Efficacy improvement of distribution of F-Latencies (DFL) in the diagnosis of peripheral neuropathy

Abstract

The nervous system, consisting of central and peripheral nervous systems, is very important for fast control of our body functions. Therefore, measurement of nerve conduction velocity (CV) is an important tool in the diagnosis of peripheral neuropathy. Distribution of F-latency (DFL) is a nerve conduction parameter introduced by our extended group at Dhaka University and this group has also developed a method for the detection of cervical spondylotic radiculopathy and myelopathy (CRM), a neurological problem causing neck pain, identifying patterns of DFL. The present work was taken up to improve this method of detection, or in other words, to improve the efficacy of DFL in the detection of CRM. For normal healthy persons DFL has a single sharp peak while for subjects with CRM, DFL may have a broad peak, double or triple peaks as determined for the median nerve for which a large experience has been built upon by our extended group over the years. However, the distinction between the single and the broad peak is very subtle and it is necessary to obtain DFL as accurately as possible for an effective detection of CRM. DFL is a statistical parameter, and we cannot obtain a large number of F-latencies as the process is slightly painful for the patients, about 40 F-latencies being the practical maximum. The present work is related to the accurate determination of the pattern of DFL under such practical limitations. For most of the studies in the present thesis, DFL data from human subjects obtained by a previous worker in our extended group was used where all the subjects had a standard diagnosis for CRM using MRI, which is the existing ‘gold standard’. Of course some data were acquired during the present work as well. For the accurate determination of DFL effort was made mainly in four directions which are described in brief in the following four paragraphs. Firstly we tried to obtain a very closely related but more fundamental distribution called Distribution of Conduction Velocity (DCV) from the raw F-latencies to find out if it is useful in the detection of CRM. Our extended group had established earlier that DCV is an approximate mirror image of DFL. The argument behind this relationship also allows us to obtain DCV directly from the CVs corresponding to the multiple F-latencies. However, since a large experience has been built up through earlier work of our extended group in relating the shapes of DFL, obtained from the median nerve, to CRM, we wanted to obtain DCVs that show similar but laterally inverted (mirrored) shapes. Our extended group used 2ms bin width to obtain DFL, so we tried to determine the most appropriate bin width for computing DCV from CVs so that the above mentioned features are retained. We related the typically used 2ms bin width of DFL for each nerve to the average of the maximum and minimum values, median and mode values of the corresponding CV data. It was found that by relating the median value of CV to the 2ms fixed bin width used for DFL, the best matching of DCV patterns was obtained. An important part of this study was the sample size of DFL, i.e., to determine how many F-latencies are adequate to make a representative DFL for a particular bin size, particularly to make it effective in the detection of CRM. For this second study, two independent sets of data were used to make the judgment in between normal and abnormal DFL patterns. One used data from 25 median nerves of 16 persons collected earlier by our extended group. These had more than 30 but less than 40 F-latencies each. Further three sets of data were acquired during the present work where a large number (about 200 or more) of non-zero F-latency values were obtained from three median nerves of two subjects. Obtaining random samples with sample sizes of 5, 10, 20 and 30 from both the old and the new data and drawing corresponding DFLs with different bin widths a decision was reached in terms of the sample size, in order to detect CRM reliably. It was found that if the bin width is chosen at 1.5ms or 2.0ms, a minimum sample of 20 F-latencies can give a representative pattern of DFL for the detection of CRM. The third study revealed a significant effect of the starting point on the DFL pattern where the number of samples is small (20 to 40), and came up with a new method to obtain a better detection of CRM. It was observed that if the starting point of DFL is shifted by half the bin width (by 1ms where the bin width is 2ms, used mostly in the DFL work by our extended group) then the DFL peak pattern may in some cases change from single to broad, broad to double or vice-versa. This observation created a great uncertainty in the prediction of CRM, particularly for the changes between single and broad peak patterns as this changes the prediction altogether. We obtained two sets of DFLs corresponding to each available data set from human subjects with the starting point shifted by 1ms (for a bin size of 2ms). Then we combined the results through a logical OR and a logical AND operations, where a ‘Yes’ corresponded to the presence of CRM. The predictions were adjudged against MRI findings. It was found from this study that a logical OR operation of the above combination gave the best prediction for CRM, and this has now been introduced in the standard detection algorithm of our extended group. Finally we tried to find alternative means of characterizing DFL patterns quantitatively, which could lead to objective determination of CRM. We tried two common statistical parameters – Skewness and Full Width Half Maximum (FWHM) – for this purpose. We calculated these values for identified single and broad peaked patterns of DFLs (the two groups under test) obtained from many subjects and performed statistical t-tests to determine significant differences, if any. It was found that skewness could not distinguish the two groups at all while FWHM could do it very well at a very high level of significance. From an observation of the FWHM values from both the groups, a preliminary threshold value was chosen at 2.5ms; a larger value would indicate a broad peak, giving an indication that this is a case of CRM. Of course a study involving a large number of data needs to be carried out to determine a more accurate value of the threshold. The present study strengthens our confidence in characterizing the DFL patterns both qualitatively and quantitatively for the diagnosis of CRM. DFL can be measured using standard EMG equipment which can be made at a much lower cost in comparison with MRI machines, investigation using which is considered as the ‘gold standard of the day’ in this diagnosis. Portable EMG units based on Laptop computers are also available. Therefore, such portable equipment for measuring DFL can be distributed widely. The present work builds up sufficient knowledge and confidence in the use of DFL for specific determination of Cervical spondylotic radiculopathy and myelopathy which will go a long way in the diagnosis of neural disorders, contributing to public health, and the society in general.