AmyloidPETNet: Classification of Amyloid Positivity in Brain PET Imaging Using End-to-End Deep Learning.

Background Visual assessment of amyloid PET scans relies on the availability of radiologist expertise, whereas quantification of amyloid burden typically involves MRI for processing and analysis, which can be computationally expensive. Purpose To develop a deep learning model to classify minimally processed brain PET scans as amyloid positive or negative, evaluate its performance on independent data sets and different tracers, and compare it with human visual reads. Materials and Methods This retrospective study used 8476 PET scans (6722 patients) obtained from late 2004 to early 2023 that were analyzed across five different data sets. A deep learning model, AmyloidPETNet, was trained on 1538 scans from 766 patients, validated on 205 scans from 95 patients, and internally tested on 184 scans from 95 patients in the Alzheimer's Disease Neuroimaging Initiative (ADNI) fluorine 18 (18F) florbetapir (FBP) data set. It was tested on ADNI scans using different tracers and scans from independent data sets. Scan amyloid positivity was based on mean cortical standardized uptake value ratio cutoffs. To compare with model performance, each scan from both the Centiloid Project and a subset of the Anti-Amyloid Treatment in Asymptomatic Alzheimer's Disease (A4) study were visually interpreted with a confidence level (low, intermediate, high) of amyloid positivity/negativity. The area under the receiver operating characteristic curve (AUC) and other performance metrics were calculated, and Cohen κ was used to measure physician-model agreement. Results The model achieved an AUC of 0.97 (95% CI: 0.95, 0.99) on test ADNI 18F-FBP scans, which generalized well to 18F-FBP scans from the Open Access Series of Imaging Studies (AUC, 0.95; 95% CI: 0.93, 0.97) and the A4 study (AUC, 0.98; 95% CI: 0.98, 0.98). Model performance was high when applied to data sets with different tracers (AUC ≥ 0.97). Other performance metrics provided converging evidence. Physician-model agreement ranged from fair (Cohen κ = 0.39; 95% CI: 0.16, 0.60) on a sample of mostly equivocal cases from the A4 study to almost perfect (Cohen κ = 0.93; 95% CI: 0.86, 1.0) on the Centiloid Project. Conclusion The developed model was capable of automatically and accurately classifying brain PET scans as amyloid positive or negative without relying on experienced readers or requiring structural MRI. Clinical trial registration no. NCT00106899 © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Bryan and Forghani in this issue.

Infrared Spectroscopy and Mass Spectrometry of CO2 Clusters during Nucleation and Growth.

CO2 clusters with 2 to 4300 molecules are characterized with mass spectrometry and infrared spectroscopy in the uniform postnozzle flow of Laval expansions at constant temperatures of ∼29 and ∼43 K. The mass spectra provide independent, accurate information on the cluster size distributions and through magic numbers also on cluster structures. The experimental results are complemented with force field, quantum chemical, and vibrational exciton calculations. We find our data to be consistent with predominantly fcc cuboctahedral structures for clusters with more than about 50 molecules. Infrared spectra of cluster size distributions with average sizes above 140-220 molecules are completely dominated by the features from the larger cuboctahedral clusters in the distribution. For very small clusters, exciton simulations predict a pronounced broadening of the infrared band as soon as the average cluster size exceeds about five molecules. The nucleation behavior of CO2 under the present conditions is found to be barrierless in agreement with similar trends previously observed for other compounds at very high supersaturation.

Water nucleation at extreme supersaturation.

We report water cluster formation in the uniform postnozzle flow of a Laval nozzle at low temperatures of 87.0 and 47.5 K and high supersaturations of lnS ∼ 41 and 104, respectively. Cluster size distributions were measured after soft single-photon ionization at 13.8 eV with mass spectrometry. Critical cluster sizes were determined from cluster size distributions recorded as a function of increasing supersaturation, resulting in critical sizes of 6-15 and 1, respectively. Comparison with previous data for propane and toluene reveals a systematic trend in the nucleation behavior, i.e., a change from a steplike increase to a gradual increase of the maximum cluster size with increasing supersaturation. Experimental nucleation rates of 5 · 1015 cm-3 s-1 and 2 · 1015 cm-3 s-1 for lnS ∼ 41 and 104, respectively, were retrieved from cluster size distributions recorded as a function of nucleation time. These lie 2-3 orders of magnitude below the gas kinetic collision limit assuming unit sticking probability, but they agree very well with a recent prediction by a master equation model based on ab initio transition state theory. The experimental observations are consistent with barrierless growth at 47.5 K, but they hint at a more complex nucleation behavior for the measurement at 87.0 K.

Toluene Cluster Formation in Laval Expansions: Nucleation and Growth.

Toluene cluster formation has been investigated in the postnozzle flows of Laval expansions at flow temperatures between ∼48 and 73 K, toluene number concentrations between ∼1013 and 1015 cm-3, and for growth times of up to ∼170 µs. The clusters were detected by soft ionization mass spectrometry to ensure minimum cluster fragmentation upon ionization. The optimum conditions were achieved with single-photon ionization using vacuum ultraviolet (VUV) photons of 13.3 eV energy and low fluences. The nature of the onset of toluene cluster formation hints at barrierless nucleation, which seems a likely scenario for the high supersaturations (>1019) of the present experiments. This contrasts with the onset behavior observed for propane in earlier studies, which suggested nucleation in the presence of a barrier. Subsequent cluster growth has been studied as a function of the growth time for various toluene partial pressures. Size-resolved growth data have been recorded for all cluster sizes from the dimer to aggregates composed of ∼2400 monomers (∼4.4 nm in size), revealing general trends in the growth behavior. The current experiments provide systematic size- and time-resolved data on cluster formation at high supersaturations as a possible benchmark for the understanding of cluster formation under such conditions.

Single Particle Analyzer of Mass: A Charge Detection Mass Spectrometer with a Multi-Detector Electrostatic Ion Trap.

A new charge detection mass spectrometer that combines array detection and electrostatic ion trapping to repeatedly measure the masses of single ions is described. This instrument has four detector tubes inside an electrostatic ion trap with conical electrodes (cone trap) to provide multiple measurements of an ion on each pass through the trap resulting in a signal gain over a conventional trap with a single detection tube. Simulations of a cone trap and a dual ion mirror trap design indicate that more passes through the trap per unit time are possible with the latter. However, the cone trap has the advantages that ions entering up to 2 mm off the central axis of the trap are still trapped, the trapping time is less sensitive to the background pressure, and only a narrow range of energies are trapped so it can be used for energy selection. The capability of this instrument to obtain information about the molecular weight distributions of heterogeneous high molecular weight samples is demonstrated with 8 MDa polyethylene glycol (PEG) and 50 and 100 nm amine modified polystyrene nanoparticle samples. The measured mass distribution of the PEG sample is centered at 8 MDa. The size distribution obtained from mass measurements of the 100 nm nanoparticle sample is similar to the size distribution obtained from transmission electron microscopy (TEM) images, but most of the smaller nanoparticles observed in TEM images of the 50 nm nanoparticles do not reach a sufficiently high charge to trigger the trap on a single pass and be detected by the mass spectrometer. With the maximum trapping time set to 100 ms, the charge uncertainty is as low as ±2 charges and the mass uncertainty is approximately 2% for PEG and polystyrene ions.

The effect of halide and iodate anions on the hydrogen-bonding network of water in aqueous nanodrops.

The hydration of halide and iodate anions was investigated using electrospray ionization (ESI) mass spectrometry and infrared photodissociation (IRPD) spectroscopy. The average cluster sizes, determined from the abundances of X-(H2O)n (X- = F-, Cl-, Br-, I-, IO3-) in the ESI mass spectra, follow the order F- > IO3- ≈ Cl- > Br- > I-. The average cluster sizes and solution hydration enthalpies of the halides increase linearly with decreasing ionic radii, but IO3- does not fit this trend. The correlation between average cluster sizes and solution hydration enthalpies indicates that there is a similar relationship between ion-water interactions in these large gas-phase clusters and in bulk solution. The abundances of odd number clusters between n = 49 and 55 for I-, Br- and Cl- are enhanced but those for F- and IO3- are not. I- and IO3- have nearly the same ionic radii, but evidence suggests that these ions interact with water molecules differently both in solution and in small clusters. IRPD spectra of I-(H2O)n and IO3-(H2O)n, measured for select cluster sizes between n = 30 and 75 reveal differences in the hydrogen-bonding network of water molecules in these two ions even for sizes around n = 50. This indicates that differences in hydration motifs reported previously for the first hydration shells of I- and IO3- propagate to water molecules past the second solvation shell, a phenomenon that has not been reported previously for singly charged anions.

Observation of propane cluster size distributions during nucleation and growth in a Laval expansion.

We report on molecular-level studies of the condensation of propane gas and propane/ethane gas mixtures in the uniform (constant pressure and temperature) postnozzle flow of Laval expansions using soft single-photon ionization by vacuum ultraviolet light and mass spectrometric detection. The whole process, from the nucleation to the growth to molecular aggregates of sizes of several nanometers (∼5 nm), can be monitored at the molecular level with high time-resolution (∼3 µs) for a broad range of pressures and temperatures. For each time, pressure, and temperature, a whole mass spectrum is recorded, which allows one to determine the critical cluster size range for nucleation as well as the kinetics and mechanisms of cluster-size specific growth. The detailed information about the size, composition, and population of individual molecular clusters upon condensation provides unique experimental data for comparison with future molecular-level simulations.

Hydration of gaseous m-aminobenzoic acid: ionic vs neutral hydrogen bonding and water bridges.

Hydration of a protonated amine and a neutral carboxylic acid were investigated for protonated m-aminobenzoic acid (MABAH(+)) with up to 15 water molecules attached using infrared photodissociation spectroscopy, laser-induced dissociation kinetics, and computational chemistry. A free COO-H stretch in the spectra of MABAH(+)·(H2O)1-5 indicates that water does not bind to the carboxylic acid H atom. This band is absent in the spectrum of MABAH(+) with six or more water molecules attached, and there is a hydrogen-bonded (HB) COO-H stretch indicating that water hydrogen bonds to the carboxylic acid H atom for these larger clusters. Photodissociation kinetic data for MABAH(+)·(H2O)6 indicate that greater than 74 ± 13% of the ion population consists of the HB COO-H isomer, consistent with this isomer being ≥0.5 kJ mol(-1) lower in energy than isomers where the carboxylic acid H atom does not donate a hydrogen bond. Calculations at the B3LYP/6-31+G** and MP2/6-31+G**//B3LYP/6-31+G** levels of theory indicate that this energy difference is 3-5 kJ mol(-1), in agreement with the experimental results. Lower effective ion heating rates, either by attenuation of the laser power or irradiation of the ions at a lower frequency, result in more time for interconversion between the free and HB COO-H isomers. These data suggest that the barrier to dissociation for the free COO-H isomer is less than that for the HB COO-H isomer but greater than the barrier for interconversion between the two isomers. These results show the competition between hydration of a primary protonated amine vs that of a neutral carboxylic acid and the effect of water bridging between the two functional groups, which provide valuable insight into the hydration of protonated amino acids and establish rigorous benchmarks for theoretical modeling of water-biomolecule interactions.

Optimal approaches to analyzing functional MRI data in glioma patients.

BACKGROUND: Resting-state fMRI is increasingly used to study the effects of gliomas on the functional organization of the brain. A variety of preprocessing techniques and functional connectivity analyses are represented in the literature. However, there so far has been no systematic comparison of how alternative methods impact observed results. NEW METHOD: We first surveyed current literature and identified alternative analytical approaches commonly used in the field. Following, we systematically compared alternative approaches to atlas registration, parcellation scheme, and choice of graph-theoretical measure as regards differentiating glioma patients (N = 59) from age-matched reference subjects (N = 163). RESULTS: Our results suggest that non-linear, as opposed to affine registration, improves structural match to an atlas, as well as measures of functional connectivity. Functionally- as opposed to anatomically-derived parcellation schemes maximized the contrast between glioma patients and reference subjects. We also demonstrate that graph-theoretic measures strongly depend on parcellation granularity, parcellation scheme, and graph density. COMPARISON WITH EXISTING METHODS AND CONCLUSIONS: Our current work primarily focuses on technical optimization of rs-fMRI analysis in glioma patients and, therefore, is fundamentally different from the bulk of papers discussing glioma-induced functional network changes. We report that the evaluation of glioma-induced alterations in the functional connectome strongly depends on analytical approaches including atlas registration, choice of parcellation scheme, and graph-theoretical measures.

MRI-based classification of IDH mutation and 1p/19q codeletion status of gliomas using a 2.5D hybrid multi-task convolutional neural network.

Background: IDH mutation and 1p/19q codeletion status are important prognostic markers for glioma that are currently determined using invasive procedures. Our goal was to develop artificial intelligence-based methods to noninvasively determine molecular alterations from MRI. Methods: Pre-operative MRI scans of 2648 glioma patients were collected from Washington University School of Medicine (WUSM; n = 835) and publicly available Brain Tumor Segmentation (BraTS; n = 378), LGG 1p/19q (n = 159), Ivy Glioblastoma Atlas Project (Ivy GAP; n = 41), The Cancer Genome Atlas (TCGA; n = 461), and the Erasmus Glioma Database (EGD; n = 774) datasets. A 2.5D hybrid convolutional neural network was proposed to simultaneously localize glioma and classify its molecular status by leveraging MRI imaging features and prior knowledge features from clinical records and tumor location. The models were trained on 223 and 348 cases for IDH and 1p/19q tasks, respectively, and tested on one internal (TCGA) and two external (WUSM and EGD) test sets. Results: For IDH, the best-performing model achieved areas under the receiver operating characteristic (AUROC) of 0.925, 0.874, 0.933 and areas under the precision-recall curves (AUPRC) of 0.899, 0.702, 0.853 on the internal, WUSM, and EGD test sets, respectively. For 1p/19q, the best model achieved AUROCs of 0.782, 0.754, 0.842, and AUPRCs of 0.588, 0.713, 0.782, on those three data-splits, respectively. Conclusions: The high accuracy of the model on unseen data showcases its generalization capabilities and suggests its potential to perform "virtual biopsy" for tailoring treatment planning and overall clinical management of gliomas.

Non-invasive classification of IDH mutation status of gliomas from multi-modal MRI using a 3D convolutional neural network.

Glioma is the most common form of brain tumor with a high degree of heterogeneity in imaging characteristics, treatment-response, and survival rate. An important factor causing this heterogeneity is the mutation of isocitrate dehydrogenase (IDH) enzyme. The current clinical gold-standard for identifying IDH mutation status involves invasive procedures that involve risk, may fail to capture intra-tumoral spatial heterogeneity or can be inaccessible in low-resource settings. In this study, we propose a deep learning-based method to non-invasively and pre-operatively determine IDH status of high- and low-grade gliomas by leveraging their phenotypical characteristics from volumetric MRI scans. For this purpose, we propose a 3D Mask R-CNN-based approach to simultaneously detect and segment glioma as well as classify its IDH status - thus obviating the requirement of any separate tumor segmentation step. The network can operate on routinely acquired MRI sequences and is agnostic to glioma grade. It was trained on patient-cases from publicly available datasets (n = 223) and tested on two hold-out datasets acquired from The Cancer Genome Atlas (TCGA; n = 62) and Washington University School of Medicine (WUSM; n = 261). The model achieved areas under the receiver operating characteristic of 0.83 and 0.87, and areas under the precision-recall curves of 0.78 and 0.79, on the TCGA and WUSM sets, respectively. The model can be used to perform a pre-operative 'virtual biopsy' of gliomas, thus facilitating treatment planning, potentially leading to better overall survival.

Deep learning-based end-to-end scan-type classification, pre-processing, and segmentation of clinical neuro-oncology studies.

Modern neuro-oncology workflows are driven by large collections of high-dimensional MRI data obtained using varying acquisition protocols. The concomitant heterogeneity of this data makes extensive manual curation and pre-processing imperative prior to algorithmic use. The limited efforts invested towards automating this curation and processing are fragmented, do not encompass the entire workflow, or still require significant manual intervention. In this work, we propose an artificial intelligence-driven solution for transforming multi-modal raw neuro-oncology MRI Digital Imaging and Communications in Medicine (DICOM) data into quantitative tumor measurements. Our end-to-end framework classifies MRI scans into different structural sequence types, preprocesses the data, and uses convolutional neural networks to segment tumor tissue subtypes. Moreover, it adopts an expert-in-the-loop approach, where segmentation results may be manually refined by radiologists. This framework was implemented as Docker Containers (for command line usage and within the eXtensible Neuroimaging Archive Toolkit [XNAT]) and validated on a retrospective glioma dataset (n = 155) collected from the Washington University School of Medicine, comprising preoperative MRI scans from patients with histopathologically confirmed gliomas. Segmentation results were refined by a neuroradiologist, and performance was quantified using Dice Similarity Coefficient to compare predicted and expert-refined tumor masks. The scan-type classifier yielded a 99.71% accuracy across all sequence types. The segmentation model achieved mean Dice scores of 0.894 (± 0.225) for whole tumor segmentation. The proposed framework can automate tumor segmentation and characterization - thus streamlining workflows in a clinical setting as well as expediting standardized curation of large-scale neuro-oncology datasets in a research setting.

Integrative Imaging Informatics for Cancer Research: Workflow Automation for Neuro-Oncology (I3CR-WANO).

PURPOSE: Efforts to use growing volumes of clinical imaging data to generate tumor evaluations continue to require significant manual data wrangling, owing to data heterogeneity. Here, we propose an artificial intelligence-based solution for the aggregation and processing of multisequence neuro-oncology MRI data to extract quantitative tumor measurements. MATERIALS AND METHODS: Our end-to-end framework (1) classifies MRI sequences using an ensemble classifier, (2) preprocesses the data in a reproducible manner, (3) delineates tumor tissue subtypes using convolutional neural networks, and (4) extracts diverse radiomic features. Moreover, it is robust to missing sequences and adopts an expert-in-the-loop approach in which the segmentation results may be manually refined by radiologists. After the implementation of the framework in Docker containers, it was applied to two retrospective glioma data sets collected from the Washington University School of Medicine (WUSM; n = 384) and The University of Texas MD Anderson Cancer Center (MDA; n = 30), comprising preoperative MRI scans from patients with pathologically confirmed gliomas. RESULTS: The scan-type classifier yielded an accuracy of >99%, correctly identifying sequences from 380 of 384 and 30 of 30 sessions from the WUSM and MDA data sets, respectively. Segmentation performance was quantified using the Dice Similarity Coefficient between the predicted and expert-refined tumor masks. The mean Dice scores were 0.882 (±0.244) and 0.977 (±0.04) for whole-tumor segmentation for WUSM and MDA, respectively. CONCLUSION: This streamlined framework automatically curated, processed, and segmented raw MRI data of patients with varying grades of gliomas, enabling the curation of large-scale neuro-oncology data sets and demonstrating high potential for integration as an assistive tool in clinical practice.

MRI-based Identification and Classification of Major Intracranial Tumor Types by Using a 3D Convolutional Neural Network: A Retrospective Multi-institutional Analysis.

PURPOSE: To develop an algorithm to classify postcontrast T1-weighted MRI scans by tumor classes (high-grade glioma, low-grade glioma [LGG], brain metastasis, meningioma, pituitary adenoma, and acoustic neuroma) and a healthy tissue (HLTH) class. MATERIALS AND METHODS: In this retrospective study, preoperative postcontrast T1-weighted MR scans from four publicly available datasets-the Brain Tumor Image Segmentation dataset (n = 378), the LGG-1p19q dataset (n = 145), The Cancer Genome Atlas Glioblastoma Multiforme dataset (n = 141), and The Cancer Genome Atlas Low Grade Glioma dataset (n = 68)-and an internal clinical dataset (n = 1373) were used. In all, a total of 2105 images were split into a training dataset (n = 1396), an internal test set (n = 361), and an external test dataset (n = 348). A convolutional neural network was trained to classify the tumor type and to discriminate between images depicting HLTH and images depicting tumors. The performance of the model was evaluated by using cross-validation, internal testing, and external testing. Feature maps were plotted to visualize network attention. The accuracy, positive predictive value (PPV), negative predictive value, sensitivity, specificity, F1 score, area under the receiver operating characteristic curve (AUC), and area under the precision-recall curve (AUPRC) were calculated. RESULTS: On the internal test dataset, across the seven different classes, the sensitivities, PPVs, AUCs, and AUPRCs ranged from 87% to 100%, 85% to 100%, 0.98 to 1.00, and 0.91 to 1.00, respectively. On the external data, they ranged from 91% to 97%, 73% to 99%, 0.97 to 0.98, and 0.9 to 1.0, respectively. CONCLUSION: The developed model was capable of classifying postcontrast T1-weighted MRI scans of different intracranial tumor types and discriminating images depicting pathologic conditions from images depicting HLTH.Keywords MR-Imaging, CNS, Brain/Brain Stem, Diagnosis/Classification/Application Domain, Supervised Learning, Convolutional Neural Network, Deep Learning Algorithms, Machine Learning Algorithms Supplemental material is available for this article. © RSNA, 2021.

Structural Investigation of the Hormone Melatonin and Its Alkali and Alkaline Earth Metal Complexes in the Gas Phase.

Gas phase infrared dissociation spectra of the radical cation, deprotonated and protonated forms of the hormone melatonin, and its complexes with alkali (Li+, Na+, and K+) and alkaline earth metal ions (Mg2+, Ca2+, and Sr2+) are measured in the spectral range 800-1800 cm-1. Minimum energy geometries calculated at the B3LYP/LACVP++** level are used to assign structural motifs to absorption bands in the experimental spectra. The melatonin anion is deprotonated at the indole-N. The indole-C linking the amide chain is the most favored protonation site. Comparisons between the experimental and calculated spectra for alkali and alkaline earth metal ion complexes reveal that the metal ions interact similarly with the amide and methoxy oxygen atoms. The amide I band undergoes a red shift with increasing charge density of the metal ion and the amide II band shows a concomitant blue shift. Another binding motif in which the metal ions interact with the amide-O and the π-electron cloud of the aromatic group is identified but is higher in energy by at least 18 kJ/mol. Melatonin is deprotonated at the amide-N with Mg2+ and the metal ion coordinates to the amide-N and an indole-C or the methoxy-O. These results provide information about the intrinsic binding of metal ions to melatonin and combined with future studies on solvated melatonin-metal ion complexes may help elucidate the solvent effects on metal ion binding in solution and the biochemistry of melatonin. These results also serve as benchmarks for future theoretical studies on melatonin-metal ion interactions. Graphical Abstract ᅟ.

Stepwise unfolding of bovine and human serum albumin by an anionic surfactant: an investigation using the proton transfer probe norharmane.

Interactions of the anionic surfactant sodium dodecyl sulfate (SDS) with the transport proteins bovine serum albumin (BSA) and human serum albumin (HSA) have been divulged using an external photoinduced proton transfer probe, norharmane (NHM). Steady-state fluorometry, time-resolved measurements, micropolarity analysis, circular dichroism (CD), and isothermal titration calorimetry (ITC) have been exploited for the study. With the gradual addition of SDS to the probe-bound proteins, the fluorometric responses of the different prototropic species of NHM exhibit an opposite pattern as to that observed while NHM binds to the proteins. The study reveals a sequential unfolding of the serum proteins with the gradual addition of SDS. ITC measures the heat changes associated with each step of the unfolding. ITC experiments, carried out at two different pH's, elucidate the nature of interaction between SDS and the two serum proteins. At a very high concentration of SDS, the external probe (NHM) is found to be dislodged from the protein environments to bind to the SDS micellar medium.