PILLAR: ParaspInaL muscLe segmentAtion pRoject - a comprehensive online resource to guide manual segmentation of paraspinal muscles from magnetic resonance imaging.

BACKGROUND: There is an increasing interest in assessing paraspinal morphology and composition in relation to low back pain (LBP). However, variations in methods and segmentation protocols contribute to the inconsistent findings in the literature. We present an on-line resource, the ParaspInaL muscLe segmentAtion pRoject (PILLAR, https://projectpillar.github.io/ ), to provide a detailed description and visual guide of a segmentation protocol by using the publicly available ITK-SNAP software and discuss related challenges when performing paraspinal lumbar muscles segmentations from magnetic resonance imaging (MRI). METHODS: T2-weighted and corresponding fat-water IDEAL axial MRI from 3 males and 3 females (2 chronic LBP and 1 control for each sex) were used to demonstrate our segmentation protocol for each lumbar paraspinal muscle (erector spinae, lumbar multifidus, quadratus lumborum and psoas) and lumbar spinal level (L1-L5). RESULTS: Proper segmentation requires an understanding of the anatomy of paraspinal lumbar muscles and the variations in paraspinal muscle morphology and composition due to age, sex, and the presence of LBP or related spinal pathologies. Other challenges in segmentation includes the presence and variations of intramuscular and epimuscular fat, and side-to-side asymmetry. CONCLUSION: The growing interest to assess the lumbar musculature and its role in the development and recurrence of LBP prompted the need for comprehensive and easy-to-follow resources, such as the PILLAR project to reduce inconsistencies in segmentation protocols. Standardizing manual muscle measurements from MRI will facilitate comparisons between studies while the field is progressively moving towards the automatization of paraspinal muscle measurements for large cohort studies.

Synergistic recruitment of UbcH7~Ub and phosphorylated Ubl domain triggers parkin activation.

The E3 ligase parkin ubiquitinates outer mitochondrial membrane proteins during oxidative stress and is linked to early-onset Parkinson's disease. Parkin is autoinhibited but is activated by the kinase PINK1 that phosphorylates ubiquitin leading to parkin recruitment, and stimulates phosphorylation of parkin's N-terminal ubiquitin-like (pUbl) domain. How these events alter the structure of parkin to allow recruitment of an E2~Ub conjugate and enhanced ubiquitination is an unresolved question. We present a model of an E2~Ub conjugate bound to the phospho-ubiquitin-loaded C-terminus of parkin, derived from NMR chemical shift perturbation experiments. We show the UbcH7~Ub conjugate binds in the open state whereby conjugated ubiquitin binds to the RING1/IBR interface. Further, NMR and mass spectrometry experiments indicate the RING0/RING2 interface is re-modelled, remote from the E2 binding site, and this alters the reactivity of the RING2(Rcat) catalytic cysteine, needed for ubiquitin transfer. Our experiments provide evidence that parkin phosphorylation and E2~Ub recruitment act synergistically to enhance a weak interaction of the pUbl domain with the RING0 domain and rearrange the location of the RING2(Rcat) domain to drive parkin activity.

Fluorescent property of carbon dots extracted from cigarette smoke and the application in bio-imaging.

Cigarette smoke is one of the six major pollution sources in the room air. It contains large number of particles with size less than 10 nm. There exist carbon dots (CDs) in cigarette smoke which have strong fluorescence and with good bio-compatibility and low toxicity. CDs in cigarette smoke can be applied in bio-imaging which has great potential applications in the integration of cancer diagnosis and treatment. In this paper, CDs were extracted from cigarette smoke. Then, sodium borohydride was added to CDs aqueous solution for reduction and the reduced CDs (R-CDs) were used for biological cell imaging. The results indicate that the CDs with the particle size <10 nm in cigarette smoke are self-assembled by the polymerizated polycyclic aromatic hydrocarbons (PAHs) and ammonium nitrite which are disk nano-structure composed of sp2/sp3 carbon and oxygen/nitrogen groups or polymers. Sodium borohydride can reduce the carbonyl group on the surface of CDs to hydroxyl group and increase the ratio of the Na 1s ratio of the CDs from 1.86 to 7.42. The CDs can emit blue fluorescence under ultraviolet irradiation. After reduction, the R-CDS have the intensity of fluorescence 7.2 times than before and the fluorescence quantum yield increase from 6.13% to 8.86%. The photoluminescence (PL) wavelength of R-CDS have red-shift of 7 nm which was due to the increasing of Na element ratio. The onion epidermal cells labeled with R-CDs show that the CDs could pass through the cell wall into the cell and reach the nucleus. The cell wall and the nucleus could be clearly visualized. CDs also shows low toxicity to human bronchial epithelial cells (BEAS-2B) with good biological activity. The obtained results indicate that the CDs and R-CDs have good fluorescent property which could be used as bio-imaging agent.

Characterizing white matter alterations subject to clinical laterality in drug-naïve de novo Parkinson's disease.

Parkinson's disease (PD) is a progressive neurodegenerative disorder that is characterized by a range of motor and nonmotor symptoms, often with the motor dysfunction initiated unilaterally. Knowledge regarding disease-related alterations in white matter pathways can effectively help improve the understanding of the disease and propose targeted treatment strategies. Microstructural imaging techniques, including diffusion tensor imaging (DTI), allows inspection of white matter integrity to study the pathogenesis of various neurological conditions. Previous voxel-based analyses with DTI measures, such as fractional anisotropy and mean diffusivity have uncovered changes in brain regions that are associated with PD, but the conclusions were inconsistent, partially due to small patient cohorts and the lack of consideration for clinical laterality onset, particularly in early PD. Fixel-based analysis (FBA) is a recent framework that offers tract-specific insights regarding white matter health, but very few FBA studies on PD exist. We present a study that reveals strengthened and weakened white matter integrity that is subject to symptom laterality in a large drug-naïve de novo PD cohort using complementary DTI and FBA measures. The findings suggest that the disease gives rise to tissue degeneration and potential re-organization in the early stage.

Terahertz optical Hall effect in p-type monolayer hexagonal boron nitride on fused silica substrate.

We demonstrate for the first time, to the best of our knowledge, that the optical Hall effect (OHE) can be observed in p-type monolayer (ML) hexagonal boron nitride (hBN) on a fused silica substrate by applying linearly polarized terahertz (THz) irradiation. When ML hBN is placed on fused silica, in which the incident pulsed THz field can create local and transient electromagnetic dipoles, proximity-induced interactions can be presented. The Rashba spin-orbit coupling can be enhanced, and the in-plane spin component can be induced, along with the lifting of valley degeneracy. Thus, in the presence of linearly polarized THz radiation, the nonzero transverse optical conductivity (or Hall conductivity) can be observed. We measure the THz transmission through ML hBN/fused silica in the temperature range from 80 to 280 K by using THz time-domain spectroscopy in combination with an optical polarization examination. The Faraday ellipticity and rotation angle, together with the complex longitudinal and transverse conductivities, are obtained. The temperature dependence of these quantities is examined. The results obtained from this work indicate that ML hBN is a valleytronic material, and proximity-induced interactions can lead to the observation of OHE in the absence of an external magnetic field.

Direct visualization and characterization of the human zona incerta and surrounding structures.

The zona incerta (ZI) is a small gray matter region of the deep brain first identified in the 19th century, yet direct in vivo visualization and characterization has remained elusive. Noninvasive detection of the ZI and surrounding region could be critical to further our understanding of this widely connected but poorly understood deep brain region and could contribute to the development and optimization of neuromodulatory therapies. We demonstrate that high resolution (submillimetric) longitudinal (T1) relaxometry measurements at high magnetic field strength (7 T) can be used to delineate the ZI from surrounding white matter structures, specifically the fasciculus cerebellothalamicus, fields of Forel (fasciculus lenticularis, fasciculus thalamicus, and field H), and medial lemniscus. Using this approach, we successfully derived in vivo estimates of the size, shape, location, and tissue characteristics of substructures in the ZI region, confirming observations only previously possible through histological evaluation that this region is not just a space between structures but contains distinct morphological entities that should be considered separately. Our findings pave the way for increasingly detailed in vivo study and provide a structural foundation for precise functional and neuromodulatory investigation.

Dimerization interface of osteoprotegerin revealed by hydrogen-deuterium exchange mass spectrometry.

Previous structural studies of osteoprotegerin (OPG), a crucial negative regulator of bone remodeling and osteoclastogenesis, were mostly limited to the N-terminal ligand-binding domains. It is now known that the three C-terminal domains of OPG also play essential roles in its function by mediating OPG dimerization, OPG-heparan sulfate (HS) interactions, and formation of the OPG-HS-receptor activator of nuclear factor κB ligand (RANKL) ternary complex. Employing hydrogen-deuterium exchange MS methods, here we investigated the structure of full-length OPG in complex with HS or RANKL in solution. Our data revealed two noteworthy aspects of the OPG structure. First, we found that the interconnection between the N- and C-terminal domains is much more rigid than previously thought, possibly because of hydrophobic interactions between the fourth cysteine-rich domain and the first death domain. Second, we observed that two hydrophobic clusters located in two separate C-terminal domains directly contribute to OPG dimerization, likely by forming a hydrophobic dimerization interface. Aided by site-directed mutagenesis, we further demonstrated that an intact dimerization interface is essential for the biological activity of OPG. Our study represents an important step toward deciphering the structure-function relationship of the full-length OPG protein.

Charge-transfer dynamics and nonlocal dielectric permittivity tuned with metamaterial structures as solvent analogues.

Charge transfer (CT) is a fundamental and ubiquitous mechanism in biology, physics and chemistry. Here, we evidence that CT dynamics can be altered by multi-layered hyperbolic metamaterial (HMM) substrates. Taking triphenylene:perylene diimide dyad supramolecular self-assemblies as a model system, we reveal longer-lived CT states in the presence of HMM structures, with both charge separation and recombination characteristic times increased by factors of 2.4 and 1.7-that is, relative variations of 140 and 73%, respectively. To rationalize these experimental results in terms of driving force, we successfully introduce image dipole interactions in Marcus theory. The non-local effect herein demonstrated is directly linked to the number of metal-dielectric pairs, can be formalized in the dielectric permittivity, and is presented as a solid analogue to local solvent polarity effects. This model and extra PH3T:PC60BM results show the generality of this non-local phenomenon and that a wide range of kinetic tailoring opportunities can arise from substrate engineering. This work paves the way toward the design of artificial substrates to control CT dynamics of interest for applications in optoelectronics and chemistry.

Population-averaged MRI atlases for automated image processing and assessments of lumbar paraspinal muscles.

PURPOSE: Growing evidence suggests an association between lumbar paraspinal muscle degeneration and low back pain (LBP). Currently, time-consuming and laborious manual segmentations of paraspinal muscles are commonly performed on magnetic resonance imaging (MRI) axial scans. Automated image analysis algorithms can mitigate these drawbacks, but they often require individual MRIs to be aligned to a standard "reference" atlas. Such atlases are well established in automated neuroimaging analysis. Our aim was to create atlases of similar nature for automated paraspinal muscle measurements. METHODS: Lumbosacral T2-weighted MRIs were acquired from 117 patients who experienced LBP, stratified by gender and age group (30-39, 40-49, and 50-59 years old). Axial MRI slices of the L4-L5 and L5-S1 levels at mid-disc were obtained and aligned using group-wise linear and nonlinear image registration to produce a set of unbiased population-averaged atlases for lumbar paraspinal muscles. RESULTS: The resulting atlases represent the averaged morphology and MRI intensity features of the corresponding cohorts. Differences in paraspinal muscle shapes and fat infiltration levels with respect to gender and age can be visually identified from the population-averaged data from both linear and nonlinear registrations. CONCLUSION: We constructed a set of population-averaged atlases for developing automated algorithms to help analyze paraspinal muscle morphometry from axial MRI scans. Such an advancement could greatly benefit the fields of paraspinal muscle and LBP research. These slides can be retrieved under Electronic Supplementary Material.

Characterizing the Structure and Oligomerization of Major Royal Jelly Protein 1 (MRJP1) by Mass Spectrometry and Complementary Biophysical Tools.

Royal jelly (RJ) triggers the development of female honeybee larvae into queens. This effect has been attributed to the presence of major royal jelly protein 1 (MRJP1) in RJ. MRJP1 isolated from royal jelly is tightly associated with apisimin, a 54-residue α-helical peptide that promotes the noncovalent assembly of MRJP1 into multimers. No high-resolution structural data are available for these complexes, and their binding stoichiometry remains uncertain. We examined MRJP1/apisimin using a range of biophysical techniques. We also investigated the behavior of deglycosylated samples, as well as samples with reduced apisimin content. Our mass spectrometry (MS) data demonstrate that the native complexes predominantly exist in a (MRJP14 apisimin4) stoichiometry. Hydrogen/deuterium exchange MS reveals that MRJP1 within these complexes is extensively disordered in the range of residues 20-265. Marginally stable secondary structure (likely antiparallel ß-sheet) exists around residues 266-432. These weakly structured regions interchange with conformers that are extensively unfolded, giving rise to bimodal (EX1) isotope distributions. We propose that the native complexes have a "dimer of dimers" quaternary structure in which MRJP1 chains are bridged by apisimin. Specifically, our data suggest that apisimin acts as a linker that forms hydrophobic contacts involving the MRJP1 segment 316VLFFGLV322. Deglycosylation produces large soluble aggregates, highlighting the role of glycans as aggregation inhibitors. Samples with reduced apisimin content form dimeric complexes with a (MRJP12 apisimin1) stoichiometry. The information uncovered in this work will help pave the way toward a better understanding of the unique physiological role played by MRJP1 during queen differentiation.

Calcium-Mediated Control of S100 Proteins: Allosteric Communication via an Agitator/Signal Blocking Mechanism.

Allosteric proteins possess dynamically coupled residues for the propagation of input signals to distant target binding sites. The input signals usually correspond to "effector is present" or "effector is not present". Many aspects of allosteric regulation remain incompletely understood. This work focused on S100A11, a dimeric EF-hand protein with two hydrophobic target binding sites. An annexin peptide (Ax) served as the target. Target binding is allosterically controlled by Ca2+ over a distance of ∼26 Å. Ca2+ promotes formation of a [Ca4 S100 Ax2] complex, where the Ax peptides are accommodated between helices III/IV and III'/IV'. Without Ca2+ these binding sites are closed, precluding interactions with Ax. The allosteric mechanism was probed by microsecond MD simulations in explicit water, complemented by hydrogen exchange mass spectrometry (HDX/MS). Consistent with experimental data, MD runs in the absence of Ca2+ and Ax culminated in target binding site closure. In simulations on [Ca4 S100] the target binding sites remained open. These results capture the essence of allosteric control, revealing how Ca2+ prevents binding site closure. Both HDX/MS and MD data showed that the metalation sites become more dynamic after Ca2+ loss. However, these enhanced dynamics do not represent the primary trigger of the allosteric cascade. Instead, a labile salt bridge acts as an incessantly active "agitator" that destabilizes the packing of adjacent residues, causing a domino chain of events that culminates in target binding site closure. This agitator represents the starting point of the allosteric signal propagation pathway. Ca2+ binding rigidifies elements along this pathway, thereby blocking signal transmission. This blocking mechanism does not conform to the commonly held view that allosteric communication pathways generally originate at the sites where effectors interact with the protein.

Investigation of morphometric variability of subthalamic nucleus, red nucleus, and substantia nigra in advanced Parkinson's disease patients using automatic segmentation and PCA-based analysis.

Subthalamic nucleus (STN) deep brain stimulation (DBS) is an effective surgical therapy to treat Parkinson's disease (PD). Conventional methods employ standard atlas coordinates to target the STN, which, along with the adjacent red nucleus (RN) and substantia nigra (SN), are not well visualized on conventional T1w MRIs. However, the positions and sizes of the nuclei may be more variable than the standard atlas, thus making the pre-surgical plans inaccurate. We investigated the morphometric variability of the STN, RN and SN by using label-fusion segmentation results from 3T high resolution T2w MRIs of 33 advanced PD patients. In addition to comparing the size and position measurements of the cohort to the Talairach atlas, principal component analysis (PCA) was performed to acquire more intuitive and detailed perspectives of the measured variability. Lastly, the potential correlation between the variability shown by PCA results and the clinical scores was explored.

Virtual and augmented reality in ventriculostomy: A systematic review.

BACKGROUND: Ventriculostomy, one of the most common neurosurgical procedures, involves inserting a draining catheter into the brain's ventricular system to alleviate excessive cerebrospinal fluid (CSF) accumulation. Tradition- ally, this procedure has relied on freehand techniques guided by anatomical landmarks, which have shown a high rate of misplacement. Recent advancements in virtual reality (VR) and augmented reality (AR) technologies have opened up new possibilities in the field. This comprehensive review aims to analyze the existing literature, examine the diverse applications of VR and AR in ventriculostomy procedures, address their limitations, and propose potential future directions. METHODS: A systematic search was conducted in Web of Science and PubMed databases to identify studies employ- ing VR and AR technologies in ventriculostomy procedures. Review papers, non-English records, studies unrelated to VR/AR technologies in ventriculostomy, and supplementary documents were excluded. In total 29 papers were included in the review. RESULTS: The development of various VR and AR systems aimed at enhancing the ventriculostomy procedure are categorized according to the Data, Visualization and View (DVV) taxonomy. The study investigates the data utilized by these systems, the visualizations employed, and the virtual or augmented environments created. Furthermore, the surgical scenarios and applications of each method, as well as the validation and evaluation metrics used, are discussed. DISCUSSION: The review delves into the fundamental challenges encountered in the implementation of VR and AR systems in ventriculostomy. Additionally, potential future directions and areas for improvement are proposed, addressing the identified limitations and paving the way for further advancements in the field.

circCPA4 induces malignant behaviors of prostate cancer via miR-491-5p/SHOC2 feedback loop.

OBJECTIVE: circCPA4 has been defined to be an oncogenic gene. This study examined whether circCPA4 regulates Prostate Cancer (PC) development and revealed its molecular mechanism. METHODS: PC tissues and PC cell lines were collected, in which circCPA4/miR-491-5p/SHOC2 levels were evaluated by RT-qPCR and immunoblot. Colony formation assay and EdU assay assessed cell proliferation, flow cytometry measured apoptosis, and Transwell assessed invasion and migration. Ki-67, cleaved caspase-3, E-cadherin, and N-cadherin were evaluated by immunoblot. Based on the luciferase reporter assay and RIP assay the authors investigated the targeting relationship between circCPA4/miR-491-5p/SHOC2. The effect of circCPA4 on tumor growth was evaluated by xenotransplantation in nude mice. RESULTS: circCPA4 and SHOC2 levels were abundant while miR-491-5p expression was low in PC. Loss of circCPA4 decreased the proliferation and EdU-positive rate of PC cells, enhanced apoptosis, and inhibited invasion, migration, and EMT. Upregulation of circCPA4 forced the malignant behaviors of PC cells, and this promotion could be abolished when miR-491-5p was overexpressed or SHOC2 was silenced. CircCAP4 competitively decoyed miR-491-5p mediating SHOC2 expression. circCAP4 suppression inhibited PC tumor growth. CONCLUSION: circCAP4 acts as a novel oncogenic factor in PC, accelerating the malignant behavior of PC cells via miR-491-5p/SHOC2 interaction. This novel ceRNA axis may be a potential target for PC drug development and targeted therapy in the future.

SCANED: Siamese collateral assessment network for evaluation of collaterals from ischemic damage.

This study conducts collateral evaluation from ischemic damage using a deep learning-based Siamese network, addressing the challenges associated with a small and imbalanced dataset. The collateral network provides an alternative oxygen and nutrient supply pathway in ischemic stroke cases, influencing treatment decisions. Research in this area focuses on automated collateral assessment using deep learning (DL) methods to expedite decision-making processes and enhance accuracy. Our study employed a 3D ResNet-based Siamese network, referred to as SCANED, to classify collaterals as good/intermediate or poor. Utilizing non-contrast computed tomography (NCCT) images, the network automates collateral identification and assessment by analyzing tissue degeneration around the ischemic site. Relevant features from the left/right hemispheres were extracted, and Euclidean Distance (ED) was employed for similarity measurement. Finally, dichotomized classification of good/intermediate or poor collateral is performed by SCANED using an optimal threshold derived from ROC analysis. SCANED provides a sensitivity of 0.88, a specificity of 0.63, and a weighted F1 score of 0.86 in the dichotomized classification.

Characterization of three lamp genes from largemouth bass (Micropterus salmoides): molecular cloning, expression patterns, and their transcriptional levels in response to fast and refeeding strategy.

Lysosomes-associated membrane proteins (LAMPs), a family of glycosylated proteins and major constituents of the lysosomal membranes, play a dominant role in various cellular processes, including phagocytosis, autophagy and immunity in mammals. However, their roles in aquatic species remain poorly known. In the present study, three lamp genes were cloned and characterized from Micropterus salmoides. Subsequently, their transcriptional levels in response to different nutritional status were investigated. The full-length coding sequences of lamp1, lamp2 and lamp3 were 1251bp, 1224bp and 771bp, encoding 416, 407 and 256 amino acids, respectively. Multiple sequence alignment showed that LAMP1-3 were highly conserved among the different fish species, respectively. 3-D structure prediction, genomic survey, and phylogenetic analysis were further confirmed that these genes are widely existed in vertebrates. The mRNA expression of the three genes was ubiquitously expressed in all selected tissues, including liver, brain, gill, heart, muscle, spleen, kidney, stomach, adipose and intestine, lamp1 shows highly transcript levels in brain and muscle, lamp2 displays highly expression level in heart, muscle and spleen, but lamp3 shows highly transcript level in spleen, liver and kidney. To analyze the function of the three genes under starvation stress in largemouth bass, three experimental treatment groups (fasted group and refeeding group, control group) were established in the current study. The results indicated that the expression of lamp1 was significant induced after starvation, and then returned to normal levels after refeeding in the liver. The expression of lamp2 and lamp3 exhibited the same trend in the liver. In addition, in the spleen and the kidney, the transcript level of lamp1 and lamp2 was remarkably increased in the fasted treatment group and slightly decreased in the refed treatment group, respectively. Collectively, our findings suggest that three lamp genes may have differential function in the immune and energetic organism in largemouth bass, which is helpful in understanding roles of lamps in aquatic species.

DiffeoRaptor: diffeomorphic inter-modal image registration using RaPTOR.

PURPOSE: Diffeomorphic image registration is essential in many medical imaging applications. Several registration algorithms of such type have been proposed, but primarily for intra-contrast alignment. Currently, efficient inter-modal/contrast diffeomorphic registration, which is vital in numerous applications, remains a challenging task. METHODS: We proposed a novel inter-modal/contrast registration algorithm that leverages Robust PaTch-based cOrrelation Ratio metric to allow inter-modal/contrast image alignment and bandlimited geodesic shooting demonstrated in Fourier-Approximated Lie Algebras (FLASH) algorithm for fast diffeomorphic registration. RESULTS: The proposed algorithm, named DiffeoRaptor, was validated with three public databases for the tasks of brain and abdominal image registration while comparing the results against three state-of-the-art techniques, including FLASH, NiftyReg, and Symmetric image Normalization (SyN). CONCLUSIONS: Our results demonstrated that DiffeoRaptor offered comparable or better registration performance in terms of registration accuracy. Moreover, DiffeoRaptor produces smoother deformations than SyN in inter-modal and contrast registration. The code for DiffeoRaptor is publicly available at https://github.com/nimamasoumi/DiffeoRaptor .

Robust landmark-based brain shift correction with a Siamese neural network in ultrasound-guided brain tumor resection.

PURPOSE: In brain tumor surgery, tissue shift (called brain shift) can move the surgical target and invalidate the surgical plan. A cost-effective and flexible tool, intra-operative ultrasound (iUS) with robust image registration algorithms can effectively track brain shift to ensure surgical outcomes and safety. METHODS: We proposed to employ a Siamese neural network, which was first trained using natural images and fine-tuned with domain-specific data to automatically detect matching anatomical landmarks in iUS scans at different surgical stages. An efficient 2.5D approach and an iterative re-weighted least squares algorithm are utilized to perform landmark-based registration for brain shift correction. The proposed method is validated and compared against the state-of-the-art methods using the public BITE and RESECT datasets. RESULTS: Registration of pre-resection iUS scans to during- and post-resection iUS images were executed. The results with the proposed method shows a significant improvement from the initial misalignment ([Formula: see text]) and the method is comparable to the state-of-the-art methods validated on the same datasets. CONCLUSIONS: We have proposed a robust technique to efficiently detect matching landmarks in iUS and perform brain shift correction with excellent performance. It has the potential to improve the accuracy and safety of neurosurgery.

Deep learning for collateral evaluation in ischemic stroke with imbalanced data.

PURPOSE: Collateral evaluation is typically done using visual inspection of cerebral images and thus suffers from intra- and inter-rater variability. Large open databases of ischemic stroke patients are rare, limiting the use of deep learning methods in treatment decision-making. METHODS: We adapted a pre-trained EfficientNet B0 network through transfer learning to improve collateral evaluation using slice-based and subject-level classification. Our method uses stacking and overlapping of 2D slices from a patient's 4D computed tomography angiography (CTA) and a majority voting scheme to determine a patient's final collateral grade based on all classified 2D MIPs. Class imbalance is handled in the evaluation process by using the focal loss with class weight to penalize the majority class. RESULTS: We evaluated our method using a nine-fold cross-validation performed with 83 subjects. Mean sensitivity of 0.71, specificity of 0.84, and a weighted F1 score of 0.71 in multi-class (good, intermediate, and poor) classification were obtained. Considering treatment effect, a dichotomized decision is also made for collateral scoring of a subject based on two classes (good/intermediate and poor) which achieves a sensitivity of 0.89 and specificity of 0.96 with a weighted F1 score of 0.95. CONCLUSION: An automatic and robust collateral assessment method that mitigates the issues with the small imbalanced dataset was developed. Computer-aided evaluation of collaterals can help decision-making of ischemic stroke treatment strategy in clinical settings.