Martin D. Meyer

Impaired Peripheral Vasomotion in Diabetes

OBJECTIVE To test the hypothesis that vasomotion, the rhythmic contraction exhibited by small arteries and arterioles, is impaired in diabetic subjects compared with healthy control subjects. RESEARCH DESIGN AND METHODS We mathematically modeled the oscillations in laser Doppler microvascular measurements taken from the pulpar surface of the index finger in 20 healthy control subjects and 20 age-matched diabetic subjects (8 with type I and 12 with type II diabetes). The mean duration of diabetes was 17.1 ± 2.3 years, and mean HbA1c was 9.1 ± 0.4%. Blood flow was measured for 5 min as subjects rested quietly in a closed room. Fast Fourier transformation was performed to provide the frequency power spectrum of each recording. Amplitude of vasomotion was correlated with six quantitative measurements of neuropathy. RESULTS Diabetic subjects had impaired low-frequency oscillation vasomotion in 75% of age-matched patients (15 of 20 patients), with mean amplitudes of 24.9 ± 6.4 vs. 129.0 ± 33.2 (P < 0.0039). Of six somatic and autonomic neuropathy variables, only the warm thermal sensory threshold correlated significantly with the mean amplitude of vasomotion (r = −0.75, P < 0.0009). CONCLUSIONS Patterns of peripheral vasomotion are clearly disordered in diabetes. The loss of low-frequency oscillations observed here suggests a peripheral vascular abnormality that extends past the capillary network to arterial vessels. It is uncertain whether the accompanying small unmyelinated nerve C-fiber dysfunction is a cause or consequence of the impaired microvascular function. Measurement of vasomotion may prove useful as a novel test for peripheral neurovascular function.

A high sampling rate delayed LMS filter architecture

The problem of implementing a high sampling rate transversal form adaptive filter is investigated. A highly pipelined systolic-type alternative to the conventional LMS adaptive filter is presented. The proposed system consists of a linear array of identical processing modules specifically suited to the computational requirements of the delayed LMS algorithm. The resulting adaptive filter structure can accommodate very high sampling rates, which are independent of the filter order. The performance of the system is analyzed in terms of computational speedup and maximum sampling rate, and the effect of adaptation delay on algorithm convergence is addressed. >

Normal Blood Flow Response and Vasomotion in the Diabetic Charcot Foot

Vasomotion, the spontaneous rhythmic contraction exhibited by small arteries and arterioles is dysregulated in patients with diabetic neuropathy. We examined the relationship between Charcot arthropathy and vasomotion at the dorsum of the foot. We studied nine diabetic patients with clinically diagnosed neuropathy and Charcot arthropathy in 13 feet (n=13), twelve subjects with diabetic neuropathy and no Charcot deformity (n=12), and 11 healthy controls (n=11). Following neuropathy assessment, blood flow was measured by laser Doppler flowmetry with local skin warming. Fast Fourier transformation was performed to provide an index of vasomotion. Subjects with Charcot osteoarthropathy had more severe somatic neuropathy and higher circulating levels of serum calcium (9.8+/-0.1 versus 9.3+/-0.1 mg/dL). Raising local temperature increased skin blood flow and vasomotion in both control subjects and Charcot subjects, but not in diabetic patients with neuropathy alone (p < 0.05 for blood flow, p < 0.02 for vasomotion). Patterns of peripheral vasomotion and blood flow which are clearly disordered in diabetic neuropathy are intact in patients with a Charcot osteoarthropathy, despite a more severe sensory nerve impairment. These findings suggest that the loss of peripheral blood flow and vasomotion often seen in diabetic neuropathy may actually be protective against Charcot arthropathy by preventing bone resorption. It remains unclear then whether the Charcot arthropathy is a direct result of a failure to decrease blood flow to bone, or is the manifestation of some other pathology.

A modular pipelined implementation of a delayed LMS transversal adaptive filter

The problem of efficiently realizing a delayed least-mean-squares (DLMS) transversal adaptive filter is investigated. A time-shifted version of the DLMS algorithm is derived. Due to its order recursive nature, the restructured algorithm is well suited to parallel implementation. In addition, a pipelined systolic-type architecture which implements the algorithm is presented. The performance of the pipelined system is analyzed, and equations for speedup are derived. The pipelined system is capable of much greater throughput than existing conventional least-mean-square (LMS) implementations, making it a good candidate for real-time applications where high sampling rates are required. Also, due to its high modular nature, the system is easily expandable.<<ETX>>

Vectorization of the DLMS transversal adaptive filter

The subject of high sampling rate realizations for transversal adaptive filters is addressed. In particular, a vectorized version of the delayed least mean squares (DLMS) algorithm is derived using look-ahead computation techniques. The resulting parallel algorithm is then mapped onto a linear array of highly pipelined processing modules, which can accept an input vector of arbitrary length, and compute the corresponding output vector in a single clock cycle. The proposed system is shown to be capable of implementing transversal adaptive filters at sampling rates which are theoretically without bound. The performance of the proposed system is analyzed and simulation results are presented to verify the convergence properties of the algorithm under varying degrees of vectorization. >

An arbitrarily high sampling rate DLMS adaptive filter

The authors propose a new scheme for implementing gradient based transversal adaptive filters at speeds which are theoretically independent of both filter order and hardware speed. The design is based on an order recursive variant of the delayed least mean squares (DLMS) algorithm. The order recursive nature of this algorithm allows the use of pipelining techniques to design high speed adaptive filters whose sampling rates are independent of the filter order. It is shown that this algorithm can be vectorized so that an arbitrary number of outputs can be computed simultaneously using parallel hardware. The proposed system is shown to be capable of implementing transversal adaptive filters at sampling rates which are theoretically without bound. The performance of the proposed system is analyzed and simulation results are presented to verify the convergence properties of the algorithm.<<ETX>>

Adaptive lattice filter implementations on pipelined multiprocessor architectures

Three pipelined multiprocessor implementations of adaptive lattice filters are examined. The three multiprocessor architectures, which can be characterized as a serial pipeline, a ladder-connected dual pipeline, and a ring pipeline, are derived directly from the computational and data-transfer requirements of adaptive lattice algorithms. The order-recursive nature of the adaptive lattice structure results in architectures which use pipelining extensively. A performance analysis is done for each multiprocessor system, with respect to two different adaptive lattice algorithms. Expressions for approximate computation time and speedup are derived for each combination of architecture and algorithm.<<ETX>>

A fault tolerant pipelined adaptive filter

A fault tolerant pipelined architecture for high sampling rate adaptive filters is presented. The architecture, which is based on the computational requirements of delayed LMS and adaptive lattice filters, offers robust performance in the presence of single hardware faults, and software faults resulting from numerical instability. The reliability of the proposed system is analyzed and compared to the existing implementations strategies, and methods for fault detection, fault location, and recovery via hardware reconfiguration are discussed. Finally, simulation results illustrating recovery from processor fault are presented.<<ETX>>