Dual-Time-Point Diagnostic 18F-FDG PET/MRI with 3-hour Delayed, Prone, Breast-Coil Acquisition for Characterization of BI-RADS 4 Nodules: A Single-Centre Diagnostic Accuracy Study

Abstract Objective Conventional PET/MRI often misclassifies BI-RADS 4 breast nodules because of high parenchymal metabolic background, coil-related ADC distortion and missed low-grade cancers. To evaluate the diagnostic performance of a combined multi-sequence dual-phase modular scanning (MSDP-MS) dataset integrating 3-h delayed 18 F-FDG PET with prone dedicated-breast-coil MRI, compared with early-phase supine-coil PET/MRI in BI-RADS 4 breast nodules. Subjects and methods: 68 women with BI-RADS 4 breast nodules were enrolled and underwent PET/MRI examinations from July 2020 to February 2025. All examinations were performed on a GE SIGNA 3.0 T integrated PET/MRI system. Image post-processing was performed on an AW 4.6 workstation. After 18 F-FDG injection, patients rested in a dim quiet room. Whole-body supine PET/MRI was performed using a 16-channel torso coil. PET acquisition duration was 8 min and total whole-body scan duration was 40 min. At approximately 3 h post-injection, a delayed targeted acquisition was performed for suspected nodules and metabolically- active lesions. The patient was positioned prone, with breasts placed within an 8- channel dedicated breast coil. PET acquisition lasted 8 min and additional MRI sequences added approximately 12 min of table-time. Two breast radiologists independently interpreted lesion findings, with histopathology serving as the reference standard. Results Among the 68 lesions, 24 were malignant and 44 were benign. Compared with early-only conventional PET/MRI, the combined MSDP-MS dataset yielded accuracy 941 % vs 824 % (P = 0.025), sensitivity 917 % vs 750 % (P = 0.041), specificity 955 % vs 864 % (P = 0.108). Malignant lesions demonstrated lower ADCmin values irrespective of 18 F-FDG uptake trend. Exploratory optimal subgroup-specific ADCmin thresholds were 0.86 × 10⁻³ mm²/s (SUV-increase subgroup) and 1.31 × 10⁻³ mm²/s (SUV-decrease subgroup). The additional delayed-phase imaging component added approximately 12 min of scanner table-time and improved overall diagnostic accuracyby 11.8 %, raising sensitivityto 91.7 %. Conclusion The combined MSDP-MS multiparametric dataset showed promising diagnostic performance for differentiating benign from malignant BI-RADS 4 breast nodules.