Deep Learning-Driven Estimation of Centiloid Scales from Amyloid PET Images with 11C-PiB and 18F-Labeled Tracers in Alzheimer's Disease

BACKGROUND

Standard methods for deriving Centiloid scales from amyloid PET images are time-consuming and require considerable expert knowledge. We aimed to develop a deep learning method of automating Centiloid scale calculations from amyloid PET images with 11C-Pittsburgh Compound-B (PiB) tracer and assess its applicability to 18F-labeled tracers without retraining.

METHODS

We trained models on 231 11C-PiB amyloid PET images using a 50-layer 3D ResNet architecture. The models predicted the Centiloid scale, and accuracy was assessed using mean absolute error (MAE), linear regression analysis, and Bland-Altman plots.

RESULTS

The MAEs for Alzheimer's disease (AD) and young controls (YC) were 8.54 and 2.61, respectively, using 11C-PiB, and 8.66 and 3.56, respectively, using 18F-NAV4694. The MAEs for AD and YC were higher with 18F-florbetaben (39.8 and 7.13, respectively) and 18F-florbetapir (40.5 and 12.4, respectively), and the error rate was moderate for 18F-flutemetamol (21.3 and 4.03, respectively). Linear regression yielded a slope of 1.00, intercept of 1.26, and R2 of 0.956, with a mean bias of -1.31 in the Centiloid scale prediction.

CONCLUSIONS

We propose a deep learning means of directly predicting the Centiloid scale from amyloid PET images in a native space. Transferring the model trained on 11C-PiB directly to 18F-NAV4694 without retraining was feasible.

Quantitative Evaluation of 18F-Flutemetamol PET in Patients With Cognitive Impairment and Suspected Alzheimer's Disease: A Multicenter Study

Background: In clinical practice, equivocal findings are inevitable in visual interpretation of whether amyloid positron emission tomography (PET) is positive or negative. It is therefore necessary to establish a more objective quantitative evaluation method for determining the indication for disease-modifying drugs currently under development.

Aims: We aimed to determine cutoffs for positivity in quantitative analysis of ¹⁸F-flutemetamol PET in patients with cognitive impairment and suspected Alzheimer's disease (AD). We also evaluated the clinical efficacy of amyloid PET in the diagnosis of AD. This study was registered in the Japan Registry of Clinical Trials (jRCTs, 031180321).

Methods: Ninety-three patients suspected of having AD underwent ¹⁸F-flutemetamol PET in seven institutions. A PET image for each patient was visually assessed and dichotomously rated as either amyloid-positive or amyloid-negative by two board-certified nuclear medicine physicians. If the two readers obtained different interpretations, the visual rating was rerun until they reached consensus. The PET images were quantitatively analyzed using the standardized uptake value ratio (SUVR) and standardized Centiloid (CL) scale with the whole cerebellum as a reference area.

Results: Visual interpretation obtained 61 positive and 32 negative PET scans. Receiver operating characteristic analysis determined the best agreement of quantitative assessments and visual interpretation of PET scans to have an area under curve of 0.982 at an SUVR of 1.13 and a CL of 16. Using these cutoff values, there was high agreement between the two approaches (kappa = 0.88). Five discordant cases had SUVR and CL values ranging from 1.00 to 1.22 and from 1 to 26, respectively. In these discordant cases, either diffuse or mildly focal elevation of cortical activity confused visual interpretation. The amyloid PET outcome significantly altered the diagnosis of AD (χ² = 51.3, p < 0.0001). PET imaging elevated the proportions of the very high likelihood category from 20.4 to 46.2% and the very low likelihood category from 0 to 22.6%.

Conclusion: Quantitative analysis of amyloid PET using ¹⁸F-flutemetamol can objectively evaluate amyloid positivity using the determined cutoffs for SUVR and CL. Moreover, amyloid PET may have added value over the standard diagnostic workup in dementia patients with cognitive impairment and suspected AD.