Clinical and phantom validation of a deep learning based denoising algorithm for F-18-FDG PET images from lower detection counting in comparison with the standard acquisition

Background

PET/CT image quality is directly influenced by the F-18-FDG injected activity. The higher the injected activity, the less noise in the reconstructed images but the more radioactive staff exposition. A new FDA cleared software has been introduced to obtain clinical PET images, acquired at 25% of the count statistics considering US practices. Our aim is to determine the limits of a deep learning based denoising algorithm (SubtlePET) applied to statistically reduced PET raw data from 3 different last generation PET scanners in comparison to the regular acquisition in phantom and patients, considering the European guidelines for radiotracer injection activities. Images of low and high contrasted (SBR = 2 and 5) spheres of the IEC phantom and high contrast (SBR = 5) of micro-spheres of Jaszczak phantom were acquired on 3 different PET devices. 110 patients with different pathologies were included. The data was acquired in list-mode and retrospectively reconstructed with the regular acquisition count statistic (PET100), 50% reduction in counts (PET50) and 66% reduction in counts (PET33). These count reduced images were post-processed with SubtlePET to obtain PET50 + SP and PET33 + SP images. Patient image quality was scored by 2 senior nuclear physicians. Peak-signal-to-Noise and Structural similarity metrics were computed to compare the low count images to regular acquisition (PET100).

Results

SubtlePET reliably denoised the images and maintained the SUV_max values in PET50 + SP. SubtlePET enhanced images (PET33 + SP) had slightly increased noise compared to PET100 and could lead to a potential loss of information in terms of lesion detectability. Regarding the patient datasets, the PET100 and PET50 + SP were qualitatively comparable. The SubtlePET algorithm was able to correctly recover the SUV_max values of the lesions and maintain a noise level equivalent to full-time images.

Conclusion

Based on our results, SubtlePET is adapted in clinical practice for half-time or half-dose acquisitions based on European recommended injected dose of 3 MBq/kg without diagnostic confidence loss.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40658-022-00465-z.

Comparison of Patient Skin Dose Evaluated Using Radiochromic Film and Dose Calculation Software

PURPOSE

To compare, in an interventional radiology setting, peak skin doses (PSDs) delivered as calculated using a dedicated software tool and as measured using radiochromic film. To assess the utility of this dose calculation software tool in routine clinical practice.

MATERIALS AND METHODS

First, radiochromic films were positioned on the examination table in the back of an adult anthropomorphic phantom to measure PSD, and X-ray examinations were simulated. Then, films were again positioned in the patient's back for 59 thoracic or abdominopelvic endovascular interventions. The results obtained with the radiochromic films were taken as a reference and were statistically compared with those of the software.

RESULTS

With measured PSDs ranging from 100 to 7000 mGy, the median software-film difference was 8.5%. Lin's concordance coefficient was 0.98 [0.97; 0.99] (p < 0.001), meaning that concordance was excellent between the two methods. For the films where PSD exceeded 1000 mGy, the median difference in the measured value was 8.7% [- 1.3; 21.1], with a maximum discrepancy of 34%. Lin's concordance coefficient was 0.98 [0.96; 1] (p < 0.001), meaning that concordance was excellent between the two methods.

CONCLUSION

Comparison between radiochromic films and the software tool showed that the software is a suitable tool for a simple and reliable estimation of PSD. The software seems to be a good alternative to films, whose use remains complex.

Imaging of the distribution of (90)y-ibritumomab tiuxetan in bone marrow and comparison with pathology

PURPOSE

Radioimmunotherapy with anti-CD20 antibodies (Abs) labeled with beta-emitters is now used in the treatment of non-Hodgkin's lymphoma (NHL). Because (90)Y is a pure beta-emitter, no direct image of its distribution can be obtained in humans. In this paper, we present in this study imaging data of (90)Y-Ab distribution in human-mantle-cell lymphoma within a mouse model. Describing the actual distribution of the radionuclide at the level of particles range may have important impact on patient dosimetry and therapy treatment planning.

EXPERIMENTAL DESIGN

NOD/SCID mice were grafted with a human NHL cell line that involves the bone marrow. The mice were treated with (90)Y-ibritumomab tiuxetan (Zevalin); Schering AG, Germany) and sacrificed 2 hours after Zevalin administration. Tissue sections were then prepared and viewed under conventional microscopy. The distribution of the radioactivity in mouse femur was determined by using digital autoradiography and subsequently correlated with immunohistochemical results.

RESULTS

Various extent of bone marrow infiltration was investigated and found to be reproducible. Zevalin uptake was heterogeneous within the bone marrow. However, unspecific mouse monoclonal uptake by accessory myeloid cells gave nonspecific background radioactivity. Treating mice with an irrelevant mouse IgG1 monoclonal antibody (mAb) before Zevalin injection controlled this unspecific uptake, and images were strongly correlated with bone marrow infiltration on histologic analysis.

CONCLUSIONS

Our model was reproducible, and allows for the study of various bone marrow involvement with good sensitivity. We demonstrated that imaging of the beta-emitter was possible with good image quality and that (90)Y-Zevalin is distributed heterogeneously within bone marrow. These data suggest that detailed pharmacokinetics may be developed with this model.

Validation of the EGS usercode DOSE3D for internal beta dose calculation at the cellular and tissue levels

Internal radiotherapy is currently focusing on beta emitters such as 90Y or 131I because of their high-energy emissions. However, conventional dosimetric methods (MIRD) are known to be limited for such applications. They are unable to take into account microscopic radionuclide distribution because standardized anthropomorphic phantoms are used, and absorbed dose is calculated at the organ level. New tools are therefore required for dose assessment at cellular and tissue level (10-100 microm). The purpose of this study was to validate, at this scale, a Monte Carlo usercode (DOSE3D), based on the MORSE combinatorial geometry package and the EGS code system. Dose point-kernel calculations in water were compared to those published by Cross et al and Simpkin and Mackie. They confirm that DOSE3D is a reliable tool for cellular dosimetry in various geometric configurations.