IOLMaster — Axial Length Measurements
Signal-to-Noise Ratio (SNR).

(Data used with permission, courtesy of Dr. W. Haigis, University Wuezburg, Germany)

1) When looking at the composite signal's SNR, is a HIGHER value better than a lower value?

In principle, the higher the SNR, the better the signal quality – up to a limit. An SNR above 2.0 for the composite signal is sufficient for a good signal quality. Significantly higher levels (e.g., SNR=100) or extremely high levels (e.g., SNR=1000) do not provide appreciably better results.

2) How is the composite SNR calculated?

SNR is calculated as signal amplitude divided by noise amplitude of the composite signal.

3) Is there a range of "good" or "bad" composite SNR?

In Version 4, our target was to get above 2. Correct, a good quality signal has an SNR above 2.0. This is still the threshold for single measurements as well as for the new composite signal in Advanced Technology Software Version 5. As long as the SNR is above 2, you will have a very good composite signal. It does not matter whether the composite SNR is 100, 200, 400, or 1000. The SNR value itself is not enough to determine which is better. Concentrate on the composite signal's shape. Does it have a clear single peak? The limits for borderline values are still 1.5 for a single measurement and 1.6 for the composite signal.

4) After a couple of axial length (AL) measurements have been performed that result in fair SNR readings, the composite signal shows a high SNR. This high SNR can be increased by an additional measurement that showed an “Error” warning, instead of an AL value. Why does an "Error" AL reading contribute to a higher composite SNR?

The “Error” warning is displayed instead of an AL value when the SNR is below 1.5, but a low SNR signal can contain significant information (i.e., significant signal reflection from the retina) that is lost in the noise. This is why Advanced Technology Software Version 5 with its composite analysis is so powerful. Noise is a random process, and statistically it will cancel out in the composite signal. In contrast, the main peak will be higher in the composite because even a tiny peak, otherwise lost in the noise, will contribute to the primary peak in the composite signal. Do not worry about SNR values or “Error” warnings (SNR<1.5) on single measurements, but concentrate on the composite signal. Do not delete single measurements (even those with the “Error” warning) because even a noisy signal can contribute significantly to the composite signal. Especially on a really dense cataract eye, you will never receive an AL on a single measurement. You will most likely receive only “Error” warnings, as the SNR is far below 1.5. With the composite signal, however, the system will often be able to determine the AL.

5) Why in some cases is the composite signal's SNR, analyzed from 5 AL measurements, lower than the composite signal's SNR, analyzed from 2 AL measurements? Shouldn't the composite signal's SNR increase with each additional measurement?

The composite signal analysis algorithm was designed and optimized to extract AL information from very low SNR signals and to process an evaluable composite signal with acceptable SNR from low SNR single signals. For high SNR signals, the algorithm is not linear. If you add high SNR measurements to already high existing ones, the composite SNR does not necessarily increase. Usually additional noise will be added, as all 2 measurements are subject to statistical fluctuations. Consequently, there will be a small deviation between the AL values in repeated measurements that will broaden the narrow AL signal peak and lead to a lower SNR. Only in cases where you hit exactly the same spot for each individual measurement will you receive an extremely high SNR of more than a 1000. A high composite SNR (green traffic light) is sufficient. On a really dense cataract eye any composite SNR above 2.0 should be considered good.

6) I noticed that V5 is slower than V4 in AL measurement. Why?

Advanced Technology Software Version 5 performs signal processing while measuring. This processing includes: Noise filtering of the single measurement signal before it is added to the composite signal. Calculation of the new composite signal (including the last measurement).

7) Have the AL readings between V4 and V5 been compared in a study?

Yes, Prof. Haigis presented the results of his “Advanced vs standard signal analysis IOLMaster” comparison at the ESCRS 2007. Please refer to Slide 3 of the presentation of Dr. Haigis at ESCRS 2007.

8) Has the accuracy of K measurements obtained with V4 and V5 been compared in a study?

Please refer to the presentation of Dr. Haigis at ESCRS 2007, Slides 11 through 14.

9) Five AL measurements have been performed. One shows an AL value with an SNR of 12.2, but 4 show the “Error” warning. Although the composite signal has an SNR of 150.1, an “Evaluation!” warning is shown instead of an AL value, and no AL is transferred to the IOL power calculation dialog. What is going on?

Advanced Technology Software Version 5 has a safety mechanism. As long as only one measurement contains significant information about the AL, the software does not show or provide an AL value for further calculation. The “Evaluation!” warning appears instead. Of course, the software is calculating the composite graph and shows an SNR.

This scenario is unlikely to happen when measuring a patient's eye. In the example above, measurements on different objects were added to the measurement on the patient's eye to achieve the error condition 3

10) I did AL measurements of my eye on two separate IOLMasters with Advanced Technology Software V5. Measurement with both instruments resulted in the same AL but different composite SNR. Is there anything wrong with the IOLMaster that showed lower composite SNR?

Not necessarily. All measurements are subject to statistical fluctuations. Only if you hit the same spot on the retina in exactly the same angle will you get the same SNR reading. This is very difficult to do on a patient's eye, as the patient is not always fixating properly (e.g., because of micro saccades).

11) A couple of AL readings were printed in "RED" and the “multiple peaks” warning was shown. I deleted them, and the composite value sometimes changed, and sometimes it remained unchanged. Why?

As long as the composite signal contains any significant AL information, each single measurement AL (if available) is compared with the composite AL. Outliers, which differ more than 0.05 mm, are printed in RED and the warning “multiple peaks” is shown if there is at least one outlier.

An outlier measurement still can contain information on the "correct" peak as well as the “outlier” peak. Do not delete measurements, even those printed in RED. Instead, take a closer look at the composite signal and the composite peak's shape ("Chrysler Building" single peak or "Sears Tower antennas" double peak).

Go through the single measurements and and try to determine which peak is the correct one, especially if the composite peak's shape is like the "Sears Tower antennas" double peak. If possible, perform additional measurements. If necessary, move the measurement cursor (circle) to the correct peak on the blue composite graph. The AL measurement for the other eye may help you decide which is the correct peak.

12) What are the guidelines for best practice with the Advanced Technology Software Version 5 and its composite signal?

• A good composite signal has a SNR above 2.0.
• A borderline value is between 1.6 and 2.0.
• Below 1.6 no AL is displayed.

Take a closer look at the composite signal peak's shape ("Chrysler Building" single peak or "Sears Tower antennas" double peak), especially if the warning "multiple peaks" is shown. Go through the single measurements and try to determine which peak is the correct one, especially if the composite peak's shape is like the "Sears Tower antennas" double peak. If possible, perform additional measurements. If necessary, move the measurement cursor (circle) to the correct peak on the blue composite graph. The AL measurement for the other eye may help you decide which is the correct peak.

Do not delete single measurements because they are identified as outliers (AL is printed in RED) or have low SNR (“Error” warning). The reason is that the software already recognized them as outliers and has printed the AL in RED. Even a low SNR signal may contain significant information for the correct AL. That is the rationale for using the composite signal.

 

The IOLMaster with software version 4 requires five measurements to be taken. The number of measurements is crucial. After the fifth measurement, if the individual axial lengths are consistent, IOLMaster calculates the arithmetic mean value, and passes it on for use in the IOL calculation, which enables an evaluation.