A Survey on Efficient Vision Transformers: Algorithms, Techniques, and Performance Benchmarking.

Vision Transformer (ViT) architectures are becoming increasingly popular and widely employed to tackle computer vision applications. Their main feature is the capacity to extract global information through the self-attention mechanism, outperforming earlier convolutional neural networks. However, ViT deployment and performance have grown steadily with their size, number of trainable parameters, and operations. Furthermore, self-attention's computational and memory cost quadratically increases with the image resolution. Generally speaking, it is challenging to employ these architectures in real-world applications due to many hardware and environmental restrictions, such as processing and computational capabilities. Therefore, this survey investigates the most efficient methodologies to ensure sub-optimal estimation performances. More in detail, four efficient categories will be analyzed: compact architecture, pruning, knowledge distillation, and quantization strategies. Moreover, a new metric called Efficient Error Rate has been introduced in order to normalize and compare models' features that affect hardware devices at inference time, such as the number of parameters, bits, FLOPs, and model size. Summarizing, this paper firstly mathematically defines the strategies used to make Vision Transformer efficient, describes and discusses state-of-the-art methodologies, and analyzes their performances over different application scenarios. Toward the end of this paper, we also discuss open challenges and promising research directions.

Parasagittal Meningiomas: The Impact of Sinus Opening and Grade of Resection on Clinical Outcome and Recurrence in a Consecutive Series of Patients.

BACKGROUND: Surgical management of parasagittal meningiomas (PMs) remains controversial in the literature. The need to pursue a resection as radical as possible and the high risk of venous injuries contribute to making the sinus opening a widely argued choice. This study aimed to analyze factors affecting the risk of recurrence and to assess clinical outcomes of patients who underwent surgical resection of PMs with conservative or aggressive management of the intrasinusal portion. METHODS: A single-institution retrospective review of all patients with PM surgically treated between January 2013 and March 2021 was conducted. RESULTS: Among 56 patients, the sinus was opened in 32 patients (57%), and a conservative approach was used in 24 patients (43%). The sinus opening was found to be a predictive factor of radical resection (Simpson grade [SG] I-II) (P = 0.007). SG was the only predictive factor of recurrence (P < 0.001). The radical resection group (SG I-II) showed recurrence-free survival at 72 months of about 90% versus 30% in the non-radical resection group (SG III-IV) (log-rank test = 14.21, P < 0.001). Aggressive management of the sinus and radical resection were not found to be related to permanent deficit (P = 0.214 and P = 0.254) or worsening of Karnofsky performance scale score (P = 0.822 and P = 0.933). CONCLUSIONS: Removal of the intrasinusal portion of the tumor using standard procedures is not associated with a higher risk of permanent deficit or worsening of Karnofsky performance scale and reduces the risk of recurrence.

Assessment of the axial resolution of a compact gamma camera with coded aperture collimator.

PURPOSE: Handheld gamma cameras with coded aperture collimators are under investigation for intraoperative imaging in nuclear medicine. Coded apertures are a promising collimation technique for applications such as lymph node localization due to their high sensitivity and the possibility of 3D imaging. We evaluated the axial resolution and computational performance of two reconstruction methods. METHODS: An experimental gamma camera was set up consisting of the pixelated semiconductor detector Timepix3 and MURA mask of rank 31 with round holes of 0.08 mm in diameter in a 0.11 mm thick Tungsten sheet. A set of measurements was taken where a point-like gamma source was placed centrally at 21 different positions within the range of 12-100 mm. For each source position, the detector image was reconstructed in 0.5 mm steps around the true source position, resulting in an image stack. The axial resolution was assessed by the full width at half maximum (FWHM) of the contrast-to-noise ratio (CNR) profile along the z-axis of the stack. Two reconstruction methods were compared: MURA Decoding and a 3D maximum likelihood expectation maximization algorithm (3D-MLEM). RESULTS: While taking 4400 times longer in computation, 3D-MLEM yielded a smaller axial FWHM and a higher CNR. The axial resolution degraded from 5.3 mm and 1.8 mm at 12 mm to 42.2 mm and 13.5 mm at 100 mm for MURA Decoding and 3D-MLEM respectively. CONCLUSION: Our results show that the coded aperture enables the depth estimation of single point-like sources in the near field. Here, 3D-MLEM offered a better axial resolution but was computationally much slower than MURA Decoding, whose reconstruction time is compatible with real-time imaging.

Utilizing generative conversational artificial intelligence to create simulated patient encounters: a pilot study for anaesthesia training.

PURPOSE OF THE STUDY: Generative conversational artificial intelligence (AI) has huge potential to improve medical education. This pilot study evaluated the possibility of using a 'no-code' generative AI solution to create 2D and 3D virtual avatars, that trainee doctors can interact with to simulate patient encounters. METHODS: The platform 'Convai' was used to create a virtual patient avatar, with a custom backstory, to test the feasibility of this technique. The virtual patient model was set up to allow trainee anaesthetists to practice answering questions that patients' may have about interscalene nerve blocks for open reduction and internal fixation surgery. This tool was provided to anaesthetists to receive their feedback and evaluate the feasibility of this approach. RESULTS: Fifteen anaesthetists were surveyed after using the tool. The tool had a median score [interquartile range (IQR)] of 9 [7-10] in terms of how intuitive and user-friendly it was, and 8 [7-10] in terms of accuracy in simulating patient responses and behaviour. Eighty-seven percent of respondents felt comfortable using the model. CONCLUSIONS: By providing trainees with realistic scenarios, this technology allows trainees to practice answering patient questions regardless of actor availability, and indeed from home. Furthermore, the use of a 'no-code' platform allows clinicians to create customized training tools tailored to their medical specialties. While overall successful, this pilot study highlighted some of the current drawbacks and limitations of generative conversational AI, including the risk of outputting false information. Additional research and fine-tuning are required before generative conversational AI tools can act as a substitute for actors and peers.

Virtual dosimetry study with three cone-beam breast computed tomography scanners using a fast GPU-based Monte Carlo code.

Objective. To compare the dosimetric performance of three cone-beam breast computed tomography (BCT) scanners, using real-time Monte Carlo-based dose estimates obtained with the virtual clinical trials (VCT)-BREAST graphical processing unit (GPU)-accelerated platform dedicated to VCT in breast imaging. Approach. A GPU-based Monte Carlo (MC) code was developed for replicatingin silicothe geometric, x-ray spectra and detector setups adopted, respectively, in two research scanners and one commercial BCT scanner, adopting 80 kV, 60 kV and 49 kV tube voltage, respectively. Our cohort of virtual breasts included 16 anthropomorphic voxelized breast phantoms from a publicly available dataset. For each virtual patient, we simulated exams on the three scanners, up to a nominal simulated mean glandular dose of 5 mGy (primary photons launched, in the order of 1011-1012per scan). Simulated 3D dose maps (recorded for skin, adipose and glandular tissues) were compared for the same phantom, on the three scanners. MC simulations were implemented on a single NVIDIA GeForce RTX 3090 graphics card.Main results.Using the spread of the dose distribution as a figure of merit, we showed that, in the investigated phantoms, the glandular dose is more uniform within less dense breasts, and it is more uniformly distributed for scans at 80 kV and 60 kV, than at 49 kV. A realistic virtual study of each breast phantom was completed in about 3.0 h with less than 1% statistical uncertainty, with 109primary photons processed in 3.6 s computing time.Significance. We reported the first dosimetric study of the VCT-BREAST platform, a fast MC simulation tool for real-time virtual dosimetry and imaging trials in BCT, investigating the dose delivery performance of three clinical BCT scanners. This tool can be adopted to investigate also the effects on the 3D dose distribution produced by changes in the geometrical and spectrum characteristics of a cone-beam BCT scanner.

Influence of Prior Endoscopic Sinus Surgery Extent on Dupilumab Effectiveness in CRSwNP Patients.

BACKGROUND: Guidelines recommend that the vast majority of patients with severe uncontrolled chronic rhinosinusitis with nasal polyps (CRSwNP) should have at least one endoscopic sinus surgery (ESS) prior to starting biologics. Because ESS can be performed with a variable extension, the aim of this study would be to evaluate the association between surgical extensiveness, as measured by ACCESS score, and outcomes collected in patients treated with Dupilumab. MATERIALS AND METHODS: This is a multicentric retrospective study; patients affected by CRSwNP who were subjected to Dupilumab therapy and who underwent at least one ESS prior to Dupilumab initiation were included. ACCESS score was assigned to each patient's pre-Dupilumab CT scan. Subjective and objective parameters (SNOT-22, NPS, VAS scores, Sniffin' Sticks) were collected before and during the administration of therapy. Statistical correlations between ACCESS scores and clinical outcomes were investigated. RESULTS: A total of 145 patients were included; mean time from last previous ESS was 68.6 months, and on average, patients were subjected to 2.2 surgeries. Many correlations with ACCESS scores were demonstrated: better NPS at all timepoints and subjective scores (30-days SNOT-22, VAS nasal obstruction, and rhinorrhea) were achieved in patients with low ACCESS score (more extensive ESS). On the other hand, significantly worse VAS loss of smell values were demonstrated in patients with lower ACCESS scores. CONCLUSION: Dupilumab patients subjected to a prior extensive ESS may have reduced size of polyps and improved subjective indicators, together with a decreased chance to recover smell, when compared with patients who underwent a minimal excision. LEVEL OF EVIDENCE: 3 Laryngoscope, 134:1556-1563, 2024.

Facial trustworthiness dampens own-gender bias in emotion recognition.

Previous research suggests that emotion recognition is influenced by social categories derived by invariant facial features such as gender and inferences of trustworthiness from facial appearance. The current study sought to replicate and extend these findings by examining the intersection of these social categories on recognition of emotional facial expressions. We used a dynamic emotion recognition task to assess accuracy and response times in the happiness and anger categorization displayed by female and male faces that differed in the degree of facial trustworthiness (i.e., trustworthy- vs. untrustworthy-looking faces). We found that facial trustworthiness was able to modulate the own-gender bias on emotion recognition, as responses to untrustworthy-looking faces revealed a bias towards ingroup members. Conversely, when faces look trustworthy, no differences on emotion recognition between female and male faces were found. In addition, positive inferences of trustworthiness lead to faster recognition of happiness in females and anger in males, showing that facial appearance was able to influence also the intersection between social categories and specific emotional expressions. Together, these results suggest that facial appearance, probably due to the activation of approach or avoidance motivational systems, is able to modulate the own-gender bias on emotion recognition.

Monte Carlo and experimental evaluation of a Timepix4 compact gamma camera for coded aperture nuclear medicine imaging with depth resolution.

PURPOSE: We designed a prototype compact gamma camera (MediPROBE4) for nuclear medicine tasks, including radio-guided surgery and sentinel lymph node imaging with a 99mTc radiotracer. We performed Monte Carlo (MC) simulations for image performance assessment, and first spectroscopic imaging tests with a 300 µm thick silicon detector. METHODS: The hand-held camera (1 kg weight) is based on a Timepix4 readout circuit for photon-counting, energy-sensitive, hybrid pixel detectors (24.6 × 28.2 mm2 sensitive area, 55 µm pixel pitch), developed by the Medipix4 Collaboration. The camera design adopts a CdTe detector (1 or 2 mm thick) bump-bonded to a Timepix4 readout chip and a coded aperture collimator with 0.25 mm diameter round holes made of 3D printed 1-mm thick tungsten. Image reconstruction is performed via autocorrelation deconvolution. RESULTS: Geant4 MC simulations showed that, for a 99mTc source in air, at 50 mm source-collimator distance, the estimated collimator sensitivity (4 × 10-4) is 292 times larger than that of a single hole in the mask; the system sensitivity is 0.22 cps/kBq (2 mm CdTe); the lateral spatial resolution is 1.7 mm FWHM. The estimated axial longitudinal resolution is 8.2 mm FWHM at 40 mm distance. First experimental tests with a 300 µm thick Silicon pixel detector bump-bonded to a Timepix4 chip and a high-resolution coded aperture collimator showed time-over-threshold and time-of-arrival capabilities with 241Am and 133Ba gamma-ray sources. CONCLUSIONS: MC simulations and validation lab tests showed the expected performance of the MediPROBE4 compact gamma camera for gamma-ray 3D imaging.

Anatomical landmarks during high-magnification microsurgery for a safe and effective resection of high-grade gliomas: how I do it.

BACKGROUND: Gross total resection, when possible, is the first crucial treatment for high-grade gliomas, as it has been demonstrated to be associated with longer survival. Different intraoperative tools, such as neuronavigation, fluorescent agents, and intra-operative ultrasound, have been developed to help neurosurgeons to extend the resection. METHODS: We describe the high-magnification microsurgery technique used during the first surgical removal for high-grade gliomas. We illustrate the key anatomical "markers" of normal brain parenchyma, which guide the surgery. CONCLUSION: High-magnification microsurgery is an anatomically based approach that allows the identification of key anatomical "markers" of normal brain parenchyma in order to resect high-grade gliomas safely and effectively.

Bridge Damage Identification Using Deep Neural Networks on Time-Frequency Signals Representation.

For the purpose of maintaining and prolonging the service life of civil constructions, structural damage must be closely monitored. Monitoring the incidence, formation, and spread of damage is crucial to ensure a structure's ongoing performance. This research proposes a unique approach for multiclass damage detection using acceleration responses based on synchrosqueezing transform (SST) together with deep learning algorithms. In particular, our pipeline is able to classify correctly the time series representing the responses of accelerometers placed on a bridge, which are classified with respect to different types of damage scenarios applied to the bridge. Using benchmark data from the Z24 bridge for multiclass classification for different damage situations, the suggested method is validated. This dataset includes labeled accelerometer measurements from a real-world bridge that has been gradually damaged by various conditions. The findings demonstrate that the suggested approach is successful in exploiting pre-trained 2D convolutional neural networks, obtaining a high classification accuracy that can be further boosted by the application of simple voting methods.

A systematic review and meta-analysis of the association between young adults' sleep habits and substance use, with a focus on self-medication behaviours.

Young adults (18-30 years) are vulnerable to sleep-wake disturbances and substance use, which are bi-directionally associated. The present work aims to organise the literature that deals with the association between sleep and substance use in young adults, also considering self-medication behaviours. We adopted a framework that accounts for the multidimensionality of sleep and the effect of different substances. We considered sleep disturbances (insomnia symptoms, sleep quality), sleep health dimensions (duration, satisfaction, efficiency, timing, daytime alertness), circadian characteristics (chronotype). Substances were alcohol, caffeine, nicotine, cannabis, others. We included 46 studies. The use of caffeine and nicotine was associated with higher odds of sleep disturbances. No significant effect was detected for sleep duration. In narrative findings, daytime dysfunction was associated with alcohol and caffeine use, and poor sleep satisfaction with nicotine use. Few evidence were available for the other sleep health dimensions. Evening chronotype was associated with alcohol, caffeine, and nicotine use. Few studies focused on cannabis or self-medication. Longitudinal results were inconclusive. We found a distinct pattern of associations between different substances and different sleep outcomes. Further investigation considering the multidimensionality of sleep would create a better understanding of the complex relationship between substance use and sleep health in young adults.

Familiarity From Facial Appearance Leads to Hypoalgesia.

Social context has been shown to influence pain perception. This study aimed to broaden this literature by investigating whether relevant social stimuli, such as faces with different levels of intrinsic (based on physical resemblance to known individuals) and episodic (acquired through a previous experience) familiarity, may lead to hypoalgesia. We hypothesized that familiarity, whether intrinsic or acquired through experience, would increase pain threshold and decrease pain intensity. Sixty-seven participants underwent pain induction (the cold pressor test) viewing previously seen faces (Episodic Group) or new faces (Non-episodic Group) that differed in the level of intrinsic familiarity (high vs low). Pain threshold was measured in seconds, while pain intensity was measured on a rating scale of 0 to 10. The results did not show an effect of episodic familiarity. However, compared to low, high intrinsic familiar faces had an attenuating effect on pain intensity, even after controlling for pain expectation. These results suggest that physical features conveying a higher feeling of familiarity induce a top-down hypoalgesic modulation, in line with the idea that familiarity may signal safety and that the presence of familiar others reduce perceived threat-related distress. This study provides further evidence on the social modulation of pain and contributes to the literature on first impressions' influence on social behavior. PERSPECTIVE: Consistent with the idea that familiar others signal safety and reduce the sense of threat, facial features conveying familiarity induce a top-down hypoalgesic modulation. This knowledge may contribute to understanding differences in pain perception in experimental and clinical contexts.

No(cebo) Vax: COVID-19 Vaccine Beliefs Are Important Determinants of Both Occurrence and Perceived Severity of Common Vaccines' Adverse Effects.

This study highlights the role of psychological influences in triggering and amplifying the adverse effects of the COVID-19 vaccine (i.e., nocebo effects). Fear, beliefs, and expectations about the COVID-19 vaccine, trust in health and scientific institutions, and stable personality traits were measured in 315 adult Italian citizens (145 men) during the 15-min waiting time after vaccination. The occurrence and severity of 10 potential adverse effects were assessed 24 hr later. Nonpharmacological variables predicted nearly 30% of the severity of the vaccine's adverse effects. Expectations are important determinants of adverse effects from vaccines, and the results of the path analyses show that these expectations stem primarily from people's vaccine beliefs and attitudes, which can be changed. Implications for increasing vaccine acceptability and limiting the nocebo effect are discussed.

Patient-specific Monte Carlo-based organ dose estimates in spiral CT via optical 3D body scanning and adaptation of a voxelized phantom dataset: proof-of-principle.

Objective. We present a method for personalized organ dose estimates obtained before the computed tomography (CT) exam, via 3D optical body scanning and Monte Carlo (MC) simulations.Approach. A voxelized phantom is derived by adapting a reference phantom to the body size and shape measured with a portable 3D optical scanner, which returns the 3D silhouette of the patient. This was used as an external rigid envelope for incorporating a tailored version of the internal body anatomy derived from a phantom dataset (National Cancer Institute, NIH, USA) matched for gender, age, weight, and height. The proof-of-principle was conducted on adult head phantoms. The Geant4 MC code provided estimates of the organ doses from 3D absorbed dose maps in the voxelized body phantom.Main results. We applied this approach for head CT scanning using an anthropomorphic voxelized head phantom derived from 3D optical scans of manikins. We compared the estimates of head organ doses with those provided by the NCICT 3.0 software (NCI, NIH, USA). Head organ doses differed up to 38% using the proposed personalized estimate and MC code, with respect to corresponding estimates calculated for the standard (non-personalized) reference head phantom. Preliminary application of the MC code to chest CT scans is shown. Real-time pre-exam personalized CT dosimetry is envisaged with adoption of a Graphics Processing Unit-based fast MC code.Significance. The developed procedure for personalized organ dose estimates before the CT exam, introduces a new approach for realistic description of size and shape of patients via voxelized phantoms specific for each patient.

Anomaly detection in railway bridges using imaging techniques.

The monitoring of the structural health of infrastructures is a very important topic in structural engineering, but unfortunately, there are few established techniques that are applicable in a wide range of situations. In this paper, we present a new method that adapts image analysis tools and methodologies, taken from the field of computer vision, and applies them to the monitoring signals of a railway bridge. We show that our method correctly identifies changes in the structural health of the bridge with very high precision, thus providing a better, simpler, and more general alternative to current methodologies used in the field.

Key role of microsurgical dissections on cadaveric specimens in neurosurgical training: Setting up a new research anatomical laboratory and defining neuroanatomical milestones.

Introduction: Neurosurgery is one of the most complex surgical disciplines where psychomotor skills and deep anatomical and neurological knowledge find their maximum expression. A long period of preparation is necessary to acquire a solid theoretical background and technical skills, improve manual dexterity and visuospatial ability, and try and refine surgical techniques. Moreover, both studying and surgical practice are necessary to deeply understand neuroanatomy, the relationships between structures, and the three-dimensional (3D) orientation that is the core of neurosurgeons' preparation. For all these reasons, a microsurgical neuroanatomy laboratory with human cadaveric specimens results in a unique and irreplaceable training tool that allows the reproduction of patients' positions, 3D anatomy, tissues' consistencies, and step-by-step surgical procedures almost identical to the real ones. Methods: We describe our experience in setting up a new microsurgical neuroanatomy lab (IRCCS Neuromed, Pozzilli, Italy), focusing on the development of training activity programs and microsurgical milestones useful to train the next generation of surgeons. All the required materials and instruments were listed. Results: Six competency levels were designed according to the year of residency, with training exercises and procedures defined for each competency level: (1) soft tissue dissections, bone drilling, and microsurgical suturing; (2) basic craniotomies and neurovascular anatomy; (3) white matter dissection; (4) skull base transcranial approaches; (5) endoscopic approaches; and (6) microanastomosis. A checklist with the milestones was provided. Discussion: Microsurgical dissection of human cadaveric specimens is the optimal way to learn and train on neuroanatomy and neurosurgical procedures before performing them safely in the operating room. We provided a "neurosurgery booklet" with progressive milestones for neurosurgical residents. This step-by-step program may improve the quality of training and guarantee equal skill acquisition across countries. We believe that more efforts should be made to create new microsurgical laboratories, popularize the importance of body donation, and establish a network between universities and laboratories to introduce a compulsory operative training program.

Lightweight and Energy-Aware Monocular Depth Estimation Models for IoT Embedded Devices: Challenges and Performances in Terrestrial and Underwater Scenarios.

The knowledge of environmental depth is essential in multiple robotics and computer vision tasks for both terrestrial and underwater scenarios. Moreover, the hardware on which this technology runs, generally IoT and embedded devices, are limited in terms of power consumption, and therefore, models with a low-energy footprint are required to be designed. Recent works aim at enabling depth perception using single RGB images on deep architectures, such as convolutional neural networks and vision transformers, which are generally unsuitable for real-time inferences on low-power embedded hardware. Moreover, such architectures are trained to estimate depth maps mainly on terrestrial scenarios due to the scarcity of underwater depth data. Purposely, we present two lightweight architectures based on optimized MobileNetV3 encoders and a specifically designed decoder to achieve fast inferences and accurate estimations over embedded devices, a feasibility study to predict depth maps over underwater scenarios, and an energy assessment to understand which is the effective energy consumption during the inference. Precisely, we propose the MobileNetV3S75 configuration to infer on the 32-bit ARM CPU and the MobileNetV3LMin for the 8-bit Edge TPU hardware. In underwater settings, the proposed design achieves comparable estimations with fast inference performances compared to state-of-the-art methods. Moreover, we statistically proved that the architecture of the models has an impact on the energy footprint in terms of Watts required by the device during the inference. Then, the proposed architectures would be considered to be a promising approach for real-time monocular depth estimation by offering the best trade-off between inference performances, estimation error and energy consumption, with the aim of improving the environment perception for underwater drones, lightweight robots and Internet of things.

The Italian Validation of the Zimbardo Time Perspective Inventory and Its Comparison with Three Time Perspective Inventories.

The Zimbardo time perspective inventory (ZTPI) is the most well-known and widely used measure of time perspective. However, the assessment of the psychometric properties of the ZTPI reveals several problems, and various short versions have been proposed to overcome these problems. In a large Italian sample (N = 2295; 1326 women; age range 18-74 years), the present study aimed to test a short version of the ZTPI (ZTPI-16) defined by high frequency items (i.e., "good" items), reviewing the items composition of previous alternative short versions of the scale. Beyond the assessment of the factorial structure of this new short ZTPI, we compared the ZTPI-16 to the original ZTPI (ZTPI-56) and another already validated version of the ZTPI in the Italian context, such as Zimbardo's Stanford time perspective inventory (ZTPI-22), the short version of the ZTPI (ZTPI-30), and the ZTPI-36 proposed analyzing the data from 24 countries. The results confirmed the psychometric problems of the ZTPI-56, whereas the ZTPI-16 reported adequate structural validity and reliability. Moderate-to-strong correlations between same temporal subscales in different ZTPI versions were also found. These data suggest that the review of the "good" items is a new direction in the development of ZTPI versions with good psychometric properties and comparable data among cultures.