Evaluating Computer Vision, Large Language, and Genome-Wide Association Models in a Limited Sized Patient Cohort for Pre-Operative Risk Stratification in Adult Spinal Deformity Surgery

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.

Anatomical and topographical variations in the distribution of brain metastases based on primary cancer origin and molecular subtypes: a systematic review

Background

While it has been suspected that different primary cancers have varying predilections for metastasis in certain brain regions, recent advances in neuroimaging and spatial modeling analytics have facilitated further exploration into this field.

Methods

A systematic electronic database search for studies analyzing the distribution of brain metastases (BMs) from any primary systematic cancer published between January 1990 and July 2020 was conducted using PRISMA guidelines.

Results

Two authors independently reviewed 1957 abstracts, 46 of which underwent full-text analysis. A third author arbitrated both lists; 13 studies met inclusion/exclusion criteria. All were retrospective single- or multi-institution database reviews analyzing over 8227 BMs from 2599 patients with breast (8 studies), lung (7 studies), melanoma (5 studies), gastrointestinal (4 studies), renal (3 studies), and prostate (1 study) cancers. Breast, lung, and colorectal cancers tended to metastasize to more posterior/caudal topographic and vascular neuroanatomical regions, particularly the cerebellum, with notable differences based on subtype and receptor expression. HER-2-positive breast cancers were less likely to arise in the frontal lobes or subcortical region, while ER-positive and PR-positive breast metastases were less likely to arise in the occipital lobe or cerebellum. BM from lung adenocarcinoma tended to arise in the frontal lobes and squamous cell carcinoma in the cerebellum. Melanoma metastasized more to the frontal and temporal lobes.

Conclusion

The observed topographical distribution of BM likely develops based on primary cancer type, molecular subtype, and genetic profile. Further studies analyzing this association and relationships to vascular distribution are merited to potentially improve patient treatment and outcomes.

Gene and protein expression in pituitary corticotroph adenomas: a systematic review of the literature

OBJECT Functional corticotroph pituitary adenomas (PAs) secrete adrenocorticotropic hormone (ACTH) and are the cause of Cushing's disease, which accounts for 70% of all cases of Cushing's syndrome. Current classification systems for PAs rely primarily on laboratory hormone findings, tumor size and morphology, invasiveness, and immunohistochemical findings. Likewise, drug development for functional ACTH-secreting PAs (ACTH-PAs) is limited and has focused largely on blocking the production or downstream effects of excess cortisol. The authors aimed to summarize the findings from previous studies that explored gene and protein expression of ACTH-PAs to prioritize potential genetic and protein targets for improved molecular diagnosis and treatment of Cushing's disease. METHODS A systematic literature review was performed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A PubMed search of select medical subject heading (MeSH) terms was performed to identify all studies that reported gene- and protein-expression findings in ACTH-PAs from January 1, 1990, to August 24, 2014, the day the search was performed. The inclusion criteria were studies on functional ACTH-PAs compared with normal pituitary glands, on human PA tissue only, with any method of analysis, and published in the English language. Studies using anything other than resected PA tissue, those that compared other adenoma types, those without baseline expression data, or those in which any pretreatment was delivered before analysis were excluded. RESULTS The primary search returned 1371 abstracts, of which 307 were found to be relevant. Of those, 178 were selected for secondary full-text analysis. Of these, 64 articles met the inclusion criteria and an additional 4 studies were identified from outside the search for a total of 68 included studies. Compared with the normal pituitary gland, significant gene overexpression in 43 genes and 22 proteins was reported, and gene underexpression in 58 genes and 15 proteins was reported. Immunohistochemistry was used in 39 of the studies, and reverse transcriptase polymerase chain reaction was used in 26 of the studies, primarily, and as validation for 4 others. Thirteen studies used both immunohistochemistry and reverse transcriptase polymerase chain reaction. Other methods used included microarray, in situ hybridization, Northern blot analysis, and Western blot analysis. Expression of prioritized genes emphasized in multiple studies were often validated on both the gene and protein levels. Genes/proteins found to be overexpressed in ACTH-PAs relative to the normal pituitary gland included hPTTG1/securin, NEUROD1/NeuroD1 (Beta2), HSD11B2/11β-hydroxysteroid dehydrogenase 2, AKT/Akt, protein kinase B, and CCND1/cyclin D1. Candidate genes/proteins found to be underexpressed in ACTH-PAs relative to the normal pituitary gland included CDKN1B/p27(Kip1), CDKN2A/p16, KISS1/kisspeptin, ACTHR/ACTH-R, and miR-493. CONCLUSIONS On the basis of the authors' systematic review, many significant gene and protein targets that may contribute to tumorigenesis, invasion, and hormone production/secretion of ACTH have been identified and validated in ACTH-PAs. Many of these potential targets have not been fully analyzed for their therapeutic and diagnostic potential but may represent candidate molecular targets for biomarker development and drug targeting. This review may help catalyze additional research efforts using modern profiling and sequencing techniques and alteration of gene expression.