Anterior view of the thyroid scintigraphy imaging with collimator-less gamma camera system using XCAT phantom and Monte-Carlo method

The deep learning (DL) techniques with medical imaging modalities has been studied for enhancing diagnostic accuracy and efficiency. This study aimed to the application of DL in thyroid scintigraphy for assessing thyroid function and detecting abnormalities using radiopharmaceuticals. Conventional thyroid scintigraphy methods require prolonged data acquisition times and increased patient exposure to radiopharmaceuticals. We suggested a DL-based approach that employs no collimator system based on the DL models. Training datasets were obtained using the XCAT phantom and Monte-Carlo simulation and U-Net models were trained to predict the thyroid scintigraphy images. This study evaluates the feasibility of using DL models for accurately predicting anterior view images of the thyroid scintigraphy using data acquired without any types of collimators. A quantitative analysis showed that our approach not only decreases the data acquisition time but also improves the signal-to-noise ratio (SNR) of the images. Moreover, the study explores the potential of using the reconstructed images for disease classification through a convolutional neural network (CNN) model, comparing its performance with images obtained from conventional methods.