Background: Adult spinal deformities (ASD) are varied spinal abnormalities, often necessitating surgical intervention when associated with pain, worsening deformity, or worsening function. Predicting post-operative complications and revision surgery is critical for surgical planning and patient counseling. Due to the relatively small number of cases of ASD surgery, machine learning applications have been limited to traditional models (e.g., logistic regression or standard neural networks) and coarse clinical variables. We present the novel application of advanced models (CNN, LLM, GWAS) using complex data types (radiographs, clinical notes, genomics) for ASD outcome prediction. Methods: We developed a CNN trained on 209 ASD patients (1549 radiographs) from the Stanford Research Repository, a CNN pre-trained on VinDr-SpineXR (10,468 spine radiographs), and an LLM using free-text clinical notes from the same 209 patients, trained via Gatortron. Additionally, we conducted a GWAS using the UK Biobank, contrasting 540 surgical ASD patients with 7355 non-surgical ASD patients. Results: The LLM notably outperformed the CNN in predicting pulmonary complications (F1: 0.545 vs. 0.2881), neurological complications (F1: 0.250 vs. 0.224), and sepsis (F1: 0.382 vs. 0.132). The pre-trained CNN showed improved sepsis prediction (AUC: 0.638 vs. 0.534) but reduced performance for neurological complication prediction (AUC: 0.545 vs. 0.619). The LLM demonstrated high specificity (0.946) and positive predictive value (0.467) for neurological complications. The GWAS identified 21 significant (p < 10⁻⁵) SNPs associated with ASD surgery risk (OR: mean: 3.17, SD: 1.92, median: 2.78), with the highest odds ratio (8.06) for the LDB2 gene, which is implicated in ectoderm differentiation. Conclusions: This study exemplifies the innovative application of cutting-edge models to forecast outcomes in ASD, underscoring the utility of complex data in outcome prediction for neurosurgical conditions. It demonstrates the promise of genetic models when identifying surgical risks and supports the integration of complex machine learning tools for informed surgical decision-making in ASD.

Background: Alzheimer’s disease (AD) is a leading worldwide cause of cognitive impairment, necessitating accurate, inexpensive diagnostic tools to enable early recognition. Methods: In this study, we present a robust deep learning approach for AD classification based on structural MRI scans, ConvNeXt, an emergent convolutional architecture inspired by vision transformers. We introduce AlzaSet, a clinically curated T1-weighted MRI dataset of 79 subjects (63 with Alzheimer’s disease [AD], 16 cognitively normal controls [NC]) acquired on a 1.5 T Siemens Aera in axial, coronal, and sagittal planes, respectively (12,947 slices in total). Images are neuroradiologist-labeled. Results are reported per plane, with awareness of the class imbalance at the subject level. We further present AlzaSet, a novel, expertly labeled clinical dataset with axial, coronal, and sagittal perspectives from AD and cognitively normal control subjects. Three ConvNeXt sizes (Tiny, Small, Base) were compared and benchmarked against existing state-of-the-art CNN models (VGG16, VGG19, InceptionV3, DenseNet121). Results: ConvNeXt-Base consistently outperformed the other models on coronal slices with an accuracy of 98.37% and an AUC of 0.992. Coronal views were determined to be most diagnostically informative, with emphasis on visualization of the medial temporal lobe. Moreover, comparison with recent ensemble-based techniques showed superior performance with comparable computational efficiency. Conclusions: These results indicate that ConvNeXt-capable models applied to clinically curated datasets have strong potential to provide scalable, real-time AD screening in diverse settings, including both high-resource and resource-constrained settings.

Alzheimer’s disease (AD) is a leading worldwide cause of cognitive impairment, necessitating accurate, inexpensive diagnostic tools to enable early recognition. In this study, we present a robust deep learning approach for AD classification based on structural MRI scans, ConvNeXt, an emergent convolutional architecture inspired by vision transformers,. We further present AlzaSet, a novel, expertly labeled clinical dataset with axial, coronal, and sagittal perspectives from AD and cognitively normal control subjects. Three ConvNeXt sizes (Tiny, Small, Base) were compared and benchmarked against existing state-of-the-art CNN models (VGG16, VGG19, InceptionV3, DenseNet1 21). ConvNeXt-Base consistently outperformed the other models on coronal slices with accuracy of 98.37% and an AUC of 0.992. Coronal views were determined to be most diagnostically informative, with emphasis on visualization of the medial temporal lobe. Moreover, comparison with recent ensemble-based techniques showed that superior performance with comparable computational efficiency. These results indicate that ConvNeXt-capable models applied to clinically curated data sets have strong potential to provide scalable, real-time AD screening in diverse settings, including both high-resource and resource-constrained settings.

Objectives: Pivotal trials have established the effectiveness of the Responsive Neurostimulation System (RNS® System) in treating focal epilepsy. In clinical trials, depth leads were primarily used to treat mesial temporal seizure onsets while cortical strip leads were used to treat neocortical seizure onsets. Here, we systematically analyze the safety and efficacy of stereoelectroencephalography (sEEG)‐guided depth leads to provide responsive stimulation to neocortical gray matter. Methods: Patients were stratified as strong responders (>median cohort seizure reduction %), weak responders (>0% and ≤median cohort seizure reduction %), and anti‐responders (≤0%) based on percent seizure reduction at 1 year post‐implant (1‐Y). Pre‐operative T1‐weighted magnetic resonance imaging and post‐operative computed tomography images were merged, and the Euclidean distance between the sEEG epileptic focus (sEEG‐EF) and the nearest RNS System depth lead contacts was calculated. Results: A total of 87 depth leads were implanted in 55 patients across neocortical brain regions. The median reduction in clinical seizures improved from 66.7% at 1‐Y to 77.5% at long‐term follow‐up (LTFU: 2.35 ± 0.95 years), with 10 patients (18.2%) achieving complete seizure freedom. Seven patients (12.7%) experienced six serious adverse events. At 1‐Y, shorter Euclidean distance between the sEEG‐EF and RNS System depth leads predicted improved seizure outcome in strong responders ( β = −0.84, p = 0.008) but not in weak responders ( β = 0.21, p = 0.9) or anti‐responders ( β = −20.34, p = 0.11). At LTFU, there was no significant relationship between Euclidean distance and seizure reduction in strong responders ( β = 0.77, p = 0.18), weak responders ( β = 2.05, p = 0.54), or anti‐responders ( β = 0.24, p = 0.99). Exploratory analyses at 1‐Y showed nominal associations between older age ( ρ = 0.32), longer epilepsy duration ( ρ = 0.27), and non‐mesial temporal sEEG‐EFs and greater seizure reduction; however, none survived Bonferroni correction (adjusted α = 0.0027; all post‐correction p > 0.0027), and no associations were observed at LTFU. Significance: In this series, neocortical depth leads for RNS therapy had favorable safety and efficacy and proximity to the sEEG‐EF drove initial outcomes for strong responders to RNS therapy. Plain Language Summary: In this multi‐center study, patients with difficult‐to‐treat seizures received brain‐responsive stimulation using a device called responsive neurostimulation (RNS), which delivers small electrical pulses to reduce seizures. We focused on patients treated with electrodes placed in the brain's outer regions (the neocortex) and guided by a mapping procedure called sEEG. On average, patients had their seizures cut by two‐thirds after one year and by more than three‐quarters with longer follow‐up, with about one in five becoming seizure‐free. The treatment was safe, and closer electrode placement to the seizure source helped explain early—but not long‐term—improvements.