HYBRID Publications (2): „Improving the comparison of PET images before and after cancer treatment”

Last year, Guilherme Domingues Kolinger published the first results of his HYBRID project in EJNMMI Research. In this paper he describes how the analysis of PET images can be made more reliable (link to publication below).

The aim of Guilherme’s project is to improve data analysis from PET images. He investigates which methods can be used to analyse PET images in a consistent and reliable way. This is essential to ensure that conclusions drawn from PET images are trustworthy and comparable for different patients and hospitals. This way, PET images allow for in depth investigation of disease features, e.g. in cancer and Alzheimer’s Disease.

In his first paper, Guilherme aimed to find the best method to analyse PET images so that scans taken on different days can be compared to each other. He investigated the ‘activity’ of tumours in lung cancer patients. His goal was to find reliable ways to analyse PET images, so that the image only reflects changes in the tumour, and not ‘normal’ changes in the patient’s body that happen from day to day.

PET scans

A PET scan uses so-called ‘tracers’ to track functions in our body. One of the most used PET tracers is a molecule called [¹⁸F]FDG. This is a molecule that is very similar to glucose, and so our body uses it like it uses glucose. But PET tracers have a special function: they are radioactive*. That way, a PET scanner can detect it and discover places in the body that use large amounts of glucose. Of note, tumours use a lot of glucose. Therefore, a PET scan can visualize them. (*PET tracers are injected in so small amounts that the radioactivity does not harm the patient.)

When a cancer patient is treated, the glucose use of a tumour before and after treatment can be measured with a PET scan. The amount of [¹⁸F]FDG (glucose) that the tumour uses is compared for both time points. The glucose use of a tumour shows which parts of the tumour are more active. In this way, PET images allow physicians to see changes in the tumour before they are visible in other images. Changes in glucose use by a tumour happen before the size changes, and images from traditional CT and X-rays only show the size and shape of the tumour. PET imaging shows how much glucose a tumour uses, and therefore how active a tumour is.

To calculate the tumour activity before and after treatment, the PET images need to be comparable. That is difficult because these images are affected by several aspects of clinical routine: the time between tracer injection and PET scanning, settings used for creation of the image, and the method for drawing the tumour in the image. In this study, Guilherme compared how all of this can affect the ‘repeatability’ of PET scans taken on two different days, and therefore how reliable analysis of PET images can be.

Tracer uptake time

The first factor that influences a PET image is the tracer uptake time. This is the time between the injection of the tracer and the time of the scan. By waiting longer, more tracer can reach the tumour, and therefore make the image brighter. But too much of the tracer in the tumour can harm the image quality. The image will then be so bright that it is not informative anymore. Differences in uptake times will also affect the data obtained from the PET images, which could affect the treatment plan of a patient. Therefore, it is very important to follow all guidelines and recommendations when performing a PET scan.

In this study, Guilherme investigated the effect of different tracer uptake times on the repeatability of PET scans. To do so, he looked at two time points after the injection of the tracer: one hour and one-and-a-half hours. The scans were repeated on the next day. Guilherme discovered that the tracer uptake time did not affect the repeatability of PET scans taken on consecutive days.

Image creation

Secondly, Guilherme investigated two different reconstruction methods. A PET ‘image’ is quite different from e.g., an X-ray. An X-ray directly results in an image, while the PET scanner measures the radioactive signal from the injected tracer and needs to “rebuild” the image from this signal. To get an image, a series of mathematical models has to be applied, the so called ‘reconstruction methods’. Differences in these methods can affect the way a PET image is analysed.

Guilherme compared two reconstruction methods. One of these generates images that are more easily comparable with other patients, while the other optimizes the visualisation of the tumour. He found out that these reconstruction methods did not affect the repeated measurements of the PET scans.

Drawing the tumour

The final factor Guilherme investigated was the delineation method. Delineation is the process of ‘drawing’ the tumour outline based on a PET image. An expert in medical imaging can do this, sometimes aided by a computer programme. Computer programmes make the process faster and more reliable.

Guilherme compared four different delineation methods and two combinations of these methods (‘consensus methods’). He also compared the accuracy of these methods to delineations performed by experts. The delineation was best performed by one of the ‘consensus methods’. But when Guilherme looked at the repeatability between two days, a single delineation method also sufficed.

Conclusion

Guilherme concludes that for a delineation that is reliable and repeatable, a ‘consensus method’ based on three or four delineation methods is the best option. When this is not possible on clinical routine software, one single delineation methods can also be an option. The tracer time uptake and image reconstruction method have less impact on the repeatability of PET scans.

This study identifies reliable methods that improve the repeatability of PET scans. This can be used for assessing whether a treatment in successful or not.

Publication: Kolinger, G.D., Vállez García, D., Kramer, G.M. et al. Repeatability of [¹⁸F]FDG PET/CT total metabolic active tumour volume and total tumour burden in NSCLC patients. EJNMMI Res 9, 14 (2019). https://doi.org/10.1186/s13550-019-0481-1

Image: PET image from iStock, not related to the publication.