Evaluation of SCALE PET Software for Quantitative Brain PET Analysis in Alzheimer's Disease

A Single-Center, Retrospective, Rater-Blinded, Crossover Study to Evaluate the Efficacy of the Medical Image Analysis Software 'SCALE PET' for Quantitative Analysis of Brain PET Images in Patients With Alzheimer's Disease

The goal of this study is to learn if medical image analysis software, Neurophet SCALE PET, can enhance readers' amyloid beta positivity performance in people with Alzheimer's Disease. The main question it aims to answer is:

Does the use of medical image analysis software assistance result in a statistically significant difference in readers' sensitivity and specificity for the classification of amyloid-β (Aβ) positivity?

Four readers will interpret amyloid PET images in two sessions (with and without software assistance), separated by a 4-week washout period, and will record Aβ positivity and diagnostic confidence using predefined rating scales.

1. Study Overview This study is a single-center, retrospective, rater-blinded, crossover reader study. The primary objective is to evaluate the efficacy of 'Neurophet SCALE PET', a medical image analysis software, in assisting clinicians with the quantitative analysis of amyloid-beta (Aβ) positivity on brain PET scans. The study will compare the performance of readers when they use the software (SW-assisted) versus when they do not use it (unassisted).

2. Data Collection and Reference Standard The investigators will retrospectively collect medical data from a total of 194 patients (97 Aβ-positive and 97 Aβ-negative cases).

   Data Sources: De-identified data will be obtained from Ajou University Hospital and the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.

3. Study Procedures (Reader Study) The reader study consists of two main reading sessions separated by a 4-week washout period to minimize memory bias and order effects. Four readers will be randomly assigned to two groups (Reader Group 1 and Reader Group 2).

   • Session 1: Reader Group 1 will evaluate Case Group A with SW assistance and Case Group B without assistance. Reader Group 2 will evaluate Case Group B with SW assistance and Case Group A without assistance.
   • Washout Period: A 4-week interval will be implemented between sessions.
   • Session 2: The conditions will be swapped following a crossover design. Readers will re-evaluate the same cases under the opposite condition (assisted vs. unassisted) compared to Session 1.

   During each session, readers will rate Aβ positivity on a 5-point scale (1: Definitely negative, 2: Probably negative, 3: Equivocal, 4: Probably positive, 5: Definitely positive) and record their diagnostic confidence level on a 3-point scale (1: low confidence, 2: moderate confidence, 3: high confidence).

4. Statistical Analysis The efficacy of the software will be assessed by comparing the diagnostic performance between the SW-assisted and unassisted conditions.

   • Primary outcomes: Differences in readers' sensitivity and specificity for Aβ positivity classification.
   • Secondary Outcomes: Differences in diagnostic accuracy (AUC), the number of cases rated as "equivocal," rater confidence levels, and inter-rater agreement.

Inclusion Criteria: Medical imaging datasets meeting all of the following criteria will be eligible for inclusion in this study.

1. Age ≥ 50 years at the time of brain magnetic resonance imaging (MRI)

2. Participants who underwent amyloid brain PET imaging using one of the following tracers: 18F-Florbetaben, 18F-Florbetapir, 18F-Flutemetamol

3. Availability of 3D T1-weighted brain MRI.

4. The interval between amyloid PET and 3D T1-weighted MRI is within 1 year.

5. Participants meeting one of the following predefined groups based on prior visual interpretation of amyloid PET:

   • Aβ-positive group: Prior clinical interpretation of amyloid PET reported as Aβ positive.
   • Aβ-negative group: Prior clinical interpretation of amyloid PET reported as Aβ negative.

Exclusion Criteria: Medical imaging datasets meeting any of the following criteria will be excluded from the study.

1. Poor image quality due to artifacts, motion, shading, or other technical issues that make visual interpretation difficult or prevent proper software processing.
2. Presence of prior brain surgery resulting in significant structural changes, or presence of intracranial metallic implants.
3. Contrast-enhanced 3D T1-weighted brain MRI.
4. Presence of structural brain lesions other than cerebral infarction.
5. Cases deemed inappropriate for study participation at the investigator's discretion.