Pre-Training, Transfer Learning and Pretext Learning for a Convolutional Neural Network Applied to Automated Assessment of Clinical PET Image Quality

TY - JOUR

T1 - Pre-Training, Transfer Learning and Pretext Learning for a Convolutional Neural Network Applied to Automated Assessment of Clinical PET Image Quality

AU - Hopson, Jessica B

AU - Neji, Radhouene

AU - Dunn, Joel T

AU - McGinnity, Colm J

AU - Flaus, Anthime

AU - Reader, Andrew J

AU - Hammers, Alexander

N1 - Publisher Copyright: © 2017 IEEE.

PY - 2023/4/1

Y1 - 2023/4/1

N2 - Positron emission tomography (PET) using a fraction of the usual injected dose would reduce the amount of radioligand needed, as well as the radiation dose to patients and staff, but would compromise reconstructed image quality. For performing the same clinical tasks with such images, a clinical (rather than numerical) image quality assessment is essential. This process can be automated with convolutional neural networks (CNNs). However, the scarcity of clinical quality readings is a challenge. We hypothesize that exploiting easily available quantitative information in pretext learning tasks or using established pre-trained networks could improve CNN performance for predicting clinical assessments with limited data. CNNs were pre-trained to predict injected dose from image patches extracted from eight real patient datasets, reconstructed using between 0.5% and 100% of the available data. Transfer learning with seven different patients was used to predict three clinically scored quality metrics ranging from 0-3: 1) global quality rating; 2) pattern recognition; and 3) diagnostic confidence. This was compared to pre-training via a VGG16 network at varying pre-training levels. Pre-training improved test performance for this task: the mean absolute error of 0.53 (compared to 0.87 without pre-training), was within clinical scoring uncertainty. Future work may include using the CNN for novel reconstruction methods performance assessment.

AB - Positron emission tomography (PET) using a fraction of the usual injected dose would reduce the amount of radioligand needed, as well as the radiation dose to patients and staff, but would compromise reconstructed image quality. For performing the same clinical tasks with such images, a clinical (rather than numerical) image quality assessment is essential. This process can be automated with convolutional neural networks (CNNs). However, the scarcity of clinical quality readings is a challenge. We hypothesize that exploiting easily available quantitative information in pretext learning tasks or using established pre-trained networks could improve CNN performance for predicting clinical assessments with limited data. CNNs were pre-trained to predict injected dose from image patches extracted from eight real patient datasets, reconstructed using between 0.5% and 100% of the available data. Transfer learning with seven different patients was used to predict three clinically scored quality metrics ranging from 0-3: 1) global quality rating; 2) pattern recognition; and 3) diagnostic confidence. This was compared to pre-training via a VGG16 network at varying pre-training levels. Pre-training improved test performance for this task: the mean absolute error of 0.53 (compared to 0.87 without pre-training), was within clinical scoring uncertainty. Future work may include using the CNN for novel reconstruction methods performance assessment.

UR - http://www.scopus.com/inward/record.url?scp=85146257082&partnerID=8YFLogxK

U2 - 10.1109/TRPMS.2022.3231702

DO - 10.1109/TRPMS.2022.3231702

M3 - Article

C2 - 37051163

SN - 2469-7311

VL - 7

SP - 372

EP - 381

JO - Transactions on Radiation and Plasma Medical Sciences

JF - Transactions on Radiation and Plasma Medical Sciences

IS - 4

ER -