The primary objective was to determine the diagnostic accuracy of PET imaging for detecting lymph node involvement in breast cancer patients undergoing upfront surgery. Secondary objectives included calculating likelihood ratios, evaluating the balance of sensitivity and precision through F1-score analysis, and assessing clinical utility metrics for surgical decision-making. Kappa coefficient to know the association between PET and HPE nodes.

This retrospective cohort study was conducted at the Department of Surgical Oncology, Yenepoya Medical College Hospital, a tertiary care teaching hospital in Mangaluru, Karnataka, India. Consecutive breast cancer patients undergoing upfront surgery between June 2022 and June 2025 who had preoperative PET/CT within 4 weeks of surgery and histopathological lymph node assessment were included. Written and informed consent taken from the participants of the study.

Inclusion criteria: 1.

Histologically confirmed invasive breast cancer,

Patients receiving neoadjuvant chemotherapy prior to surgery were excluded from this analysis to maintain treatment-pathway homogeneity.

The study cohort comprised 70 consecutive eligible patients ( Figure 1). Patient demographics, clinicopathological features, imaging characteristics, and surgical pathology findings were extracted from medical records. Axillary lymph node assessment is by either sentinel node, low axillary dissection, complete axillary dissection, or a combination of Allprocedures. Nodal metastasis was defined according to American Joint Cancer Committee (AJCC) criteria as either micrometastatic disease (tumor deposits >0.2 mm and ≤2 mm) or macrometastatic disease (tumor deposits >2 mm), and both categories were classified as node‑positive for this analysis. Isolated tumor cells (ITCs; single cells or clusters ≤0.2 mm) were not considered nodal metastases and were excluded from the study. All histopathological examinations were performed by experienced Oncopathologists using standard protocols.

No formal a priori sample size calculation was performed; the sample size was determined by the number of consecutive eligible upfront surgery patients with preoperative PET/CT during the study period.

2.2.1 Tracer type and dose

All patients underwent whole‑body 18F‑FDG PET/CT on a Discovery IQ Gen 2 model, manufactured by GE Healthcare. 18F‑FDG was administered intravenously at a dose of 0.11 mci/kg according to institutional protocol.

2.2.2 Patient preparation and uptake time

Patients fasted for at least 6 hours before tracer injection, and blood glucose was confirmed to be below 200 mg/dL prior to imaging. After injection, images were acquired following a standardized uptake period of 45 minutes with patients resting comfortably in a quiet room.

2.2.3 Imaging protocol and acquisition parameters

PET/CT was performed from skull base to mid thigh in supine position with arms elevated when feasible. A low dose CT scan was obtained for attenuation correction and anatomical localization using [100–120 kVp, 20–80 mAs, 0.8–1.5 pitch, 3–5 mm slice thickness]. PET data were acquired in [N] bed positions with an acquisition time of 1.3 to 4 minutes per patient, using [2D/3D] mode.

2.2.4 Reconstruction and corrections

PET images were reconstructed using [ordered subsets expectation maximization (OSEM)] with 2–4 iterations and 8–24 subsets, and a 3–6 mm Gaussian post filter. Standard corrections for attenuation (CT based), scatter, randoms, dead time, and decay were applied according to manufacturer recommendations.

2.2.5 Quantification and SUV definitions

Quantitative analysis was performed using standardized uptake values (SUVs). SUVmax was defined as the maximum voxel value within a region of interest (ROI) drawn over the primary breast lesion and suspicious lymph nodes on attenuation corrected PET images, normalized to patient body weight. For each nodal station, a volume of interest (VOI) was placed over the most metabolically active portion of the node, and nodal SUVmax was recorded.

2.2.6 Criteria for positivity, ROIs/VOIs

Lymph nodes were considered PET‑positive if they demonstrated focal FDG uptake above background blood pool or surrounding soft tissue, in conjunction with corresponding nodal enlargement or morphological abnormalities on CT, according to institutional criteria. ROIs/VOIs were drawn by a nuclear medicine physician with 10 years of experience, blinded to histopathology, and equivocal findings were resolved by consensus review.

Diagnostic accuracy metrics were calculated using standard formulas: •

Sensitivity = True Positive/(True Positive+False Negative) — probability of positive test given disease present

Diagnostic accuracy metrics were calculated using standard formulas for sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), likelihood ratios, accuracy, and Cohen’s kappa. The F1‑score and Youden’s index were additionally derived as summary measures of the balance between sensitivity and precision and overall discriminative ability, respectively, but were considered secondary to the primary diagnostic metrics for clinical interpretation.

Ninety-five percent confidence intervals (95% CI) for sensitivity and specificity were calculated using the Wilson score method. Disease prevalence was calculated as the proportion of patients with HPE-positive nodes in the cohort.

A confusion matrix was constructed to visualize classification accuracy (true positives(TP), true negatives(TN), false positives(FP), false negatives(FN)).

Data was analysed using SPSS v21. Kappa statistics was done for assessing the agreement between PET imaging and HPE nodal status.

The analysis included 70 breast cancer patients with mean age 53.9 ± 11.5 years undergoing upfront surgery. Luminal B predominates with 41.5% (n = 29) followed by Luminal A with 34.3% (n = 24), TNBC 17.1% (n = 12), Her-2 Enriched 7.1% (n = 5). Refer for Table 1 for demographic data. 50.8%(n = 36) have >20% ki-67 proliferation index. 53 (75.7%) were Estrogen receptor positive, whereas 47 (67.1%) are Progesterone receptor positive making it a hormone receptor positive predominant cohort.

Among the 70 surgical patients, 26 patients (37.14%) had histopathologically confirmed lymph node involvement. Forty-four patients (62.86%) had node-negative disease.

Detailed results are presented in Table 2 and Figure 2. PET imaging achieved: •

Sensitivity: 69.23% (95% CI: 48.27-85.67%) — correctly identified 18 of 26 patients with nodal disease

The positive likelihood ratio was 3.81 (95% CI: 1.75-8.31), indicating that a positive PET result is 3.81 times more likely in patients with actual lymph node disease compared to those without nodal involvement. This moderate LR+ suggests that positive findings warrant additional clinical correlation but provide meaningful evidence for nodal involvement.

The negative likelihood ratio was 0.38 (95% CI: 0.20-0.71), indicating that a negative PET result is substantially less likely in patients with actual nodal disease. This favorable LR- suggests that negative findings provide reasonable reassurance against nodal involvement in surgical planning.

True Positives (TP): 18 patients (25.71% of cohort) — correctly identified with nodes

False Positives (FP): 8 patients (11.43%) — incorrectly called positive without nodes

False Negatives (FN): 8 patients (11.43%) — missed nodal involvement

True Negatives (TN): 36 patients (51.43%) — correctly identified without nodes

The balanced false positive and false negative rates (both 11.43%) indicate comparable error distribution. The false negative rate of 30.77% (8 of 26 actual positive cases) represents the clinically meaningful limitation of PET sensitivity in this cohort.

The F1‑score was 0.6923, which numerically reflects the balance between sensitivity and PPV but is mainly reported as a composite summary statistic rather than for direct clinical decision‑making. The Youden’s index of 0.5105 is consistent with moderate overall discriminative ability of PET/CT for distinguishing node‑positive from node‑negative. In the upfront surgery cohort, agreement between PET/CT nodal status and histopathologic nodal status was moderate, with a Cohen’s kappa of 0.51 (SE 0.11; 95% CI 0.30–0.72; p = 0.001; n = 70).

Table 2 presents comprehensive diagnostic accuracy metrics. PET imaging demonstrated clinically meaningful performance for lymph node detection in upfront surgical patients, with strengths in negative predictive value (81.82%) and specificity (81.82%), indicating reliable ability to exclude nodal disease.

This study is among the few that specifically evaluate PET/CT diagnostic accuracy in a homogeneous cohort of breast cancer patients treated with upfront surgery. The findings reveal that PET imaging demonstrates clinically meaningful sensitivity (69.23%) coupled with high negative predictive value (81.82%) and specificity (81.82%) for lymph node detection in immediate surgical settings.

The 69.23% sensitivity indicates that PET successfully identifies approximately two-thirds of patients with actual nodal involvement. Among the 26 patients with pathologically confirmed nodal disease, PET correctly identified 18 cases. This performance is clinically relevant because it provides surgeons with meaningful information regarding nodal status before operative intervention.

Equally important, the 81.82% NPV indicates that when PET indicates nodal absence, there is reasonable confidence in this negative result for surgical planning purposes. Among 44 PET-negative patients, 36 (81.82%) genuinely lacked nodal disease, whereas only 8 (18.18%) had missed nodal involvement that was subsequently identified at histopathology.

The diagnostic profile of PET in upfront surgical patients differs meaningfully from neoadjuvant therapy settings, where treatment intensification depends primarily on positive node identification. In contrast, surgeons performing upfront surgery require confidence in nodal status to determine operative extent, staging accuracy, and prognosis discussion with patients.

Strengths of PET in This Context: 1.

High Specificity (81.82%) provides reliable identification of truly node-negative patients, avoiding unnecessary concern about occult nodal involvement and preventing unnecessary axillary dissection escalation.

Limitations Requiring Clinical Acknowledgment: 1.

False Negative Rate of 30.77% (8 of 26 actual positive cases) represents a clinically meaningful limitation. Approximately one-third of patients with actual nodal involvement were missed by PET. In practical terms, a false‑negative rate of 30.77% (8 of 26) means that reliance on PET/CT alone would result in a substantial proportion of patients with occult nodal disease being understaged. PET/CT should therefore be interpreted as an adjunct to, rather than a replacement for, standard axillary assessment, including clinical examination, ultrasound with image‑guided sampling where appropriate, and sentinel lymph node biopsy according to existing guidelines. Negative PET/CT findings can increase confidence in nodal negativity and may support operative planning, but they must not preclude routine sentinel node procedures or other protocol‑mandated staging steps in upfront surgical patients. Conversely, positive PET/CT findings should prompt careful correlation with conventional imaging and pathological confirmation when escalation of axillary surgery or systemic therapy is being considered. This false negative rate suggests that negative PET findings should not entirely exclude nodal assessment through standard staging approaches.

Limited prior studies have evaluated PET performance specifically for lymph nodal staging in upfront surgical populations. Most existing literature combines patients across treatment modalities or focuses on neoadjuvant settings. The sensitivity of 69.23% observed in the present cohort aligns with sensitivity ranges reported in mixed breast cancer populations (62-75%), ^{8, 9} though our cohort is uniquely focused on immediate surgical patients.

The specificity of 81.82% is comparable to or slightly lower than neoadjuvant cohorts in our institution’s data, possibly reflecting differences in disease characteristics and patient selection between upfront surgical and chemotherapy-naive neoadjuvant patients. ¹⁰ In the Study done by Kim et al. ¹¹ where he compared different imaging for lymph node metastasis elastography showed high sensitivity whereas PET/CT showed reasonable sensitivity.

Kasem et al ⁸ noted that combining PET/CT with FNA improved specificity, but PET/CT by itself had relatively low specificity for axillary disease (21% false positives) despite high sensitivity. Regional lymph nodes were incorrectly staged at 18F-FDG PET in 14% of cases in Estrogen receptor positive patients where as 18% of cases in our study which is slightly higher. ¹² Compared with the overall study population, estrogen receptor–positive patients demonstrated more accurate detection on PET imaging.

The superior specificity (81.82%) over sensitivity (69.23%) in upfront surgical patients may reflect multiple factors: 1.

Disease burden differences: Patients selected for upfront surgery may have different patterns of nodal involvement compared to neoadjuvant patients, potentially affecting detection patterns.

Based on these findings, we propose the following framework for incorporating PET findings into preoperative assessment of surgical patients:

When PET is POSITIVE (n = 26 in this cohort): •

Likelihood of true nodal involvement is 69.23% (PPV).

When PET is NEGATIVE (n = 44 in this cohort): •

Likelihood of true nodal absence is 81.82% (NPV)

Strengths: •

Homogeneous cohort of upfront surgical patients (treatment-pathway specific)

Limitations: •

Retrospective design with inherent selection bias