Quantitative optimisation of indocyanine green fluorescence for parathyroid imaging in robotic BABA thyroidectomy: a prospective cohort study

To develop and validate a quantitative framework for intraoperative indocyanine green (ICG) fluorescence imaging of the parathyroid glands during robotic bilateral axillo-breast approach (BABA) thyroidectomy and to identify technical and clinical factors associated with optimal image quality. Background ICG fluorescence is increasingly used to aid parathyroid identification and perfusion assessment in thyroid surgery, but there is no consensus on optimal dosing, image quantification, or patient selection, particularly in the context of robotic thyroidectomy. A robust, objective definition of “good” imaging is needed to standardise protocols and support broader clinical adoption. Methods In this single-centre prospective study (June 2024–January 2025), 72 consecutive patients with low-risk papillary thyroid carcinoma underwent robotic BABA thyroidectomy with ICG fluorescence using the da Vinci Xi system. A cyclic dosing protocol (1.5, 2.0, 2.5, 3.0 mL) was applied in eligible cases. For each patient, a 12-s fluorescence sequence was recorded from the onset of visible ICG uptake, and 12 frames (1 frame per second) were extracted. In each frame, the parathyroid region (Zone A) and periparathyroid background (Zone C) were manually segmented, and the relative fluorescence intensity ratio (RFIR = A/C) was calculated. Image quality was independently graded on a 3-point scale (0–2) and classified as clear (mean score ≥ 1) or unclear (< 1). Receiver operating characteristic analysis was used to derive an RFIR threshold for “good” imaging. Univariable and multivariable logistic regression were performed to explore associations between imaging quality and clinical variables (demographics, biochemistry, thyroiditis, fasting blood glucose [FBG], intraoperative haemodynamics) and ICG dose. Results An RFIR cut-off of 1.07 yielded an area under the curve of 0.832 for discrimination between clear and unclear images. At the case level, imaging was defined as good when at least 6 of 12 frames (≥ 50%) had RFIR ≥ 1.07. In univariable analysis, higher body mass index was positively associated with good imaging (odds ratio [OR] 1.18, 95% confidence interval [CI] 1.03–1.36; P = 0.020), whereas an ICG dose of 3.0 mL was negatively associated (OR 0.11, 95% CI 0.02–0.50; P = 0.004). In multivariable analysis, elevated FBG (OR 0.25, 95% CI 0.08–0.86; P = 0.027) and an ICG dose of 3.0 mL (OR 0.01, 95% CI 0.00–0.17; P = 0.001) remained independently associated with reduced odds of good imaging, while other clinical and haemodynamic variables were not significant. Interaction analyses showed that low FBG combined with lower doses (1.5–2.0 mL) provided the highest predicted probability of optimal imaging, whereas high FBG with higher doses (2.5–3.0 mL) yielded the poorest performance. Conclusions ICG fluorescence imaging of the parathyroid glands during robotic BABA thyroidectomy can be standardised using an RFIR threshold of 1.07 and a frame-based majority rule to define good imaging. Lower ICG doses (1.5–2.0 mL) and favourable metabolic status (low FBG) are key determinants of optimal image quality, whereas higher doses (3.0 mL) and elevated FBG are associated with over-saturation and reduced contrast. These findings support tailored, low-dose ICG protocols and provide a quantitative benchmark for future multi-centre studies linking fluorescence metrics to postoperative calcium and parathyroid function outcomes.