This is to our knowledge the first study on hippocampal volumes after low dose radiation to the basal part of the adult human brain. Although the dose to the hippocampus could not be calculated directly, the estimates of radiation dose to relevant areas are probably more exact than in any previous study. The small study size is an effect of our ambitions to minimize the influence of confounding factors in an original sample of 101 patients. Moreover, the patient sample is homogeneous in terms of treatment with the exception of two patients who received a higher radiation dose; these were treated separately in the statistical analysis.
Automatic methods for hippocampal volumetry are rapidly gaining acceptance. They have undisputed advantages in terms of cost, inter-rater reliability and comparability between studies. However, for small-scale studies involving only intra-study comparisons, we would argue that manual segmentation is still superior. This is even more so since the issue of ICV normalization has not been satisfactorily resolved for the most used automatic method .
The reliability results for the volumetric method are acceptable. Since the main results in the study are based on differences between patients or groups, the most relevant measure when comparing the results of the two raters is consistency ICC which does not take systematic (non-random) differences between the raters into account. Importantly, consistency ICC was considerably higher than absolute measure ICC which reflects that the latter was strongly influenced by such a systematic difference. When interpreting the reliability figures, one should also bear in mind that the method included segmentation of the hippocampal tail, which is the most difficult part and adds variation in comparison with not including the tail (data not shown).
The size of the observed interindividual variation in hippocampal volumes as reported in Table 4 accords with recently published data from healthy subjects . Other facts that speak in favour of the validity of our measurements are that the observed volumetric differences between left and right hippocampal volumes and between men and women, as well as the negative correlation with age in the male group, are in general accord with main trends among earlier findings [31, 36, 38, 39]. The L/R difference and the age correlation were similar in patients and controls. The influence of handedness could not be tested properly since the number of left-handed subject was too low to admit any statistically meaningful subgroup calculation.
Intracranial volume, ICV, was measured with a highly reliable manual method. Normalization of hippocampal volumes with respect to ICV eliminated the gender differences and reduced the overall variance. Somewhat surprisingly, the latter does not hold for the female group. This is probably a statistical artifact since there were only seven females while the normalization was based on a regression in the whole sample of 30 subjects.
It could be argued that the sample is small and that the study therefore has insufficient power. This argument would have had a point if our only result had been that the mean volume difference between patients and controls was not significantly different from zero. However, all observed mean volume differences were close to zero, and the 95% confidence interval for the difference in total normalized volume does not include larger deficits in the patient group than 8% of the group mean. Using the data from the second rater would have given very similar results. Hence our results constitute solid positive evidence that low dose radiation to the basal brain in adults does not cause a lasting, major volume reduction of the hippocampi. The lowered quality of life in our patient group stands in need of some other explanation.
The neurogenic cells in the gyrus dentatus are the most radiosensitive elements of the hippocampus and a subregion analysis would have added important information. However, such an analysis is not feasible on 1.5 T data. Further, experimental and clinical evidence support the thought that the cognitive effects seen long after low dose radiation to the brain are at least partly mediated by indirect effects on other structures than the gyrus dentatus.
The method described in this work may be of value in the future considering the change in radiation treatment techniques that are being introduced, such as IMRT (intensity modulated radiation therapy) and SRT (stereotactic radiation therapy). The radiation dose will be better targeted and controlled with these techniques, but the areas receiving low dose will be much larger than after the traditional methods used in our study. It is not known what this means biologically and clinically, and it has to be studied and documented carefully. Our study should be seen as a part of this work.