Improved prediction of outcome in Parkinson's disease using radiomics analysis of longitudinal DAT SPECT images

Abstract

No disease modifying therapies for Parkinson's disease (PD) have been found effective to date. To properly power clinical trials for discovery of such therapies, the ability to predict outcome in PD is critical, and there is a significant need for discovery of prognostic biomarkers of PD. Dopamine transporter (DAT) SPECT imaging is widely used for diagnostic purposes in PD. In the present work, we aimed to evaluate whether longitudinal DAT SPECT imaging can significantly improve prediction of outcome in PD patients. In particular, we investigated whether radiomics analysis of DAT SPECT images, in addition to use of conventional non-imaging and imaging measures, could be used to predict motor severity at year 4 in PD subjects. We selected 64 PD subjects (38 male, 26 female; age at baseline (year 0): 61.9 ± 7.3, range [46,78]) from the Parkinson's Progressive Marker Initiative (PPMI) database. Inclusion criteria included (i) having had at least 2 SPECT scans at years 0 and 1 acquired on a similar scanner, (ii) having undergone a high-resolution 3 T MRI scan, and (iii) having motor assessment (MDS-UPDRS-III) available in year 4 used as outcome measure. Image analysis included automatic region-of-interest (ROI) extraction on MRI images, registration of SPECT images onto the corresponding MRI images, and extraction of radiomic features. Non-imaging predictors included demographics, disease duration as well as motor and non-motor clinical measures in years 0 and 1. The image predictors included 92 radiomic features extracted from the caudate, putamen, and ventral striatum of DAT SPECT images at years 0 and 1 to quantify heterogeneity and texture in uptake. Random forest (RF) analysis with 5000 trees was used to combine both non-imaging and imaging variables to predict motor outcome (UPDRS-III: 27.3 ± 14.7, range [3,77]). The RF prediction was evaluated using leave-one-out cross-validation. Our results demonstrated that addition of radiomic features to conventional measures significantly improved (p < 0.001) prediction of outcome, reducing the absolute error of predicting MDS-UPDRS-III from 9.00 ± 0.88 to 4.12 ± 0.43. This shows that radiomics analysis of DAT SPECT images has a significant potential towards development of effective prognostic biomarkers in PD.

Machine learning methods for optimal prediction of motor outcome in Parkinson’s disease

“…Overall, in prediction of year 4 MDS-UPDRS-III (outcome range ), predictions errors (MAE) as low as 4.32 ± 0.19 were achieved. Such prediction performance far exceeds results in other prior works where absolute errors of the order of 9 were obtained when using similar features [20]. Best results were observed when features MDS-UPDRS I (year 0 and 1), MDS-UPDRS-III (years 0 and 1), putamen as well as caudate uptake (both left and right; years 0 and 1), age and gender were in the sets.…”

“…Importantly, we saw that not using imaging information (features selected by NSGAII) can lower performance to around 4.48 (in prediction error) which was not statistically significantly than our 4.32 performance (see Supplementary materials, section III, Supplemental Table 1&2). In our other investigations we also found that conventional imaging measures do not correlate well with clinical measures (correlation coefficient = −0.008, p-value = 0.94) [29] nor improve prediction (9 ± 0.88) [20]. However, radiomics analysis of DAT SPECT images, going beyond conventional imaging measures, was seen to provide significant improvements (4.12 ± 0.43, p < 0.001) in both tasks.…”

“…Novel biomarkers may enable more accurate stratification of PD based on the expected prognosis. There is also a significant interest in using the biomarkers for prediction of disease outcome [20][21][22][23][24][25]28], to properly adapt clinical trial studies as applied to appropriate patients.…”

“…Approaches based on machine learning aim to build classification or prediction algorithms automatically by capturing statistically robust patterns present in the analyzed data [26]. In our past efforts we used radiomic (texture) measures to show improved correlation with clinical measures [29] and, when combined with clinical measures, improved prediction of motor outcome (MDS_UPDRS III) [20]. Furthermore, our recent studies confirmed significant improvement of outcome prediction based on discovering of patterns in images using deep learning (see the Section 4) [30,31].…”

From Handcrafted to Deep-Learning-Based Cancer Radiomics: Challenges and Opportunities

“…In recent years, Radiomics has been applied to many health- investigations, such as identifying the coronary plaques [33]. In neurology, it is widely applicable for detecting Alzheimer's disease [34] and Parkinson's disease [35]. However, among all the applications of the Radiomics, cancer-related topics have been the focus of interest.…”