One tumour: three subpopulations

Now, a comprehensive molecular imaging study has shown that many tumours actually comprise three distinct subpopulations of cells. At the American Association of Physicists in Medicine (AAPM) annual meeting, held this week in Philadelphia, PA, Chihwa Song from the University of Wisconsin-Madison explained the findings.

"Tumours are known to be heterogeneous for many characteristics, such as metabolism, cell proliferation and hypoxia," Song told the AAPM delegates. "Molecular imaging can provide spatial distributions of these characteristics within the tumour."

Song and colleagues performed PET/CT scans on a series of patients with head-and-neck cancer. ¹⁸F-FDG was used to measure metabolism, ¹⁸F-FLT to assess cell proliferation and ⁶¹Cu-CuATSM to gauge hypoxia. All three of these factors can impact how a tumour reacts to treatment.

The researchers used clustering algorithms to classify voxels within the PET images of 11 patients, grouping the voxels according to the similarity of their values. They used these data to partition tumours into subpopulations, with the aim of studying the characteristics of each of the groups. Most of the tumours could be optimally divided into three distinct subpopulations, spatially separated into peripheral, core and intermediate regions, according to their fraction of surface voxels.

The mean standard uptake values (SUV) were calculated for each subpopulation and normalized to the mean SUV for the entire tumour. Song noted that the three subpopulations exhibited distinct characteristics. Peripheral cells were seen to exhibit low uptake of FDG and FLT, cells in the core exhibited high uptake of both FDG and FLT, while the intermediate subpopulation exhibited high FDG and medium FLT uptake. The uptake of CuATSM was similar for all three.

The researchers note that this classification into distinct subregions may be generalizable to many different kinds of cancer. Such information could be employed for dose painting, by increasing the dose delivered to radiation-resistant cells and lowering the dose to radiation-sensitive cells, for example. Ultimately, this could enable better design of personalized cancer therapy.

"Tumour subpopulations with distinct characteristics were spatially separated," Song concluded. "Tumour phenotypes were more strongly correlated within each subpopulation than in the whole tumour."