Overall, 307 patients, with a mean age of 72.1 ± 8.4 years (ranging from 48 to 91 years), were included in this study. The mean length of education was 9.5 ± 4.8 years (n = 184). The cohort comprised 131 males (42.7%) and 176 females (57.3%). Based on the BAPL score assessments by the physician, 108 and 199 patients were classified as negative (BAPL 1) and positive, respectively. Among the positive group, 42 patients had BAPL 2, while 157 patients had BAPL 3. Thus, 199 and 108 patients were classified as positive and negative, respectively (Table 1).

Table 1 shows no significant differences in variables such as sex, body mass index (BMI), height, and body weight across the groups, indicating these factors may not significantly influence the detection of amyloid deposition through PET scanning in this cohort. Additionally, no significant variations were observed in the mean injection dose of the tracer and the uptake time for the PET scans among the groups. In contrast, the data showed statistically significant differences in age and education level (p-values of 0.027 and 0.001, respectively). The lower p-value for education suggests a stronger association.

The study also highlights significant differences in cognitive performance over time between patients with positive and negative PET scans. Patients with positive scans exhibited a decline in cognitive function, as shown by initial and follow-up scores. Initially, the CDR-SB score for the positive group (3.2 ± 2.3, range 0–16) was significantly higher than that for the negative group (2.2 ± 1.6, range 0–8), with a p-value of 0.006. Furthermore, the initial MMSE scores were significantly lower in the positive group (21.3 ± 4.9, range 6–29) than in the negative group (24.4 ± 4.2, range 13–30), with a p-value of < 0.001. Follow-up scores showed similar trends, with increased CDR-SB scores and decreased MMSE scores over time (p < 0.001), indicating a deterioration in cognitive function associated with amyloid pathology.

We identified 199 positive and 108 negative PET scans using visual interpretation. In the positive group, SUVR values were 1.57 ± 0.21 (global), 1.63 ± 0.23 (frontal), 1.70 ± 0.25 (PCC-precuneus), 1.52 ± 0.22 (lateral temporal), and 1.57 ± 0.21 (lateral parietal). In contrast, the negative group exhibited SUVR values of 1.19 ± 0.06, 1.21 ± 0.07, 1.19 ± 0.08, 1.13 ± 0.06, and 1.19 ± 0.07, respectively.

Similarly, the Centiloid scale values in the positive group were 73.25 ± 32.56 (global), 49.00 ± 27.2 (frontal), 57.76 ± 31.06 (PCC-precuneus), 35.06 ± 26.51 (lateral temporal), and 41.52 ± 25.30 (lateral parietal). For the negative group, the corresponding Centiloid values were 1.23 ± 0.09, 1.25 ± 0.10, 1.26 ± 0.14, 1.17 ± 0.09, and 1.23 ± 0.10, respectively.

Significant differences in total SUVR values were observed between the visually positive and negative groups (p < 0.001), while the Centiloid scale also showed significant differences between these groups (p < 0.001).

The SUVR values demonstrated stability over time (Fig. 2a and c), indicating that the mean SUVR values for positive and negative groups did not significantly vary across imaging durations of 5 min, 10 min, 15 min, and 20 min in all five subregions. The box plots showed relatively stable median values, suggesting consistent tracer uptake in the examined brain regions.

Stability of amyloid PET tracer uptake in positive and negative groups. A comparative analysis of total SUVR and Centiloid values are measured at the different scanning times (5 min, 10 min, 15 min, and 20 min post-injection) based on various reference brain subregions, namely global, frontal, PCC-precuneus, lateral temporal, and lateral parietal. No significant differences were observed across these imaging durations in positive and negative groups across all five subregions, as indicated by total SUVR and Centiloid values. In each subregion, pairwise duration group comparisons were conducted using t-tests. Supp. Tables 1, 2, 3, and 4 show the corresponding p-values. Abbreviations: PET, positron emission tomography; SUVR, standard uptake value ratio; PCC, posterior cingulate cortex. Positive (a, b) and negative (c, d) groups. No significant changes in tracer uptake were observed across the examined time points within each group (p > 0.1)

In the positive group (Fig. 2a and Supp. Table 1), the SUVR values across intervals at 5, 10, 15, and 20 min imaging durations within each target subregion did not differ significantly. Specifically, the global SUVR values were recorded as 1.56 ± 0.20, 1.57 ± 0.21, 1.58 ± 0.20, and 1.58 ± 0.20, respectively. Pairwise comparisons resulted in p-values above the threshold for statistical significance (p > 0.1), indicating no substantial changes in the SUVR measurements over time. Similarly, in the frontal region, mean SUVR values ranged from 1.61 ± 0.22 at 5 min to 1.64 ± 0.22 at 20 min. The PCC-precuneus, lateral temporal, and lateral parietal regions also showed stable SUVR values and non-significant differences in pairwise duration comparisons across these subregions (p > 0.1).

Similarly, the negative group exhibited no statistically significant changes in SUVR values across the scanning times (5, 10, 15, and 20 min) within each analysed subregion. The global SUVR values were 1.18 ± 0.06 at 5 min, 1.19 ± 0.06 at 10 and 15 min, and 1.20 ± 0.06 at 20 min. Pairwise comparisons resulted in p-values > 0.1, indicating stable amyloid deposition over time. In the frontal region, the SUVR values remained consistent at 1.21 ± 0.07, indicating no significant changes over time (p > 0.1). Similar stability was observed in the PCC-precuneus (approximately 1.19 ± 0.08 across all durations) and lateral temporal regions (1.12 ± 0.06), with all p-values > 0.1 (Fig. 2c and Supp. Table 2).

The trend of gradually increasing SUVR values over time was observed in positive and negative groups. From 5 to 20 min, the mean SUVR values in the positive group increased by 0.03, which was higher but not statistically significant, compared to the 0.01 increase in the negative group. In the positive group, SUVR values at 5 min for the global, frontal, PCC-precuneus, lateral temporal, and lateral parietal regions were 1.56 ± 0.20, 1.61 ± 0.22, 1.68 ± 0.24, 1.51 ± 0.21, and 1.55 ± 0.20, respectively. These values increased by 0.03 ± 0.01 at 20 min (Fig. 3a and Supp. Table 1). In the negative group, starting values were 1.18 ± 0.06 (global), 1.20 ± 0.07 (frontal), 1.19 ± 0.08 (PCC-precuneus), 1.12 ± 0.06 (lateral temporal), and 1.18 ± 0.07 (lateral parietal) at 5 min. These values increased by 0.02 ± 0.00 at 20 min (Fig. 3c and Supp. Table 2).

The trend line graphs illustrate the changes in the mean values over four different scanning times. The data is presented for the total patient group, as well as positive and negative groups separately. The figures show the comparison of the total regional SUVR with the total Centiloid in each target region, namely frontal, PCC-precuneus, lateral temporal, and lateral parietal. From 5 to 20 min, the mean SUVR in the positive group increased by 0.03, which was higher but not statistically significant, compared to the 0.01 increase in the negative group. Similarly, the mean Centiloid values in the positive group increased by 4.60 ± 0.44, while in the negative group, they increased by 2.38 ± 0.17. Abbreviations: PET, positron emission tomography; SUVR, standard uptake value ratio; PCC, posterior cingulate cortex. Positive (a, b) and negative (c, d) groups

Regarding magnitude, SUVR values in the positive group were consistently higher than in the negative group, indicating that the values of the positive in all regions were higher than 1.51 ± 0.21, while those in the negative group were lower than 1.21 ± 0.07 in all areas (Fig. 3a and c). In both groups, the PCC-precuneus region showed the highest uptake in the positive group, which peaked at 1.71 ± 0.25 after 20 min, while the lowest SUVR values in the negative group were in the lateral temporal region at 1.12 ± 0.06. No significant difference was observed between the values at 5 and 20 min, with p-values > 0.10 across all target regions (Supp. Tables 1 and 2).

The trends in Total Centiloid measurement mirrored those observed in the Total SUVR analysis (Fig. 2b and d). The positive and negative groups exhibited consistency across the evaluated time points, with no significant differences observed in Centiloid values at each time frame across the five reference regions.

In the positive group, the mean Centiloid values over four scan time were 73.24 ± 1.82 (global), 48.98 ± 1.72 (frontal), 57.75 ± 31.06 (PCC-precuneus), 35.04 ± 26.51 (lateral temporal), and 41.50 ± 25.30 (parietal) (Fig. 2B and Supp. Table 3). In contrast, the values of the negative group were 3.13 ± 0.78 (global), −4.93 ± 0.89 (frontal), −6.7 ± 13 (PCC-precuneus), -14.71 ± 0.82 (lateral temporal), and -6.87 ± 0.96 (parietal) (Fig. 2d and Supp. Table 4). Pairwise comparisons resulted in p-values > 0.05.

The trend of the mean Centiloid values in subregions increased gradually from 5–20 min in both groups. In the positive group, values rose by 4.60 ± 0.44, while in the negative group, they increased by 2.38 ± 0.17 (Figs. 3b and d). The distribution of positive and negative lines differed in the chart, with mean values in the positive group consistently higher than 33.10 ± 25.87. In contrast, values in the negative group remained below 4.20 ± 9.23, indicating a clear distinction between the two groups (Supp. Tables 3, 4).

The Bland–Altman plots in Fig. 4 illustrate the agreement between PET imaging measurements for 5 min and 20 min. In the positive group (Fig. 4a), the mean difference in SUVR values between the 5 and 20 min scans was -0.02 (LoA −0.07 to 0.02, p = 0.265), indicating tight consistency. In the negative group, the mean difference was −0.01 (LoA −0.04 and 0.02, p = 0.129), further demonstrating stable SUVR values over time (Fig. 4c).

Extended Bland–Altman plot analysis for differences between the 5 and 20 min images in SUVR (a, c) and Centiloid (b, d) based on Global values. The presentation of small dots denoting all patients (a, b: positive group; c, d: negative group), confidence interval limits for the mean (blue dashed lines) and limits of agreement (orange shaded areas), demonstrate a small mean difference between the 5 min and 20 min images

In the positive group, Fig. 4b shows the mean difference in global Centiloid measurements between images for 5 and 20 min was −4.88 (LoA −13.34 to 3.59, p = 0.133). In the negative group, the mean difference was −2.20 (LoA −6.92 to 2.53, p = 0.082) (Fig. 4d). These findings indicate no significant differences between the measurements at various time points, confirming the stability of the Centiloid scale over time.

In our quantitative analysis, we utilized a Total Centiloid threshold of 15 for negative and uncertain results and 25 for uncertain and positive results. We observed changes in scan outcomes in 6 of 37 cases from 5 to 20 min. Five patients transitioned from uncertain to positive, while one changed from negative to uncertain at 10 min and subsequently to positive at 20 min. Specifically, one patient converted at 10 min, two at 15 min, and three at 20 min (Supp. Table 5).

The ROC curves (Fig. 5a–d) determined the optimal quantitative assessment of amyloid burden employing total Centiloid values across brain subregions at different scanning times. Among these, the Global Centiloid region consistently exhibited the highest AUC, indicating superior classification accuracy for identifying positive or negative PET scans. Specifically, the optimal threshold values for Global Centiloids did not significantly increase over time: 21.86 at 5 min post-injection (AUC = 0.985) (Fig. 5a), 24.70 at 10 min (AUC = 0.984) (Fig. 5b), 24.60 at 15 min (AUC = 0.985) (Fig. 5c), and 27.92 at 20 min (AUC = 0.986) (Fig. 5d).

Receiver operator characteristic (ROC) curves for different Centiloid regions and visual assessments at four scanning times. The ROC curves for correct subregions are shown for each duration: a 5 min, b 10 min, c 15 min, and d 20 min. The figure also displays the corresponding optimal threshold and AUC values for each subregion at different imaging durations. Overall, the Global Centiloid value has the best threshold across all four imaging durations

We employed ROC curves (Table 2) to assess the sensitivity and specificity of Centiloid measurements at 5, 10, 15, and 20 min. High AUC values at all time points, starting with 0.985 at the early 5 min mark and slightly increasing to 0.986 at 20 min, indicate that the test consistently maintains high sensitivity in accurately identifying positive cases. The specificity was also high at 5 min (0.972), comparable to that of the 20 min duration (1.000). The sensitivity at the 5 min imaging duration was particularly significant at 0.950, showing a marginally greater ability to detect early amyloid deposition than at 20 min (0.935).

Figure 6 depicts the statistical comparison of ROC curves to evaluate the classification performance across pairwise scanning times during Total Centiloid PET scans in the global region. No statistically significant differences were observed in diagnostic performance across these imaging durations (all p > 0.1). Specifically in Fig. 6a, p values were 0.489 for the 5 min versus 10 min comparison, 0.497 for 5 min versus 15 min in Fig. 6b, and 0.504 for 5 min versus 20 min in Fig. 6c, demonstrating consistent classification capability over time. For comparisons between later scanning times, the p-values were 0.539 for 10 min versus 15 min (Fig. 6d), 0.550 for 10 min versus 20 min (Fig. 6e), and 0.524 for 15 min versus 20 min (Fig. 6f), further confirming the stability and reliability of the Centiloid measure’s classification accuracy across these durations.

Statistical comparison of ROC curves for pairwise scanning times of global total Centiloid measures. The ROC curves represent the classification performance for pairwise scanning intervals of global total Centiloid measures, including their corresponding AUC value and p-value. The comparisons are as follows: a 5 and 10 min, b 5 and 15 min, c 5 and 20 min, d 10 and 15 min, e 10 and 20 min, and f 15 and 20 min. All p-values > 0.1 indicate no statistically significant differences observed between the scanning times for all classification pairs

Upon applying the 5 min optimal threshold of 21.86, six patients shifted from uncertain to positive amyloid status. However, one patient remained in the uncertain category despite this threshold. Figure 7 provides a detailed PET image representation of this outlier patient across diverse brain regions.

Representative PET images of an outlier case. The figure was obtained at 5 min, 10 min, 15 min, and 20 min post-injection in the global, frontal, lateral temporal, lateral parietal, and PCC-precuneus regions. The positive image is indicated with a red square, uncertain with a yellow box, and negative image with a white box. Each lobe was graded on a 0–3 scale for all slides using a semi-quantitative range of gray matter