Quantitative salivary gland SPECT/CT using deep convolutional neural networks

Abstract

Quantitative single-photon emission computed tomography/computed tomography (SPECT/CT) using Tc-99m pertechnetate aids in evaluating salivary gland function. However, gland segmentation and quantitation of gland uptake is challenging. We develop a salivary gland SPECT/CT with automated segmentation using a deep convolutional neural network (CNN). The protocol comprises SPECT/CT at 20 min, sialagogue stimulation, and SPECT at 40 min post-injection of Tc-99m pertechnetate (555 MBq). The 40-min SPECT was reconstructed using the 20-min CT after misregistration correction. Manual salivary gland segmentation for %injected dose (%ID) by human experts proved highly reproducible, but took 15 min per scan. An automatic salivary segmentation method was developed using a modified 3D U-Net for end-to-end learning from the human experts (n = 333). The automatic segmentation performed comparably with human experts in voxel-wise comparison (mean Dice similarity coefficient of 0.81 for parotid and 0.79 for submandibular, respectively) and gland %ID correlation (R2 = 0.93 parotid, R2 = 0.95 submandibular) with an operating time less than 1 min. The algorithm generated results that were comparable to the reference data. In conclusion, with the aid of a CNN, we developed a quantitative salivary gland SPECT/CT protocol feasible for clinical applications. The method saves analysis time and manual effort while reducing patients’ radiation exposure.

CT-free quantitative SPECT for automatic evaluation of %thyroid uptake based on deep-learning

“…Deep-learning-based organ segmentation on CT has been actively investigated for single [ 8 , 10 ] or multiple organs [ 38 , 39 ]. Thyroid segmentation has been a major concern by radio-oncologists seeking to save the thyroid from external radiation therapy for head and neck cancer.…”

“…Recent development of deep-learning may change the concept of CTAC because CT acquisition may be omitted through either μ-map generation from SPECT (indirect approach) [ 2 – 5 ] or creation of attenuation-corrected SPECT (direct approach) [ 6 , 7 ]. Deep-learning was also useful in organ segmentation [ 8 – 10 ].…”

“…The quantitative SPECT/CT is an emerging nuclear medicine technique that employs the AC, scatter correction (SC), and resolution recovery (RR) and provides truly quantitative imaging voxels in units of radioactivity (i.e., kilo Becquerel or micro Curie) [ 11 ]. Organ segmentation is also critical in quantitative SPECT/CT because organ-level radioactivity is an important parameter of disease activity (i.e., %thyroid uptake) [ 8 , 10 ].…”

“…Technetium-99m (Tc-99m) is one of the most widely used radionuclides. Currently, quantitative SPECT/CT, which employs Tc-99m, is being actively studied in various conditions, including bone/articular [ 12 – 17 ], parathyroid [ 18 ], kidney [ 8 , 19 ], and salivary gland diseases [ 10 , 20 ]. Tc-99m pertechnetate uptake in the thyroid gland has been accurately measured using the same principle of quantitative SPECT/CT [ 21 ].…”

A deep learning model for the differential diagnosis of benign and malignant salivary gland tumors based on ultrasound imaging and clinical data

“…Moreover, DL methods have been extensively utilized in CT and MRI image analysis for PTs, including differential diagnosis of BPGTs and MPGTs, TNM classification of PTs, and evaluation of prognosis (21)(22)(23)(24). However, to the best of our knowledge, there is a paucity of data exploring the application of DL-based ultrasound (US) imaging analysis to differentially diagnose BPGT and MPGT.…”