Diagnostic accuracy in FDG-PET imaging highly depends on the operating procedures. In this clinical study on dementia, we compared the diagnostic accuracy at a single-subject level of a) Clinical Scenarios, b) Standard FDG Images and c) Statistical Parametrical (SPM) Maps generated via a new optimized SPM procedure. We evaluated the added value of FDG-PET, either Standard FDG Images or SPM Maps, to Clinical Scenarios. In 88 patients with neurodegenerative diseases (Alzheimer's Disease—AD, Frontotemporal Lobar Degeneration—FTLD, Dementia with Lewy bodies—DLB and Mild Cognitive Impairment—MCI), 9 neuroimaging experts made a forced diagnostic decision on the basis of the evaluation of the three types of information. There was also the possibility of a decision of normality on the FDG-PET images. The clinical diagnosis confirmed at a long-term follow-up was used as the gold standard. SPM Maps showed higher sensitivity and specificity (96% and 84%), and better diagnostic positive (6.8) and negative (0.05) likelihood ratios compared to Clinical Scenarios and Standard FDG Images. SPM Maps increased diagnostic accuracy for differential diagnosis (AD vs. FTD; beta 1.414, p = 0.019). The AUC of the ROC curve was 0.67 for SPM Maps, 0.57 for Clinical Scenarios and 0.50 for Standard FDG Images. In the MCI group, SPM Maps showed the highest predictive prognostic value (mean LOC = 2.46), by identifying either normal brain metabolism (exclusionary role) or hypometabolic patterns typical of different neurodegenerative conditions.
Artificial intelligence (AI) models are playing an increasing role in biomedical research and healthcare services. This review focuses on challenges points to be clarified about how to develop AI applications as clinical decision support systems in the real-world context. Methods: A narrative review has been performed including a critical assessment of articles published between 1989 and 2021 that guided challenging sections. Results: We first illustrate the architectural characteristics of machine learning (ML)/radiomics and deep learning (DL) approaches. For ML/radiomics, the phases of feature selection and of training, validation, and testing are described. DL models are presented as multi-layered artificial/convolutional neural networks, allowing us to directly process images. The data curation section includes technical steps such as image labelling, image annotation (with segmentation as a crucial step in radiomics), data harmonization (enabling compensation for differences in imaging protocols that typically generate noise in non-AI imaging studies) and federated learning. Thereafter, we dedicate specific sections to: sample size calculation, considering multiple testing in AI approaches; procedures for data augmentation to work with limited and unbalanced datasets; and the interpretability of AI models (the so-called black box issue). Pros and cons for choosing ML versus DL to implement AI applications to medical imaging are finally presented in a synoptic way. Conclusions: Biomedicine and healthcare systems are one of the most important fields for AI applications and medical imaging is probably the most suitable and promising domain. Clarification of specific challenging points facilitates the development of such systems and their translation to clinical practice.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.