Objective: Montreal Cognitive Assessment (MoCA) as a rapid nonmotor-screening test assesses different aspects of cognitive dysfunction. Early prediction of these symptoms may facilitate better temporal therapy, disease control and identification of disease mechanisms. We aimed to predict MoCA scores in PD patients (year 4) from year 0 and 1 data. Methods: We collected 210 samples from the PPMI, and predicted their MoCA scores (range: 0-30) in year-4. In our study, 951 features including clinical features, conventional imaging features, and radiomics features were extracted from each dorsal striatum on DAT SPECT via our standardized SERA radiomics package. Four datasets (normalized by the z-score) were generated: use of the mentioned features in only (i) year-0, or (ii) year-1, iii) longitudinal-data (putting cross-sectional-datasets longitudinally next to each other), and iv) timeless-data (doubling dataset size, listing both cross-sectional-datasets separately). To predict MoCA, various Hybrid-Machine-Learning-Systems (HMLS) including 14 feature-extraction (FEA) or 10 features-selection algorithms (FSA) followed by 28 prediction algorithms (optimized by 5-fold-cross-validation and grid-search-technique) were applied on the datasets.

Objectives: Parkinson’s Disease (PD) is progressive and heterogeneous. Predicting and personalizing drug amount consistently to treat PD patients holds significant promises to enhance chances of successful temporal symptomatic therapy. Here, we aim to predict the amount of Levodopa prescribed by physicians, using Hybrid-Machine-Learning-Systems (HMLS). Methods: We selected 264 patients from the PPMI. Seven datasets were generated with 950 features including clinical-features (CF) and radiomics-features (RFs) extracted from DAT-SPECT. We selected eight outcomes including patients being on/off drug, amount of dose in specific years, and increase in drug amount, in different years. For each outcome, we considered seven datasets normalized by Z-score techniques: i,ii,iii) data in year 0, 1, or 2; iv,v) longitudinal-datasets in years 0&1, and 0&1&2, vi,vii) timeless datasets in years 0&1, and years 0&1&2 (doubling&tripling data-samples respectively). HMLSs utilized included 14 feature-extraction (FEA) and 10 feature-selection-algorithms (FSA) followed by 25 regressors and 17 classifiers (optimized by 5-fold-cross-validation and Grid-search).

Objective: Accurate prognostic stratification of Head-and-Neck-Squamous-Cell-Carcinoma (HNSCC) patients can be an important clinical reference when designing therapeutic strategies. We set to predict 4 outcomes: overall-survival (OS), distant-metastasis (DM), loco-regional-recurrence (LR), and progression-free-survival (SP). We studied hybrid-machine-learning-systems (HMLS) incorporating multiple dimensionality reduction as well as survival prediction algorithms, applied to datasets with radiomics-features. Methods: In this multi-center-study, 408 HNSCC patients were extracted from the Cancer-Imaging-Archive. PET images were registered to CT, enhanced, and cropped. 215 radiomics features were extracted from each region-of-interest via our standardized SERA radiomics package. We employed multiple HMLSs: 14 feature-extraction (FEA) or 10 feature-selection-algorithms (FSA) linked with 8 survival-prediction-algorithms (SPA) optimized by 5-fold cross-validation, applied to PET-only, CT-only and 4 PET-CT datasets generated by image-level fusion strategies. Datasets were normalized by Z-score-technique, with c-index reported to compare models.