Image-derived input functions for [18F]LW223 and [18F]SynVesT-1 PET in the rodent determined using an autoencoder...

Area of Science:

• Nuclear medicine
• Machine learning applications
• Quantitative imaging

Background:

• Dynamic positron emission tomography (PET) requires accurate arterial input function (AIF) for quantitative analysis.
• Current AIF methods, arterial blood sampling and conventional image-derived input function (IDIF), have limitations including invasiveness, cost, and significant errors.
• Developing a non-invasive, accurate, and data-efficient AIF method is crucial for advancing quantitative PET imaging.

Purpose of the Study:

• To develop and validate a novel machine learning-based image-derived arterial input function (IDIF) method for dynamic PET scans.
• To assess the performance of the proposed IDIF method compared to conventional techniques in determining kinetic parameters.
• To demonstrate the generalizability of the method across different animal models and PET tracers.

Main Methods:

• A customized autoencoder (AE) machine learning model was trained using voxel-by-voxel time-activity curves from dynamic PET scans.
• The autoencoder learns a compressed representation of the data, enabling the extraction of the IDIF from the model's weights.
• The method was evaluated on rat PET scans using [18F]LW223 tracer and adapted for mouse scans with [18F]SynVesT-1 tracer.

Main Results:

• The novel IDIF method, termed IDIF-AE, demonstrated superior accuracy in estimating the volume of distribution (VT) compared to conventional IDIF.
• IDIF-AE achieved a mean absolute percentage error of ±32% for VT, significantly outperforming the conventional IDIF's ±54%.
• The method was successfully applied to different animal models (rats and mice) and neuro PET tracers, showcasing its adaptability.

Conclusions:

• The developed machine learning-based IDIF method offers a promising alternative for accurate AIF estimation in dynamic PET.
• This approach overcomes the limitations of traditional methods by providing accurate kinetic parameter quantification without requiring external training data.
• The IDIF-AE method has the potential to enhance the reliability and efficiency of quantitative PET imaging in research and clinical settings.