Automated differential diagnosis of dementia syndromes using FDG PET and machine learning

Abstract

BackgroundMetabolic brain imaging with 2-[18F]fluoro-2-deoxy-D-glucose positron emission tomography (FDG PET) is a supportive diagnostic and differential diagnostic tool for neurodegenerative dementias. In the clinic, scans are usually visually interpreted. However, computer-aided approaches can improve diagnostic accuracy. We aimed to build two machine learning classifiers, based on two sets of FDG PET-derived features, for differential diagnosis of common dementia syndromes.MethodsWe analyzed FDG PET scans from three dementia cohorts [63 dementia due to Alzheimer’s disease (AD), 79 dementia with Lewy bodies (DLB) and 23 frontotemporal dementia (FTD)], and 41 normal controls (NCs). Patients’ clinical diagnosis at follow-up (25 ± 20 months after scanning) or cerebrospinal fluid biomarkers for Alzheimer’s disease was considered a gold standard. FDG PET scans were first visually evaluated. Scans were pre-processed, and two sets of features extracted: (1) the expressions of previously identified metabolic brain patterns, and (2) the mean uptake value in 95 regions of interest (ROIs). Two multi-class support vector machine (SVM) classifiers were tested and their diagnostic performance assessed and compared to visual reading. Class-specific regional feature importance was assessed with Shapley Additive Explanations.ResultsPattern- and ROI-based classifier achieved higher overall accuracy than expert readers (78% and 80% respectively, vs. 71%). Both SVM classifiers performed similarly to one another and to expert readers in AD (F1 = 0.74, 0.78, and 0.78) and DLB (F1 = 0.81, 0.81, and 0.78). SVM classifiers outperformed expert readers in FTD (F1 = 0.87, 0.83, and 0.63), but not in NC (F1 = 0.71, 0.75, and 0.92). Visualization of the SVM model showed bilateral temporal cortices and cerebellum to be the most important features for AD; occipital cortices, hippocampi and parahippocampi, amygdala, and middle temporal lobes for DLB; bilateral frontal cortices, middle and anterior cingulum for FTD; and bilateral angular gyri, pons, and vermis for NC.ConclusionMulti-class SVM classifiers based on the expression of characteristic metabolic brain patterns or ROI glucose uptake, performed better than experts in the differential diagnosis of common dementias using FDG PET scans. Experts performed better in the recognition of normal scans and a combined approach may yield optimal results in the clinical setting.

Combined 18F‐FDG PET‐CT markers in dementia with Lewy bodies

“…Our findings confirm a previous report showing a very high specificity and PPV but a relatively low sensitivity of FDG ‐ PET to diagnose DLB, when evaluated by human expert readers , 39,42 . Indeed, as suggested by Perovnik et al., 42 lower sensitivity in combination with high specificity shows that expert readers need to recognize clear visual features of the disease and in their study, when they made a final call, they had fewer false‐positive readings than their AI model. Of interest, increasing diagnostic accuracy with longer clinical experience was demonstrated 39 …”

“…However, a relatively high number of false‐negative FDG‐PET were seen in patients with clinical DLB diagnosis, with consequent slight reduction of sensitivity for this group compared with AD group. Our findings confirm a previous report showing a very high specificity and PPV but a relatively low sensitivity of FDG ‐ PET to diagnose DLB, when evaluated by human expert readers , 39,42 . Indeed, as suggested by Perovnik et al., 42 lower sensitivity in combination with high specificity shows that expert readers need to recognize clear visual features of the disease and in their study, when they made a final call, they had fewer false‐positive readings than their AI model.…”

“…However, visual assessment is still prone to errors and inter‐rater variability, especially when the readers are not experts and fully automated tools for assessment of FDG‐PET are required 39 . Before these tools are integrated in clinical practice, a head‐to head comparison with expert's reading is warranted 42,43 …”

An artificial intelligence-derived metabolic network predicts psychosis in Alzheimer’s disease

“…The 3D ResNet101was trained and ne-tuned on the FDG PET images from the AD + P and HEC training sets to identify the ADPN and validated on their respective testing sets. The performance of the ADPN classi er was then compared to that of a conventional PET-based classi er [31], achieved through 95 FDG PET regions of interest based on the AAL atlas [32], and a support vector machine.…”

Longitudinal changes in metabolic network activity in early Alzheimer's disease

“…5,10 Indeed, ADRP expression has been found to correlate with cognitive performance in multiple AD samples. 6,7,9 While AD patients typically exhibit significant elevations in ADRP expression, which can be helpful in differentiation from other common dementia syndromes, 11 other networks may also be involved in this disease. The default mode network (DMN) is relevant in this regard.…”