Introduction Visual amyloid positron emission tomography (PET) interpretation is standard for detecting cerebral amyloid-β but is limited by subjectivity, inter-reader variability, and tracer heterogeneity. We evaluated the impact of adjunctive quantitative analysis in real-world clinical practice. Materials and methods This retrospective multicentre study included patients who underwent amyloid PET with F-18 florbetaben (FBB), flutemetamol (FMM), or florapronol (FPN) and T1-weighted brain MRI between 2015 and 2022. Scans were interpreted by visual assessment alone and then with adjunctive quantitative standardized uptake value ratio and Centiloid analysis. Primary outcomes were changes in reader confidence and inter-reader agreement; agreement metrics with an expert consensus adjudication (with quantitative and clinical information unblinded only when initial visual reads were discordant) were assessed as a secondary endpoint. Results In 950 patients, adjunctive quantitative analysis significantly increased reader confidence, with high-confidence interpretations rising from 76.3% to 83.7% (p < 0.001). Inter-reader agreement also improved significantly (κ: 0.746 to 0.826, p < 0.001, respectively). Accuracy for the expert consensus adjudication increased modestly (0.930 to 0.940, p = 0.003), driven by higher specificity (0.915 to 0.944, p < 0.001) with a slight reduction in sensitivity (0.943 to 0.937, p = 0.52). Conclusion Adjunctive quantitative analysis increases reader confidence and improves inter-reader agreement in real-world amyloid PET interpretation. Although agreement and diagnostic metrics with expert consensus adjudication also increased, this should not be interpreted as independent diagnostic validation. Overall, quantitative metrics appear most valuable as an adjunct to enhance consistency, particularly in equivocal cases. Key Points <Emphasis Type="BoldItalic">Question</Emphasis> Does the use of adjunctive quantitative analysis significantly increase the diagnostic confidence of readers compared to visual interpretation alone? <Emphasis Type="BoldItalic">Findings</Emphasis> Adjunctive quantitative support significantly increased reader confidence and inter-reader agreement. <Emphasis Type="BoldItalic">Clinical relevance</Emphasis> Quantitative tools, when used as adjuncts to expert visual interpretation and supported by automated workflows, can improve consistency and confidence in amyloid PET reading, particularly in borderline cases, thereby supporting clinical decision-making in routine practice. Graphical Abstract

Impact of adjunctive quantitative analysis on visual interpretation of amyloid PET: a multiple tracer, multicentre study

期刊论文
发表日期：

2026年6月18日

查看全文

Seunggyun Ha,
Byungwook Choi,
Soo Jin Kwon,
Sungmin Kang,
Junik Son,
Shin Young Jeong,
Sang-Woo Lee,
Yeong Sim Choe,
Donghyeon Kim,
Chae Moon Hong

摘要

Introduction

Visual amyloid positron emission tomography (PET) interpretation is standard for detecting cerebral amyloid-β but is limited by subjectivity, inter-reader variability, and tracer heterogeneity. We evaluated the impact of adjunctive quantitative analysis in real-world clinical practice.

Materials and methods

This retrospective multicentre study included patients who underwent amyloid PET with F-18 florbetaben (FBB), flutemetamol (FMM), or florapronol (FPN) and T1-weighted brain MRI between 2015 and 2022. Scans were interpreted by visual assessment alone and then with adjunctive quantitative standardized uptake value ratio and Centiloid analysis. Primary outcomes were changes in reader confidence and inter-reader agreement; agreement metrics with an expert consensus adjudication (with quantitative and clinical information unblinded only when initial visual reads were discordant) were assessed as a secondary endpoint.

Results

In 950 patients, adjunctive quantitative analysis significantly increased reader confidence, with high-confidence interpretations rising from 76.3% to 83.7% (p < 0.001). Inter-reader agreement also improved significantly (κ: 0.746 to 0.826, p < 0.001, respectively). Accuracy for the expert consensus adjudication increased modestly (0.930 to 0.940, p = 0.003), driven by higher specificity (0.915 to 0.944, p < 0.001) with a slight reduction in sensitivity (0.943 to 0.937, p = 0.52).

Conclusion

Adjunctive quantitative analysis increases reader confidence and improves inter-reader agreement in real-world amyloid PET interpretation. Although agreement and diagnostic metrics with expert consensus adjudication also increased, this should not be interpreted as independent diagnostic validation. Overall, quantitative metrics appear most valuable as an adjunct to enhance consistency, particularly in equivocal cases.

Key Points

Question Does the use of adjunctive quantitative analysis significantly increase the diagnostic confidence of readers compared to visual interpretation alone?

Findings Adjunctive quantitative support significantly increased reader confidence and inter-reader agreement.

Clinical relevance Quantitative tools, when used as adjuncts to expert visual interpretation and supported by automated workflows, can improve consistency and confidence in amyloid PET reading, particularly in borderline cases, thereby supporting clinical decision-making in routine practice.

Graphical Abstract