approaches. Modern computers consist of multiple processors, each equipped with multiple cores, each with a set of arithmetic/logical units. Thus, a modern computer may be composed of several thousand units capable of doing arithmetic operations like addition and multiplication. Graphic processors, in addition may offer some thousand such units. In both cases, single instruction multiple data and multiple instruction multiple data parallelism must be exploited. We review the principles of algorithms which exploit this parallelism and focus also on the memory issues when multiple processing units access main memory through caches. This is important for many applications of health, such as ECG, EEG, CT, SPECT, fMRI, DTI, ultrasound, microscopy, dermascopy, etc.
翻译:在本文中,我们审查健康应用的数据挖掘方法。我们的重点是以硬件为中心的方法。现代计算机由多个处理器组成,每个处理器配备多个核心,每个处理器配备一套计算/逻辑单位。因此,现代计算机可能由数千个能够进行算术操作的单位组成,例如添加和乘法。图形处理器可能提供大约1000个此类单位。在这两种情况下,必须利用单一指示多重数据和多重指示多重数据平行法。我们审查了利用这种平行法的算法原则,并在多个处理器通过缓存获取主存储器时也关注记忆问题。这对于许多健康应用都很重要,例如ECG、EEG、CT、SPECT、FMRI、DTI、超声波、显微镜、皮肤镜等。