Patient characteristics

From January 2014 to October 2022, a total of 83 patients were included in the study (38 in PET/CT cohort, and 45 in PET/MR cohort). The median intervals (25th-75th) from the completion of the last chemotherapy to PET/CT and PET/MR scans were 27 days (18–45 days) and 25 days (14–50 days), respectively.

The baseline clinical characteristics of these patients are outlined in Table 1. All patients were high-grade B-cell lymphoma, the majority of which were diffuse large B-cell lymphoma. In the PET/CT cohort, 20 patients (53%) were male, with a median age of 57 years (range 22–74). At the time of diagnosis, multiple intracranial tumor lesions were detected in 21 patients (55%), while the remaining 17 patients (45%) presented with a single lesion. In the PET/MR cohort, 24 patients (53%) were male, with a median age of 63 years (range 15–83). Multiple lesions were identified in 29 patients (64%) at diagnosis, while 16 patients (36%) had a single lesion. The median follow-up was 34.6 months in the PET/CT cohort, and the median PFS and OS were 31.2 and 38.6 months, respectively. In the PET/MR cohort, the median PFS and OS were 20.8 and 21.4 months, respectively, with a median follow-up 21.4 months. During the follow-up period, 12 patients (32%) in the PET/CT cohort experienced disease progression or relapse, and 7 patients (18%) died. In the PET/MR cohort, 8 patients (18%) experienced disease progression or relapse, and 3 patients (7%) died.

Table 1 Patient characteristicsIMS and/or DS in the PET/CT and PET/MR cohorts

Figure 1b shows the SUVmax of the reference normal structures for both the IMS and DS. The SUVmax of gray matter, with a mean ± standard deviation (SD) of 9.45 ± 2.01, was significantly greater than that of the other reference structures. The second tier included the liver (2.96 ± 0.42) and white matter (2.81 ± 0.58), and no significant difference in the SUVmax was observed (p = 0.178). The third tier comprised the mediastinal blood pool (1.80 ± 0.36) and cerebrospinal fluid (CSF) (1.50 ± 0.34). Although the SUVmax of the mediastinal blood pool was significantly greater than that of the cerebrospinal fluid (p = 0.001), the difference in values was minimal (1.80 vs. 1.50).

Within the cohort of 38 patients undergoing EOT-PCTs, the distribution of the IMS was as follows: 5 patients (13%) were assigned IMS 5, 5 patients (13%) IMS 4, 11 patients (29%) IMS 3, 9 patients (24%) IMS 2, and 8 patients (21%) IMS 1. Meanwhile, the DS was distributed as follows: 5 patients (13%) were assigned DS 5, no patients were assigned DS 4, 5 patients (13%) DS 3, 2 patients (5%) DS 2, and 26 patients (68%) DS 1.

For the 45 patients in the PET/MR cohort, the DS was inapplicable due to the absence of the mediastinal blood pool and liver. The IMS was distributed as follows: 5 patients (11%) were assigned to IMS 5, 9 patients (20%) to IMS 4, 12 patients (27%) to IMS 3, 9 patients (20%) to IMS 2, and 10 patients (22%) to IMS 1.

The kappa co-efficient for labeling the response assessment score on IMS and DS criteria were 0.910 and 0.958, respectively.

Survival analysis of the PET/CT cohort

Figure 2 and Supplementary Fig. 1 showed the Kaplan‒Meier survival curves for PFS and OS in the PET/CT cohort. After stratifying the patients into five distinct groups based on the IMS 1–5, significant disparities were observed among the groups in terms of PFS (p < 0.001) and OS (p < 0.001) survival rates. Further dichotomization of the patients according to the IMS into two groups (IMS 1–3 versus IMS 4–5 and IMS 1–4 versus IMS 5) revealed that the IMS 4–5 was associated with a markedly poorer prognosis than was the IMS 1–3 (PFS, HR 9.04, 95% CI 2.08–39.29, p < 0.001; OS, HR 23.01, 95% CI 3.69-143.31, p < 0.001), and the IMS 5 was linked to a significantly worse outcome than was the IMS 1–4 (PFS, HR 9.50, 95% CI 1.03–87.43, p < 0.001; OS, HR 15.09, 95% CI 1.01-227.47, p < 0.001). When the cohort was further divided into three groups (IMS 1–2, IMS 3–4 and IMS 5), it was found that an IMS 5 was associated with worse outcomes than both an IMS 1–2 and an IMS 3–4 (PFS, p < 0.001; OS, p < 0.001).

Fig. 2

Survival analysis of PFS and OS in the PET/CT cohort according to the IMS and DS scales

Significant differences in PFS and OS were observed among patients stratified into five groups based on the DS (p < 0.001; Supplementary Fig. 1). When patients were dichotomized using the DS (DS 1–3 versus DS 4–5 and DS 1–4 versus DS 5), DS 4–5 was associated with significantly worse outcomes than those with DS 1–3 (PFS, HR 9.50, 95% CI 1.03–87.43, p < 0.001); OS, HR 15.09, 95% CI 1.01-227.47, p < 0.001); and DS 5 was linked to significantly inferior outcomes than those with DS 1–4 (PFS, HR 9.50, 95% CI 1.03–87.43, p < 0.001; OS, HR 15.09, 95% CI 1.01-227.47, p < 0.001). When patients were divided into three groups (DS 1–2, DS 3–4 and DS 5), DS 5 was associated with worse PFS (p < 0.001) and OS (p < 0.001) than those with both DS 1–2 and DS 3–4.

As demonstrated in Table 2, Univariate and Multivariate Cox regression models were performed to evaluate the prognostic significance of clinical characteristics on PFS and OS, including age, gender, lesion number, site of tumor lesion, ECOG score, IELSG score, MSKCC score and PET results (DS and IMS). Multivariate analysis demonstrated the IMS (HR 9.04, 95% CI 2.08–39.29; p < 0.001) and DS (HR 9.50, 95% CI 1.03–87.43; p < 0.001) were the independent prognostic indicators for PFS in the PET/CT cohort. Similarly, for OS, the IMS (HR 18.34, 95% CI 2.15-156.67; p = 0.008) and DS (HR 15.18, 95% CI 2.68–85.92; p = 0.002) were identified as independent prognostic indicators in the PET/CT cohort (Table 2).

Table 2 Univariate and multivariate analysis for PFS and OS in PET/CT cohort of PCNSL patientsSurvival analysis of the PET/MR cohort

Figure 3 and Supplementary Fig. 2 shows the Kaplan‒Meier survival curves for PFS and OS in the PET/MR cohort. After stratifying the patients into five distinct groups based on the IMS 1–5, significant disparities were observed among the groups in terms of PFS (p < 0.001) survival. However, there was no significant difference between the 5 groups in OS (p = 0.440). Further dichotomization of the patients according to the IMS into two groups (IMS 1–3 versus IMS 4–5 and IMS 1–4 versus IMS 5) revealed that the IMS 4–5 was associated with markedly poorer PFS than was the IMS 1–3 (HR 15.49, 95% CI 3.47–69.22, p < 0.001), and the IMS 5 was linked to significantly worse PFS than was the IMS 1–4 (HR 35.47, 95% CI 1.25-1007.71, p < 0.001). However, there was no difference in OS between IMS 1–3 and IM 4–5 (HR 4.16, 95%CI 0.37–46.59, p = 0.204), or between IMS 1–4 and IMS 5 (HR 5.62, 95%CI 0.09-350.03, p = 0.109). When patients were categorized into three groups (IMS 1–2, IMS 3–4, and IMS 5), it was found that an IMS 5 was associated with shorter PFS than both an IMS 1–2 and an IMS 3–4 (p < 0.001). However, there was no difference in OS between these 3 groups (p = 0.154).

Fig. 3

Survival analysis of PFS and OS in the PET/MR cohort according to the IMS scale

Similarly, univariate and Multivariate Cox regression models were also performed to evaluate the prognostic significance of clinical characteristics on PFS and OS, including age, gender, lesion number, site of tumor lesion, ECOG score, IELSG score, MSKCC score and IMS as showed in Table 3. Multivariate analysis revealed that the IMS (HR 15.49, 95% CI 3.47–69.22; p < 0.001) was an independent prognostic indicator for PFS in the PET/MR cohort. For OS, no independent prognostic indicator was identified in the PET/MR cohort (Table 3).

Table 3 Univariate and multivariate analysis for PFS and OS in PET/MR cohort of PCNSL patientsPredictive values of PET/CT and PET/MR according to assessment methods

ROC curves were constructed to evaluate the 5-year PFS and OS in patients from both the PET/CT and PET/MR cohorts, as depicted in Fig. 4. Within the PET/CT cohort, the AUC for the ability of the IMS to predict 5-year PFS in PCNSL patients was 0.929 (95% CI, 0.797–0.987; p < 0.001), surpassing the AUC of the DS scale of 0.763 (95% CI, 0.597–0.885; p <0.001). Similarly, the AUC for the IMS in forecasting 5-year OS was 0.926 (95% CI, 0.793–0.986; p < 0.001), which exceeded the AUC of the DS scale (0.866; 95% CI, 0.717–0.955; p = 0.003). In the PET/MR cohort, the AUCs for the IMS for predicting 5-year PFS and OS were 0.934 (95% CI, 0.818–0.986; p < 0.001) and 0.782 (95% CI, 0.634–0.891; p = 0.106), respectively.

Fig. 4

ROC curves of the IMS and DS in the PET/CT and PET/MR cohorts

In the PET/CT cohort, 33 (87%) patients were classified as negative category and 5 (13%) patients as positive if DS 4 and 5 were considered as positive. Given that DS 4 and 5 were considered as positive, the sensitivity, specificity, PPV, and NPV of EOT-PCT for PFS were 42%, 100%, 100% and 79%, respectively and for OS were 57%, 97%, 80% and 91%, respectively (Table 4). 10 (26%) patients were classified as positive category if IMS 4 and 5 were considered as positive, and 5 (13%) patients were classified as positive category if IMS 5 were considered as positive. Given that IMS 4 and 5 were considered as positive, the sensitivity, specificity, PPV, and NPV of EOT-PCT for PFS were 67%, 92%, 80% and 86%, respectively and for OS were 86%, 87%, 60% and 96%, respectively. Given that IMS 5 was considered as positive, the sensitivity, specificity, PPV, NPV of EOT-PCT for PFS were 42%, 100%, 100% and 79%, respectively and for OS were 57%, 97%, 80% and 91% (Table 4).

Table 4 Sensitivity, specificity, predictive values and AUC for outcomes

In the PET/MR cohort, 14 (37%) patients were classified as positive category if IMS 4 and 5 were considered as positive, and 5 (13%) patients were classified as positive category if IMS 5 were considered as positive. Given that IMS 4 and 5 were considered as positive, the sensitivity, specificity, PPV, and NPV of EOT-PMR for PFS were 88%, 81%, 50% and 97%, respectively and for OS were 67%, 71%, 14% and 97%, respectively. Given that IMS 5 was considered as positive, the sensitivity, specificity, PPV, NPV of EOT PET for PFS were 63%, 100%, 100% and 92%, respectively and for OS were 33%, 91%, 20% and 93% (Table 4).