Demographic and clinical data

In total, 12 PCA patients, 19 AD patients, and 30 HC were included in the study. There was no significant difference between the three groups in demographic data, including age, sex proportion, and years of education, as well as routine optical examination results, such as sphere, cylinder, axial, and intraocular pressure (all p > 0.05). The PCA and AD groups had lower MMSE and MoCA scores than the HC group (all p < 0.001), and the PCA group displayed a lower MoCA score than the AD group (8.75 ± 4.48 vs 13.21 ± 5.79, p < 0.05). CDR scores showed no difference between the AD and PCA groups (1.79 ± 0.86 vs 2.00 ± 0.60, p = 0.655) (Table 1).

Table 1 Demographical and clinical data of participants

Following evaluation by two ophthalmologists, eyes with low image quality and/or ocular diseases that met the exclusion criteria were excluded. Finally, 50 eyes of HC (oculus dexter (OD) n = 25, oculus sinister (OS) n = 25), 30 eyes of AD (OD n = 16, OS n = 14), and 20 eyes of PCA patients (OD n = 9, OS n = 11) were available in the FAZ analysis; while 49 eyes of HC (OD n = 26, OS n = 23), 34 eyes of AD (OD n = 17, OS n = 17) and 18 eyes of PCA patients (OD n = 8, OS n = 10) were available for the SS-OCT and OCTA assessments. Because of high eccentricity in most eyes, we aborted the data of the peripheral circle (i.e., regions from 3 to 6 mm in the ETDRS grid) in the subsequent statistical analysis to assure the accuracy of the results. All optical parameters were compared between three groups in OD separately, OS separately, and OD mixing OS.

Fundus structural differences assessed by SS-OCT

The thickness of six separate or combined layers of the fundus were compared between the three groups in seven regions (0–1 mm circle, 0–3 mm circle, 1–3 mm ring, superior inner (SI), temporal inner (TI), inferior inner (II), nasal inner (NI) (Fig. 1A)).

Choroidal thickness was greater in the PCA group than in the AD group in the 0–1 mm circle (353.67 ± 89.50 vs 289.91 ± 89.59 μm, p = 0.040) and TI (349.86 ± 85.13 vs 285.11 ± 83.72 μm, p = 0.024). No differences were observed between the HC group and the PCA or AD groups (both p > 0.05) (Supplementary Table 1).

Retinal thickness was less in the PCA group than in the HC group in the 0–3 mm circle (315.97 ± 12.32 vs 321.29 ± 13.32 μm, p = 0.021) and NI (326.32 ± 14.63 vs 335.39 ± 14.25 μm, p = 0.003) (Supplementary Table 2). GCL + IPL thickness was less in the PCA group than in the HC group in the 0–3 mm circle (76.91 ± 9.17 vs 79.56 ± 5.62 μm, p = 0.003), 1–3 mm ring (83.69 ± 9.72 vs 86.54 ± 5.90 μm, p = 0.008), SI (85.02 ± 8.31 vs 88.07 ± 5.96 μm, p = 0.018), and NI (83.47 ± 12.13 vs 87.69 ± 6.34 μm, p = 0.0201) (Supplementary Table 3). INL thickness was less in the PCA group than in the HC group in NI (45.38 ± 2.84 vs 47.07 ± 4.06 μm, p = 0.035) (Supplementary Table 4). RNFL thickness was less in the PCA group than in the HC group in the 0–3 mm circle (23.86 ± 3.40 vs 24.66 ± 2.21 μm, p = 0.034), 1–3 mm ring (25.21 ± 3.77 vs 26.06 ± 2.77 μm, p = 0.041), and NI (23.68 ± 3.77 vs 25.03 ± 2.86 μm, p = 0.008) (Supplementary Table 5). No difference was observed in the GCL + IPL + RNFL thickness (p > 0.05) (Supplementary Table 6). The AD group presented no significant difference with the HC and PCA groups in all five structural thicknesses (all p > 0.05).

Fundus capillary differences assessed by SS-OCTA

Capillary density and flow area of the SCP, ICP, and DCP, six outcomes in total, were compared between the three groups in the same seven regions (0–1 mm circle, 0–3 mm circle, 1–3 mm ring, SI, TI, II, NI).

For the capillary density of the SCP, this was lower in the AD group than in the HC group in the 0–3 mm circle (p = 0.001), 1–3 mm ring (p = 0.001), SI (p = 0.002), TI (p = 0.019), II (p = 0.009), and NI (p = 0.007); while it was lower in the PCA group than in the HC group in the 0–1 mm circle (p = 0.008), 0–3 mm circle (p = 0.024), 1–3 mm ring (p = 0.048), and NI (p = 0.020) (Table 2 and Fig. 2). For the capillary density of the ICP, this was lower in the AD group than in the HC group in the 0–3 mm circle (p = 0.020), 1–3 mm ring (p = 0.008), TI (p = 0.046), and II (p = 0.013); while it was lower in the PCA group than in the AD group in the 0–1 mm circle (p = 0.044) (Table 2 and Fig. 2). There was no significant difference in the capillary density of the DCP between the three groups in any quadrant (all p > 0.05) (Supplementary Table 7).

Table 2 Comparison of flow area and vascular density of ICP and SCP in three groupsFig. 2

Heat maps showing the difference of mean level of flow area and capillary density. A Flow area (mm2) in the superficial capillary plexus (SCP) (left) and the intermediate capillary plexus (ICP) (right). B Capillary density (%) in the SCP (left) and ICP (right). AD, PCA vs HC: *p < 0.05, **p < 0.01, ***p < 0.001

Regarding the flow area of the SCP, this was smaller in the AD group than in the HC group in the 0–3 mm circle (p < 0.001), 1–3 mm ring (p < 0.001), SI (p = 0.001), TI (p = 0.005), II (p = 0.004), and NI (p = 0.001); while this was smaller in the PCA group than in the AD group in the 0–1 mm circle (p = 0.030) and smaller than in the HC group in the 0–1 mm circle (p = 0.004), 0–3 mm circle (p = 0.009), 1–3 mm ring (p = 0.024), SI (p = 0.045), II (p = 0.040), and NI (p = 0.009) (Table 2 and Fig. 2). The flow area of the ICP was smaller in the AD group than in the HC group in the 0–3 mm circle (p = 0.002), 1–3 mm ring (p = 0.002), SI (p = 0.010), TI (p = 0.007), II (p = 0.007), and NI (p = 0.021); while it was smaller in the PCA group than in the HC group in the 0–1 mm circle (p = 0.009) and 0–3 mm circle (p = 0.027) (Table 2 and Fig. 2). For the flow area of the DCP, only that of the TI area of the AD group was significantly smaller than in the HC group (0.3263 ± 0.0867 vs 0.3685 ± 0.0537 mm2, p = 0.011) (Supplementary Table 8).

FAZ assessment

There was no significant difference between the PCA, AD, and HC groups in the FAZ area (0.3948 ± 0.1239 vs 0.3995 ± 0.1278 vs 0.4144 ± 0.1094 mm2, p = 0.769), perimeter (2.622 ± 0.4205 vs 2.684 ± 0.3852 vs 2.718 ± 0.3679 mm, p = 0.641), circularity index (0.7064 ± 0.0578 vs 0.6805 ± 0.0779 vs 0.6969 ± 0.0736, p = 0.424), or fractal dimension (40.50 ± 4.60 vs 39.54 ± 4.34 vs 41.49 ± 3.44, p = 0.106) (Supplementary Table 9).

Correlation analysis of optic outcomes and cognitive state

To verify the matching of different OCTA data in one region, the correlation analysis between the capillary density and flow area in the ICP and SCP was conducted, showing a high correlation in all seven quadrants (r ranged from 0.8359 to 0.8707 for the ICP and from 0.9110 to 0.9723 for the SCP, all p < 0.0001).

On the basis of the analysis results above, indicators that differed significantly between groups were selected for further correlation analysis. The capillary density and flow area of the SCP was of weak to moderate correlation with the RNFL thickness, and of moderate to high correlation with the GCL + IPL in all seven quadrants (Table 3), whereas the capillary density and flow area of the ICP only displayed a strong correlation with the GCL + IPL in the 0–1 mm circle (r = 0.6439, r = 0.7848, respectively, both p < 0.0001) and a moderate to strong correlation with the INL in the 0–1 mm circle (r = 0.4645, r = 0.6588, respectively, both p < 0.0001).

Table 3 Correlation between capillary density, flow area of SCP and thickness of RNFL, GCL + IPL

A weak correlation of vessel density and perfusion area with the MMSE and MoCA scores was found in some quadrants of the SCP and ICP in all participants (Table 4). For subdomains in MMSE, only 0–3 mm circle, 1–3 mm ring, and II areas of ICP showed a weak correlation with visuospatial ability (Supplementary Table 10).

Table 4 Correlation between capillary density, flow area of SCP and ICP with MMSE and MoCA scoreROC curves of logistic regression models in distinguishing PCA, AD, and HC

For efficiency of every single optical parameter, to differentiate AD patients from HC, both vascular density and flow area in the SCP and ICP displayed low to moderate diagnostic value separately in almost all quadrants. The flow area performed slightly better than the vascular density, with the AUC of the former ranging from 0.6714 to 0.7161 in the SCP (p ranged from 0.0009 to 0.0082) and 0.6468 to 0.7203 in the ICP (p ranged from 0.0007 to 0.0236), while the AUC of the latter ranged from 0.6483 to 0.6987 in the SCP (p ranged from 0.0026 to 0.0222) and 0.6399 to 0.6897 in the ICP (p ranged from 0.0034 to 0.0310) (Fig. 3). Neither vascular density nor flow area in the 0–1 mm circle showed any capability for differentiation.

Fig. 3

ROC curve of vessel density and flow area to distinguish AD from HC. A. Vessel density of the SCP; B. Flow area of the SCP; C. Vessel density of the ICP; and D. Flow area of the ICP

Although the PCA group displayed a significant difference from the HC group in flow area and vessel density of the SCP, only the flow area of five quadrants in the ICP showed a low to moderate efficiency in distinguishing PCA patients from HC: 0–3 mm circle (AUC = 0.7132, p = 0.0078), 1–3 mm ring (AUC = 0.6950, p = 0.0150), SI (AUC = 0.6933, p = 0.0159), II (AUC = 0.7086, p = 0.0092), and NI (AUC = 0.6661, p = 0.0382) (Supplementary Fig. 1). No individual indicator was able to distinguish between the AD and PCA groups.

For multivariable diagnostic models, confounding demographic data as age, sex proportion and years of education was used as independent factors in all models. MMSE performed better than any combination of optical parameters in identifying AD and PCA from HC (AUC = 1, p < 0.001). Flow area of SCP in 1–3 mm ring was the best optical predictor for AD out of HC (AUC = 0.768, p < 0.001), as flow area of SCP in 0–1 mm circle and 1–3 mm ring was the best for PCA out of HC (AUC = 0.898, p < 0.001). In addition, PCA could be distinguished from AD using the optimal combination of MoCA, retinal thickness and vascular density of ICP in the 1–3 mm ring, with flow area of ICP in the 0–1 mm circle (AUC = 0.944, p < 0.001) (Fig. 4).

Fig. 4

ROC curve of multivariate Logistic regression models of identifying PCA from AD. Confounding demographic data as age, sex proportion and years of education was used as independent factors in model 1 (null model). MoCA scores were added as clinical data into the null model to get model 2. Retinal thickness and vascular density of ICP in the 1-3 mm ring, with flow area of ICP in the 0-1 mm circle were added as optical parameters into the model 2 to get model 3