Make your own free website on

Densitometers & Densitometry: Points to consider


Good image analysis has to begin with good images! Even the analysis of single-dimension gel or TLC images is still a computational problem in three dimensions—length, breadth (area) and optical density. The image acquisition system should optimally have sufficient resolution to capture sufficient data points within the smallest bands of interest and be operating within its dynamic range. Sensitivity may also be an issue for gels being imaged by fluorescence or radiation imagers. Frequently this is a simple matter of checking because most manufacturers supply the necessary information. For accurate image analysis, calibration of the image capture device is essential and this will be critical if the device is operating outside its linear response region (assuming it has a linear response region). For critical work a standard such as a neutral density standard can be acquired at the same time as the experimental image to confirm the response curve of the image capture device and alert the experimenter to readings that are probably outside the range of the imaging device being used. Other considerations apply with stained samples, such as the uniformity of staining and destaining. Inhomogeneities can be partially compensated by suitable internal standards and by band matching and averaging. In summary, for good results, attention to experimental design and operating conditions will repay the experimenter. Hints and suggestions in these pages are design to assist.

Strip Densitometers

1D gels can be quantitated using this type of densitometr (e.g.Joyce-lobel). A wide range of optical wedges can be used, leading to a wide dynamic range (up to and possibly beyond 3 OD). The drawback is the limited throughput with this instrumentation. It is not really well suiuted to 2D gel measurement.

Scanning Densitometers

A colimated beam (frequently now a laser) is scanned over the film surface, sometimes driven around the light beam mounted on a drum. Transmitted and reflected incident beams can be measured. These instruments can cope with a broad dynamic range an have good spatial resolution making them very suitable for 2D gels.

Requirements for Densitometry


Well Separated Bands Merged bands are a source of error.
Well Separated Lanes Use well formers (combs) that provide for good lane separation rather than high sample throughput. Diffusion between lanes will add to error.
Avoid Overloading & Underloading Must work within linear range of image capture device
Suitable Visualisation Method Choose a suitable stain or visualisation method: eg. Whilst silver stain is very sensitive it can give rise to significant quantitation errors.


Traditional Densitometers Optically superior to modern instruments, but analytically - less effective. eg. LKB laser densitometer, Joyce Loebl Chromoscan, Shimadzu, etc.
Imaging Densitometers Molecular Dynamics
Radiation / Fluorescence Imagers Molecular Dynamics, Fuji, Bio-Rad, Canberra Packard & others
CCD Camera Systems Standard Molecular Biology systems from UVP, Bio/gene, Pharmacia, Appligene, Stratagene, Bio-Rad, etc.
Very low light systems for fluorescence and chemiluminescence. eg. Photometrics.
Desk top Scanners Desktop scanners: Agfa, Sharp, Epson, HP, etc. - Often sold as densitometers from companies such as Phoretix, Bio-Rad and Pharmacia.

Optical Density (OD) Range and Linearity

Media / Device

Optical Density Range

CBB stained PAGE Gels

0 to 1.5

DNA / Ethidiun Bromide

0 to 1.0

General x-ray film

0 to 1.5

High quality x-ray film

0 to 3.0

Document scanner

0 to 1.8

High grade scanner

0 to 3.0

Imaging densitometer

0 to 3.5

Traditional densitometer

0 to 4.0

Radiation Imager

0 to 6.0

Intensity Calibration

As light passes through increasing levels of darkness, the absorbance of the light is increased exponentially which means that image reproduction is non-linear. This can easily be seen by scanning an optical density photographic tablet and analysing the scan with a product such as Phoretix 1D. There are several ways of overcoming this nonlinearity. The first is a simple mathematical conversion which provides a reasonable approximation. The second is with a calibration procedure whereby measured values of image intensity are associated with calibration optical density values. In Phoretix software this is known as intensity calibration. Some instruments are pre-calibrated, or can be calibrated to adjust for nonlinearity of image reproduction. Document scanners and video cameras are often calibrated for reflected light rather than for light absorbance.

Quantity Calibration

Whilst Intensity Calibration is undoubtedly important it can become unnecessary if it is decided to calibrate according to spots or bands on the image which have known values. Several staining methods are not linear. In such cases Quantity Calibration is the preferred method. This allows all results to be calculated from a standard curve created from the measured values of standard bands or spots with their known values. Frequently, a few lanes are set aside for this purpose for samples prepared using serial dilution. It is not necessary to use both intensity calibration and quantity calibration.

Background Subtraction

The background of most gel images varies significantly throughout the image. This means that background subtraction should account for differences between lanes and along the length of each lane. Constant value methods are general not very helpful. For standard 1D gels choose a method that is visually acceptable on the lane profile. Local area background subtraction per band is only useful if you are certain that no part of the band is included in the calculation of the background value (NB: Be careful about judging this from examination of the image because the eye does not see all the grey levels). In Phoretix 1D background subtraction can be calculated from the profile or from the image. The profile methods are much more useful for standard 1D electrophoresis, or TLC applications. In 2D electrophoresis the preferred methods of background subtraction relate to the values around each individual spot. The accuracy of this can be dependent on the parameters used for spot detection and the methods of spot editing used. If it is clear that spot detection is such that generous spot boundaries are present then the average of each spot boundary would be a good selection. With less generous areas for spot detection, lowest of boundary would be better.

Band limit determination

Regardless of whether all band material (“whole band analysis”), or an averaged slice through band material is used during quantitation, careful consideration should be given to the determination of band or spot boundaries. Traditionally in standard 1D gels, band boundaries have been detected by computer analysis of the lane profile with manual editing (as per HPLC etc.). This method is fine, especially if the lane image is shown together with the graphical profile. Standard imaging methods to determine boundaries should be viewed with caution, depending on the algorithms used. This is because they often do not take into account background variations. Of course this is not the case with the Phoretix 2D spot detection algorithm which calculates the boundaries of each spot independently. One common source of error is to use “objects” of a constant size for identifying the area for measurement which approximate the shape of the spot or band. For example the use of circles for dot blots. When such a method is used the size of the object should either be variable between spots, or should approximate the size of the largest spot. Otherwise a significant proportion of material may be omitted from measurement.

Merged bands

The presence of merged bands or spots is always a problem. There are mathematical methods of separating and recalculating merged spots and bands. However they are dependant on certain assumptions relating to the shape of the spots and bands that may not be true. In 1D electrophoresis the shape of a band profile can be dependent on a number of factors such as the uniformity of the gel matrix, the amount of band material, the amount of band diffusion, the speed of electrophoresis and the method of electrophoresis. Thus, mathematical deconvolution methods can be prone to error. Often the best approximation is to simply divide merged bands at the merge point. However, it should be noted that when large bands are merged with small bands, there is often material belonging solely to the large band on the other side of the small band and that the hight of the small band is largely influenced by material from the large band.

Margins of error

Strictly speaking, each analysis should be performed at least three times. A “best” analysis and two analyses using alternative worst case scenarios for background subtraction and band edge determination will enable identification of margins of error.