Primary graft dysfunction (PGD) is a common complication after lung transplantation associated with poor outcomes. Although risk factors have been identified, the complex interactions between clinical variables affecting PGD risk are not well understood, which can complicate decisions about donor lung acceptance. Previously, we developed a machine learning (ML) model to predict grade 3 PGD using donor and recipient electronic health record (EHR) data, but it lacked granular information from donor lung CT scans, which are routinely assessed during offer review. In this study, we used a gated approach to determine optimal methods for analyzing donor lung CT scans among patients receiving first-time, bilateral lung transplants at a single center over 10 years. We assessed four computer vision approaches and fused the best with EHR data at three points in the ML process. A total of 160 patients had donor-lung CT scans for analysis. The best imaging-only approach employed a 3D ResNet model, yielding a median (IQR) AUROC and AUPRC of 0.63 (0.49 - 0.72) and 0.48 (0.35 - 0.6), respectively. Combining imaging with clinical data using late fusion provided the highest performance, with a median AUROC and AUPRC of 0.74 (0.59 - 0.85) and 0.61 (0.47 - 0.72), respectively.
FIFor The Alzheimer'S Disease Neuroimaging Initiative
For The Alzheimer'S Disease Neuroimaging Initiative
+8
INTRODUCTION: Previous studies have applied normative modeling on a single neuroimaging modality to investigate Alzheimer disease (AD) heterogeneity. We employed a deep learning‐based multimodal normative framework to analyze individual‐level variation across ATN (amyloid‐tau‐neurodegeneration) imaging biomarkers. METHODS: We selected cross‐sectional discovery ( n = 665) and replication cohorts ( n = 430) with available T1‐weighted magnetic resonance imaging (MRI), amyloid, and tau positron emission tomography (PET). Normative modeling estimated individual‐level abnormal deviations in amyloid‐positive individuals compared to amyloid‐negative controls. Regional abnormality patterns were mapped at different clinical group levels to assess intra‐group heterogeneity. An individual‐level disease severity index (DSI) was calculated using both the spatial extent and magnitude of abnormal deviations across ATN. RESULTS: Greater intra‐group heterogeneity in ATN abnormality patterns was observed in more severe clinical stages of AD. Higher DSI was associated with worse cognitive function and increased risk of disease progression. DISCUSSION: Subject‐specific abnormality maps across ATN reveal the heterogeneous impact of AD on the brain. Highlights: Normative modeling examined AD heterogeneity across multimodal imaging biomarkers. Heterogeneity in spatial patterns of gray matter atrophy, amyloid, and tau burden. Higher within‐group heterogeneity for AD patients at advanced dementia stages. Patient‐specific metric summarized extent of neurodegeneration and neuropathology. Metric is a marker of poor brain health and can monitor risk of disease progression.
Differences in amyloid positron emission tomography (PET) radiotracer pharmacokinetics and binding properties lead to discrepancies in amyloid‐β uptake estimates. Harmonization of tracer‐specific biases is crucial for optimal performance of downstream tasks. Here, we investigated the efficacy of ComBat, a data‐driven harmonization model, for reducing tracer‐specific biases in regional amyloid PET measurements from [18F]‐florbetapir (FBP) and [11C]‐Pittsburgh compound‐B (PiB). One hundred thirteen head‐to‐head FBP‐PiB scan pairs, scanned from the same subject within 90 days, were selected from the Open Access Series of Imaging Studies 3 (OASIS‐3) dataset. The Centiloid scale, ComBat with no covariates, ComBat with biological covariates, and GAM‐ComBat with biological covariates were used to harmonize both global and regional amyloid standardized uptake value ratios (SUVR). Variants of ComBat, including longitudinal ComBat and PEACE, were also tested. Intraclass correlation coefficient (ICC) and mean absolute error (MAE) were computed to measure the absolute agreement between tracers. Additionally, longitudinal amyloid SUVRs from an anti‐amyloid drug trial were simulated using linear mixed effects modeling. Differences in rates‐of‐change between simulated treatment and placebo groups were tested, and change in statistical power/Type‐I error after harmonization was quantified. In the head‐to‐head tracer comparison, ComBat with no covariates was the best at increasing ICC and decreasing MAE of both global summary and regional amyloid PET SUVRs between scan pairs of the same group of subjects. In the clinical trial simulation, harmonization with both Centiloid and ComBat increased statistical power of detecting true rate‐of‐change differences between groups and decreased false discovery rate in the absence of a treatment effect. The greatest benefit of harmonization was observed when groups exhibited differing FBP‐to‐PiB proportions. ComBat outperformed the Centiloid scale in harmonizing both global and regional amyloid estimates. Additionally, ComBat improved the detection of rate‐of‐change differences between clinical trial groups. Our findings suggest that ComBat is a viable alternative to Centiloid for harmonizing regional amyloid PET analyses.
10 October 2025 in Biomedical Signal Processing and Control
Complex deep learning models show high prediction tasks in various clinical prediction tasks but their inherent complexity makes it more challenging to explain model predictions for clinicians and healthcare providers. Existing research on explainability of deep learning models in healthcare have two major limitations: using post-hoc explanations and using raw clinical variables as units of explanation, both of which are often difficult for human interpretation. In this work, we designed a self-explaining deep learning framework using the expert-knowledge driven clinical concepts or intermediate features as units of explanation. The self-explaining nature of our proposed model comes from generating both explanations and predictions within the same architectural framework via joint training. We tested our proposed approach on a publicly available Electronic Health Records (EHR) dataset for predicting patient mortality in the ICU. In order to analyze the performance-interpretability trade-off, we compared our proposed model with a baseline having the same set-up but without the explanation components. Experimental results suggest that adding explainability components to a deep learning framework does not impact prediction performance and the explanations generated by the model can provide insights to the clinicians to understand the possible reasons behind patient mortality.
Alzheimer’s Disease (AD), neurodegenerative disorder, results in memory loss and cognitive decline, affecting millions of individuals globally. For patients with mild cognitive impairment (MCI), a prodromal phrase of AD, it is important to monitor their risk of progressing to AD over time. Using longitudinal multimodal data on imaging, cognition and clinical data from a publicly available AD dataset, we developed and validated HiMAL a novel multimodal Hierarchical Multi-task Auxiliary Learning framework to estimate at every visit timepoint if a MCI patient will progress to AD within the next 6 months. HiMAL predicted forecasted neuropsychological composite function scores as auxiliary tasks at every timepoint and used the predicted forecasted scores to the predict the main task of predicting progression. The hierarchical approach allows HiMAL to effectively use task dependencies, resulting in more informative features and better performance than state-of-the-art multimodal and multi-task methods. Additionally, HiMAL provided longitudinal explanations that can inform clinicians 6 months in advance the potential cognitive function decline that can lead to progression to AD in future. Built on routinely collected EHR data, HiMAL's flexibility in the selection of input modalities and auxiliary tasks suggest that it can be applied to various AD datasets and other clinical problems.