An Artificial Intelligence-Based Support Tool for Lumbar Spinal Stenosis Diagnosis from Self-Reported History Questionnaire.

OBJECTIVES: Symptomatic lumbar spinal stenosis (LSS) leads to functional impairment and pain. While radiologic characterization of the morphological stenosis grade can aid in the diagnosis, it may not always correlate with patient symptoms. Artificial intelligence (AI) may diagnose symptomatic LSS in patients solely based on self-reported history questionnaires. METHODS: We evaluated multiple machine learning (ML) models to determine the likelihood of LSS using a self-reported questionnaire in patients experiencing low back pain and/or numbness in the legs. The questionnaire was built from peer-reviewed literature and a multidisciplinary panel of experts. Random forest, lasso logistic regression, support vector machine, gradient boosting trees, deep neural networks, and automated machine learning models were trained and performance metrics were compared. RESULTS: Data from 4827 patients (4690 patients without LSS: mean age 62.44, range 27-84 years, 62.8% females, and 137 patients with LSS: mean age 50.59, range 30-71 years, 59.9% females) were retrospectively collected. Among the evaluated models, the random forest model demonstrated the highest predictive accuracy with an area under the receiver operating characteristic curve (AUROC) between model prediction and LSS diagnosis of 0.96, a sensitivity of 0.94, a specificity of 0.88, a balanced accuracy of 0.91, and a Cohen's kappa of 0.85. CONCLUSIONS: Our results indicate that ML can automate the diagnosis of LSS based on self-reported questionnaires with high accuracy. Implementation of standardized and intelligence-automated workflow may serve as a supportive diagnostic tool to streamline patient management and potentially lower health care costs.

A neural network model for detection and classification of lumbar spinal stenosis on MRI.

OBJECTIVES: To develop a three-stage convolutional neural network (CNN) approach to segment anatomical structures, classify the presence of lumbar spinal stenosis (LSS) for all 3 stenosis types: central, lateral recess and foraminal and assess its severity on spine MRI and to demonstrate its efficacy as an accurate and consistent diagnostic tool. METHODS: The three-stage model was trained on 1635 annotated lumbar spine MRI studies consisting of T2-weighted sagittal and axial planes at each vertebral level. Accuracy of the model was evaluated on an external validation set of 150 MRI studies graded on a scale of absent, mild, moderate or severe by a panel of 7 radiologists. The reference standard for all types was determined by majority voting and in case of disagreement, adjudicated by an external radiologist. The radiologists' diagnoses were then compared to the diagnoses of the model. RESULTS: The model showed comparable performance to the radiologist average both in terms of the determination of presence/absence of LSS as well as severity classification, for all 3 stenosis types. In the case of central canal stenosis, the sensitivity, specificity and AUROC of the CNN were (0.971, 0.864, 0.963) for binary (presence/absence) classification compared to the radiologist average of (0.786, 0.899, 0.842). For lateral recess stenosis, the sensitivity, specificity and AUROC of the CNN were (0.853, 0.787, 0.907) compared to the radiologist average of (0.713, 0.898, 805). For foraminal stenosis, the sensitivity, specificity and AUROC of the CNN were (0.942, 0.844, 0.950) compared to the radiologist average of (0.879, 0.877, 0.878). Multi-class severity classifications showed similarly comparable statistics. CONCLUSIONS: The CNN showed comparable performance to radiologist subspecialists for the detection and classification of LSS. The integration of neural network models in the detection of LSS could bring higher accuracy, efficiency, consistency, and post-hoc interpretability in diagnostic practices.

Determining Prior Authorization Approval for Lumbar Stenosis Surgery With Machine Learning.

STUDY DESIGN: Medical vignettes. OBJECTIVES: Lumbar spinal stenosis (LSS) is a degenerative condition with a high prevalence in the elderly population, that is associated with a significant economic burden and often requires spinal surgery. Prior authorization of surgical candidates is required before patients can be covered by a health plan and must be approved by medical directors (MDs), which is often subjective and clinician specific. In this study, we hypothesized that the prediction accuracy of machine learning (ML) methods regarding surgical candidates is comparable to that of a panel of MDs. METHODS: Based on patient demographic factors, previous therapeutic history, symptoms and physical examinations and imaging findings, we propose an ML which computes the probability of spinal surgical recommendations for LSS. The model implements a random forest model trained from medical vignette data reviewed by MDs. Sets of 400 and 100 medical vignettes reviewed by MDs were used for training and testing. RESULTS: The predictive accuracy of the machine learning model was with a root mean square error (RMSE) between model predictions and ground truth of .1123, while the average RMSE between individual MD's recommendations and ground truth was .2661. For binary classification, the AUROC and Cohen's kappa were .959 and .801, while the corresponding average metrics based on individual MD's recommendations were .844 and .564, respectively. CONCLUSIONS: Our results suggest that ML can be used to automate prior authorization approval of surgery for LSS with performance comparable to a panel of MDs.

Performance of hybrid artificial intelligence in determining candidacy for lumbar stenosis surgery.

PURPOSE: Lumbar spinal stenosis (LSS) is a condition affecting several hundreds of thousands of adults in the United States each year and is associated with significant economic burden. The current decision-making practice to determine surgical candidacy for LSS is often subjective and clinician specific. In this study, we hypothesize that the performance of artificial intelligence (AI) methods could prove comparable in terms of prediction accuracy to that of a panel of spine experts. METHODS: We propose a novel hybrid AI model which computes the probability of spinal surgical recommendations for LSS, based on patient demographic factors, clinical symptom manifestations, and MRI findings. The hybrid model combines a random forest model trained from medical vignette data reviewed by surgeons, with an expert Bayesian network model built from peer-reviewed literature and the expert opinions of a multidisciplinary team in spinal surgery, rehabilitation medicine, interventional and diagnostic radiology. Sets of 400 and 100 medical vignettes reviewed by surgeons were used for training and testing. RESULTS: The model demonstrated high predictive accuracy, with a root mean square error (RMSE) between model predictions and ground truth of 0.0964, while the average RMSE between individual doctor's recommendations and ground truth was 0.1940. For dichotomous classification, the AUROC and Cohen's kappa were 0.9266 and 0.6298, while the corresponding average metrics based on individual doctor's recommendations were 0.8412 and 0.5659, respectively. CONCLUSIONS: Our results suggest that AI can be used to automate the evaluation of surgical candidacy for LSS with performance comparable to a multidisciplinary panel of physicians.

Immunoavidity-Based Capture of Tumor Exosomes Using Poly(amidoamine) Dendrimer Surfaces.

Tumor-derived blood-circulating exosomes have potential as a biomarker to greatly improve cancer treatment. However, effective isolation of exosomes remains a tremendous technical challenge. This study presents a novel nanostructured polymer surface for highly effective capture of exosomes through strong avidity. Various surface configurations, consisting of multivalent dendrimers, PEG, and tumor-targeting antibodies, were tested using exosomes isolated from tumor cell lines. We found that a dual layer dendrimer configuration exhibited the highest efficiency in capturing cultured exosomes spiked into human serum. Importantly, the optimized surface captured a > 4-fold greater amount of tumor exosomes from head and neck cancer patient plasma samples than that from healthy donors. Nanomechanical analysis using atomic force microscopy also revealed that the enhancement was attributed to multivalent binding (avidity) and augmented short-range adhesion mediated by dendrimers. Our results support that the dendrimer surface detects tumor exosomes at high sensitivity and specificity, demonstrating its potential as a new cancer liquid biopsy platform.

Distinct interaction modes of an AKAP bound to two regulatory subunit isoforms of protein kinase A revealed by amide hydrogen/deuterium exchange.

The structure of an AKAP docked to the dimerization/docking (D/D) domain of the type II (RIIalpha) isoform of protein kinase A (PKA) has been well characterized, but there currently is no detailed structural information of an AKAP docked to the type I (RIalpha) isoform. Dual-specific AKAP2 (D-AKAP2) binds in the nanomolar range to both isoforms and provided us with an opportunity to characterize the isoform-selective nature of AKAP binding using a common docked ligand. Hydrogen/deuterium (H/D) exchange combined with mass spectrometry (DXMS) was used to probe backbone structural changes of an alpha-helical A-kinase binding (AKB) motif from D-AKAP2 docked to both RIalpha and RIIalpha D/D domains. The region of protection upon complex formation and the magnitude of protection from H/D exchange were determined for both interacting partners in each complex. The backbone of the AKB ligand was more protected when bound to RIalpha compared to RIIalpha, suggesting an increased helical stabilization of the docked AKB ligand. This combined with a broader region of backbone protection induced by the AKAP on the docking surface of RIalpha indicated that there were more binding constraints for the AKB ligand when bound to RIalpha. This was in contrast to RIIalpha, which has a preformed, localized binding surface. These distinct modes of AKAP binding may contribute to the more discriminating nature of the RIalpha AKAP-docking surface. DXMS provides valuable structural information for understanding binding specificity in the absence of a high-resolution structure, and can readily be applied to other protein-ligand and protein-protein interactions.

Mapping intersubunit interactions of the regulatory subunit (RIalpha) in the type I holoenzyme of protein kinase A by amide hydrogen/deuterium exchange mass spectrometry (DXMS).

Protein kinase A (PKA), a central locus for cAMP signaling in the cell, is composed of regulatory (R) and catalytic (C) subunits. The C-subunits are maintained in an inactive state by binding to the R-subunit dimer in a tetrameric holoenzyme complex (R(2)C(2)). PKA is activated by cAMP binding to the R-subunits which induces a conformational change leading to release of the active C-subunit. Enzymatic activity of the C-subunit is thus regulated by cAMP via the R-subunit, which toggles between cAMP and C-subunit bound states. The R-subunit is composed of a dimerization/docking (D/D) domain connected to two cAMP-binding domains (cAMP:A and cAMP:B). While crystal structures of the free C-subunit and cAMP-bound states of a deletion mutant of the R-subunit are known, there is no structure of the holoenzyme complex or of the cAMP-free state of the R-subunit. An important step in understanding the cAMP-dependent activation of PKA is to map the R-C interface and characterize the mutually exclusive interactions of the R-subunit with cAMP and C-subunit. Amide hydrogen/deuterium exchange mass spectrometry is a suitable method that has provided insights into the different states of the R-subunit in solution, thereby allowing mapping of the effects of cAMP and C-subunit on different regions of the R-subunit. Our study has localized interactions with the C-subunit to a small contiguous surface on the cAMP:A domain and the linker region. In addition, C-subunit binding causes increased amide hydrogen exchange within both cAMP-domains, suggesting that these regions become more flexible in the holoenzyme and are primed to bind cAMP. Furthermore, the difference in the protection patterns between RIalpha and the previously studied RIIbeta upon cAMP-binding suggests isoform-specific differences in cAMP-dependent regulation of PKA activity.

Rapid refinement of crystallographic protein construct definition employing enhanced hydrogen/deuterium exchange MS.

Crystallographic efforts often fail to produce suitably diffracting protein crystals. Unstructured regions of proteins play an important role in this problem and considerable advantage can be gained in removing them. We have developed a number of enhancements to amide hydrogen/high-throughput and high-resolution deuterium exchange MS (DXMS) technology that allow rapid identification of unstructured regions in proteins. To demonstrate the utility of this approach for improving crystallization success, DXMS analysis was attempted on 24 Thermotoga maritima proteins with varying crystallization and diffraction characteristics. Data acquisition and analysis for 21 of these proteins was completed in 2 weeks and resulted in the localization and prediction of several unstructured regions within the proteins. When compared with those targets of known structure, the DXMS method correctly localized even small regions of disorder. DXMS analysis was then correlated with the propensity of such targets to crystallize and was further used to define truncations that improved crystallization. Truncations that were defined solely on DXMS analysis demonstrated greatly improved crystallization and have been used for structure determination. This approach represents a rapid and generalized method that can be applied to structural genomics or other targets in a high-throughput manner.

Dissecting interdomain communication within cAPK regulatory subunit type IIbeta using enhanced amide hydrogen/deuterium exchange mass spectrometry (DXMS).

cAMP-dependent protein kinase (cAPK) is a heterotetramer containing a regulatory (R) subunit dimer bound to two catalytic (C) subunits and is involved in numerous cell signaling pathways. The C-subunit is activated allosterically when two cAMP molecules bind sequentially to the cAMP-binding domains, designated A and B (cAB-A and cAB-B, respectively). Each cAMP-binding domain contains a conserved Arg residue that is critical for high-affinity cAMP binding. Replacement of this Arg with Lys affects cAMP affinity, the structural integrity of the cAMP-binding domains, and cAPK activation. To better understand the local and long-range effects that the Arg-to-Lys mutation has on the dynamic properties of the R-subunit, the amide hydrogen/deuterium exchange in the RIIbeta subunit was probed by electrospray mass spectrometry. Mutant proteins containing the Arg-to-Lys substitution in either cAMP-binding domain were deuterated for various times and then, prior to mass spectrometry analysis, subjected to pepsin digestion to localize the deuterium incorporation. Mutation of this Arg in cAB-A (Arg230) causes an increase in amide hydrogen exchange throughout the mutated domain that is beyond the modest and localized effects of cAMP removal and is indicative of the importance of this Arg in domain organization. Mutation of Arg359 (cAB-B) leads to increased exchange in the adjacent cAB-A domain, particularly in the cAB-A domain C-helix that lies on top of the cAB-B domain and is believed to be functionally linked to the cAB-B domain. This interdomain communication appears to be a unidirectional pathway, as mutation of Arg230 in cAB-A does not effect dynamics of the cAB-B domain.

Protein structure change studied by hydrogen-deuterium exchange, functional labeling, and mass spectrometry.

An automated high-throughput, high-resolution deuterium exchange HPLC-MS method (DXMS) was used to extend previous hydrogen exchange studies on the position and energetic role of regulatory structure changes in hemoglobin. The results match earlier highly accurate but much more limited tritium exchange results, extend the analysis to the entire sequence of both hemoglobin subunits, and identify some energetically important changes. Allosterically sensitive amide hydrogens located at near amino acid resolution help to confirm the reality of local unfolding reactions and their use to evaluate resolved structure changes in terms of allosteric free energy.

Dynamics of cAPK type IIbeta activation revealed by enhanced amide H/2H exchange mass spectrometry (DXMS).

cAMP-dependent protein kinase (cAPK) is a key component in numerous cell signaling pathways. The cAPK regulatory (R) subunit maintains the kinase in an inactive state until cAMP saturation of the R-subunit leads to activation of the enzyme. To delineate the conformational changes associated with cAPK activation, the amide hydrogen/deuterium exchange in the cAPK type IIbeta R-subunit was probed by electrospray mass spectrometry. Three states of the R-subunit, cAMP-bound, catalytic (C)-subunit bound, and apo, were incubated in deuterated water for various lengths of time and then, prior to mass spectrometry analysis, subjected to digestion by pepsin to localize the deuterium incorporation. High sequence coverage (>99%) by the pepsin-digested fragments enables us to monitor the dynamics of the whole protein. The effects of cAMP binding on RIIbeta amide hydrogen exchange are restricted to the cAMP-binding pockets, while the effects of C-subunit binding are evident across both cAMP-binding domains and the linker region. The decreased amide hydrogen exchange for residues 253-268 within cAMP binding domain A and for residues 102-115, which include the pseudosubstrate inhibitory site, support the prediction that these two regions represent the conserved primary and peripheral C-subunit binding sites. An increase in amide hydrogen exchange for a broad area within cAMP-binding domain B and a narrow area within cAMP-binding domain A (residues 222-232) suggest that C-subunit binding transmits long-distance conformational changes throughout the protein.

Phosphorylation driven motions in the COOH-terminal Src kinase, CSK, revealed through enhanced hydrogen-deuterium exchange and mass spectrometry (DXMS).

Previous kinetic studies demonstrated that nucleotide-derived conformational changes regulate function in the COOH-terminal Src kinase. We have employed enhanced methods of hydrogen-deuterium exchange-mass spectrometry (DXMS) to probe conformational changes on CSK in the absence and presence of nucleotides and thereby provide a structural framework for understanding phosphorylation-driven conformational changes. High quality peptic fragments covering approximately 63% of the entire CSK polypeptide were isolated using DXMS. Time-dependent deuterium incorporation into these probes was monitored to identify short peptide segments that exchange differentially with solvent. Regions expected to lie in loops exchange rapidly, whereas other regions expected to lie in stable secondary structure exchange slowly with solvent implying that CSK adopts a modular structure. The ATP analog, AMPPNP, protects probes in the active site and distal regions in the large and small lobes of the kinase domain, the SH2 domain, and the linker connecting the SH2 and kinase domains. The product ADP protects similar regions of the protein but the extent of protection varies markedly in several crucial areas. These areas correspond to the activation loop and helix G in the kinase domain and several inter-domain regions. These results imply that delivery of the gamma phosphate group of ATP induces unique local and long-range conformational changes in CSK that may influence regulatory motions in the catalytic pathway.

Application of P450RGS reporter gene system as a bioindicator of sediment PAH contamination in the vicinity of Incheon Harbor, Korea.

Sixty seven sediment samples were collected from Kyeonggi Bay, Korea, including the mouth of Han River, Incheon Harbor, the Namdong industrial complex, and the open sea. Collections were conducted in December, 1995 and samples were maintained frozen (-20 °C) until analysis. Dichloromethane extracts were analysed for their content of CYP1A1-inducing compounds with a P450RGS (reporter gene system) assay, and for polycyclic aromatic hydrocarbons (PAHs). Sediment samples were also analysed for organic carbon (OC) content and grain size, to aid in evaluating the relationship between contamination and physical nature of the sediments. The responses of the P450RGS assay to the sediment extracts were expressed as µg of benzo[a]pyrene toxic equivalents per g dry weight (µg g(-1) BaPTEQ), and these values correlated well (r(2) = 0.624 with total PAHs. BaPTEQ values were also highly correlated with the OC content of the sediments. The determination of P450RGS BaPTEQ is a useful tool, because it is both a rapid and inexpensive means of assessing the potential toxicity of organic compounds in environmental sediment samples. These values represent an estimate of the levels of compounds in the sediment that are potentially available to organisms through chronic exposure to pore water or ingestion of benthic species. We believe BaPTEQ values are more useful than tables of specific PAH concentrations, if the purpose of the investigation is to either obtain a rapid screening of an area or to develop some form of ecological or human health risk assessment.