Version of DPABI (DPARSFA) and SPM affects fMRI preprocessing result

Dear researchers and experts, 

As the COVID-19 forces us to work remotely, I deployed the latest DPABI (V4.3) and DPARSFA (V5.0_200401) on my personal computer. When processing using the same data and the same settings, I found the results were different than previously obtained. Here is what I did:

1. Preprocess resting-state fMRI data (nii). The preprocessing pipeline includes removing 10 time points, slice time correction, head motion correction, normalization, smoothing, detrend, covariates regression and filtering;
2. Construct brain networks using atlases (e.g., AAL) and save networks as a matrix (this is performed using other software);
3. Measure the maximum connection-wise difference of newly preprocessed network matrix and previously preprocessed network matrix. In other words, max(abs(net_new - net_old)) was calculated. The (max) difference was around 0.05 - 0.14 for different scans. 

Data and settings used for net_new and net_old was the same. However, after quite a few experiments, I found that it was the version of spm that affected preprocessing result. Specifically, 

1. net_old was calculated on lab HPC with DPARSFA V3.2_150710, spm12_v6470, MATLAB 2018b
2. net_new was calculated on personal PC with DPARSFA V4.3_200401, spm12_v7771, MATLAB 2019b

I tested that none of DPARSFA version, MATLAB version or brain network calculation program would influence network value, and finally found out that it was spm. 

I'm wondering if anyone know what has been changed from spm12_v6470 to spm12_v7771? How it would affect preprocessing result? Which version should I use? And regarding DPARSFA, which step might cause this difference. 

Thank you

Yunxiang Ge