Development and in silico imaging trial evaluation of a deep-learning-based transmission-less attenuation...

Quantitative measures of dopamine transporter (DaT) uptake in the caudate, putamen, and globus pallidus (GP) derived from DaT-single-photon emission computed tomography (SPECT) images are being investigated as biomarkers to diagnose, assess disease status, and track the progression of Parkinsonism. Reliable quantification from DaT-SPECT images requires performing attenuation compensation (AC), typically with a separate x-ray CT scan. Such CT-based AC (CTAC) has multiple challenges, a key one being the non-availability of x-ray CT components on many clinical SPECT systems. Even when a CT is available, the additional CT scan leads to increased radiation dose, costs, and complexity; potential quantification errors due to SPECT-CT misalignment; and higher training and regulatory requirements.

To overcome the challenges with the requirement of a CT scan for AC in DaT SPECT, we develop a transmission-less AC method for DaT SPECT and validate the method in a clinically realistic setting using an in silico imaging trial.

Integrating concepts from physics and deep learning (DL), we propose a DL-based transmission-less AC method for DaT-SPECT (DaT-CTLESS). In this method, an initial attenuation map reconstructed from scatter-energy window projection is segmented into different regions using a U-net-based network trained on CT scans. Each region is assigned a predefined attenuation coefficient, yielding an attenuation map for AC. An in silico imaging trial, titled ISIT-DaT, was designed to evaluate the performance of DaT-CTLESS on the regional uptake quantification task. In this trial, DaT SPECT scans of a virtual patient population, curated from CT and MR images of real patients, were generated with simulated SPECT scanners from two vendors. The Society of Nuclear Medicine (SNM) guidelines suggest using a uniform attenuation map for AC (UAC) when a CT scan is unavailable. Thus, the primary objective of ISIT-DaT was to assess whether the correlation between the activity uptake obtained using DaT-CTLESS and CTAC, as quantified using the intraclass correlation coefficient (ICC), was higher than the correlation between UAC and CTAC. Secondary objectives included evaluating DaT-CTLESS on the task of distinguishing patients with normal versus reduced DaT-specific binding ratio (SBR) of putamen, evaluating the repeatability of DaT-CTLESS in a test-retest study, assessing the generalizability across two SPECT scanners, evaluating DaT-CTLESS using fidelity-based figures of merit (FoMs), and evaluating the sensitivity of DaT-CTLESS to intra-regional uptake heterogeneity. Finally, we compared DaT-CTLESS with two other deep-learning transmission-less AC methods on regional uptake quantification across different training dataset sizes.

In the ISIT-DaT trial, data from 150 virtual patients were used to train, and another 47 were used to evaluate the DaT-CLTESS method. We observed that DaT-CTLESS yielded a significantly higher correlation with CTAC than the correlation between UAC and CTAC on the regional DaT uptake quantification task. Further, DaT-CLTESS had an excellent agreement with CTAC (ICC: 0.96, 95% CI: [0.94, 0.97], p < 0.05) on this task, significantly outperformed UAC in distinguishing patients with normal versus reduced putamen SBR, and on fidelity-based FoMs, yielded good generalizability across two scanners, was generally insensitive to intra-regional uptake heterogeneity, demonstrated good repeatability in the test-retest study, exhibited robust performance even as the size of the training data was reduced, and generally outperformed the other considered DL methods on the task of quantifying regional uptake across different training dataset sizes.

The proposed DaT-CTLESS method, as evaluated in ISIT-DaT trial, was observed to yield reliable performance for transmission-less AC in DaT-SPECT, providing a strong motivation for further clinical evaluation.