A convolutional neural network based approach for generating full PET/CT image series from shorter scan time

2020 Fiscal Year Research-status Report

• Project/Area Number: 19K20685
• Research Institution: National Institutes for Quantum and Radiological Science and Technology
• Principal Investigator: BhusalChhatkuli Ritu 国立研究開発法人量子科学技術研究開発機構, 放射線医学総合研究所 分子イメージング診断治療研究部, 博士研究員(任非) (50836591)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: Neural Network / PET/CT / list mode PET / RNN / Prediction

Outline of Annual Research Achievements

-Acquisition of patient data from the hospital, data preprocessing and mining.
-Using Convolutional Neural Network and recurrent neural network for creating the prediction model for predicting delayed image slices. Prediction using early PET images and dynamic images.
-Successful prediction of delayed images.Qualitative and quantitative analysis.

Current Status of Research Progress

3: Progress in research has been slightly delayed.

Reason

Delayed in obtaining significant results from the prediction models. Slight change in the approach.

Strategy for Future Research Activity

-Completing the quantitative analysis based on the requirement from the doctors in the hospital.
-Submitting the result to conference.
- Publishing the paper
-Analysis with the new set of data acquired from the newly installed PET/Imaging device in hospital

Causes of Carryover

Several society attendance was cancelled due to the ongoing covid-19 pandemic.
Due to the less number of data, the previously bought server was enough for calculation so far but will have to buy big machine in future once the data increases.