A History of Innovation

MBF Bioscience has deep roots in the scientific research community that go back before its official founding in 1988 by Dr. Edmund Glaser and Jack Glaser, the current company president.  In the 1960's, Dr. Edmund Glaser and his colleague, Dr. Hendrick Van der Loos, created the first computer-assisted neuron reconstruction system - Neurolucida.  They published a paper in 1965 on the system entitled “A semi quantitative computer-microscope for the analysis of neuronal morphometry" that appeared in the IEEE Transactions on Biomedical Engineering.

From its inception, MBF Bioscience has been a pioneer in developing and improving microscope and computer-based scientific research systems.  Neurolucida has evolved into a powerful system that has been cited in over 5,000 research papers - more than any other neuron tracing software.  With the addition of imaging modules, Neurolucida can now automatically trace neurons and automatically detect dendritic spines.  We started developing Stereo Investigator for unbiased stereology in 1995.  It has been cited in over 4,000 research papers - more than any other stereology system.  

The company's technological innovation was recognized with a Tibbetts Award in 2013. The award was given to MBF Bioscience by the U.S. Small Business Administration (SBA) during a ceremony at the White House on May 16, 2013.

MBF Bioscience was founded on a philosophy of innovation and customer service, and continues to advance science and serve the research community.


A Little History: The Origins of Neurolucida

Peter Ohara, Ph.D.
University of California, San Francisco

Dr. Edmund Glaser recently sent me a copy of a paper he co-authored with Dr. Hendrick Van der Loos entitled “A semi quantitative computer-microscope for the analysis of neuronal morphometry” that appeared in the IEEE Transactions on Biomedical Engineering in January 1965. This paper describes what might be the earliest version of Neurolucida. Basically the paper describes a system for attaching X-Y-Z transducers to a microscope stage, tracing the branches of a Golgi-stained neuron and outputting the result to a plotter. This system was the first analog reconstruction system and Dr. Glaser informs me that it was in use at the Institute of Anatomy in Lausanne, Switzerland until 1994. A digital reconstruction system was described by Woolsey et al. in 1975, ( IEEE Transactions on Biomedical Engineering, 1973 Jul, 20(4):233-47)

I imagine there are many users of modern cell reconstruction equipment who were not born in 1965 and who don’t have much idea of what it was like in those days. The paper is written in an easily-readable style and for anyone interested in the genesis of many things we take for granted, it’s well worth the read. I have extracted two small sections below, which I hope you will find interesting.

For those who might complain about complex interfaces on modern systems and can’t find what they want on a pull down menu, try this sequence for cell reconstruction:

The sequence of events in operating the instrument is as follows:

  1. A reference point is chosen. Usually this is the center of the cell body
  2. Front panel zeroing potentiometers are used to null out the readings of the transducers at the reference point. This is done at the summing junction of the input amplifier.
  3. A reference point is chosen on the plotting board, by means of its centering potentiometers, in accordance with the shape of the dendrite system of the neuron to be mapped.
  4. The microscope stage is moved to the first point to be mapped and measured.
  5. The operator presses a foot switch. This produces the following sequence of operations:
  • The distance moved is measured and printed.
  • After printing is completed the new coordinates are stored in the holding circuits, erasing the old ones if any.
  • The plotting board pen moves to the position corresponding to the new X and Y coordinates.

The end of this cycle is indicated by a signal light. The operator then continues the analysis by repeating steps 4 and 5 as often as is necessary.

It should be noted that a penlift switch enables the operator to raise and lower the plotting-board pen at will. He would raise it if, for example, he returns from a dendrite ending to an already passed branch point in order to then measure and plot the other branch originating at that point. In such return tracings the printer does not record the distance traversed.

The above sequence draws a 2-D image on the plotter and prints a list of measurements of the segments drawn. To determine the real 3-D distance from the X-Y-Z coordinates the Pythagorean theorem is used and of course this requires squaring and square rooting a lot of numbers. How do we do this? Import the values into Excel? Press the square root button on the calculator? No. We use a circuit that incorporates the QUADRATRON. If Captain Kirk had one of these he wouldn’t have had to rely on Dr. Spock so much. Here’s what the quadraton does:

Squaring and square rooting: The output of each of the input amplifiers is the difference, in voltage, between the present coordinate and the past coordinate measurement. The required measurement is the distance, in three dimensions, between the two consecutive chord points:

s= [(Xi – Xi-1)2 + (Yi – Yi-1)2 + (Zi – Zi-1)2]1/2

 Camera lucida (a) and computer-microscope (b) drawings of the dendrite systems on a pyramidal cell in rabbit cortex.This value is conveniently obtained in analog circuitry by means of the Quadratron. This device was developed by Kovach and Comley and employs temperature compensated varistors. The accuracy of the Douglas Quadratrons is specified to be 0.2 per cent of the full scale output of 100 V (square transfer function: eo = 0.01ei2). Here, this corresponds to 100C and so the error produced by squaring is less than 0.2µ The operation of square rooting introduces a similar 0.2 per cent error. The total error produced by the triangulation circuitry will be a function of the magnitude of the individual coordinate distances. A chord whose components are 100µ in each of the three coordinates will be measured with a maximum error of about 0.3 per cent. On the other hand, a chord whose components are 10µ in each of the three coordinates will be measured with a maximum error on the order of 3 per cent. This error is smaller than that introduced by the linear motion transducers for distances smaller than about 50µ.

A necessary step before the squaring and square-rooting operations is the taking of the absolute value of the coordinate distances measured. This is performed by means of a circuit described by Howe requiring two operational amplifiers per channel.

So in the end you have a drawn cell, the X-Y-Z coordinates and 3-D measurements. The figure above shows the hand drawn and plotter reconstructed neuron.