Optimizing Vision Transformers for Histopathology: Pretraining and Normalization in Breast Cancer Classification.

This paper introduces a self-attention Vision Transformer model specifically developed for classifying breast cancer in histology images. We examine various training strategies and configurations, including pretraining, dimension resizing, data augmentation and color normalization strategies, patch overlap, and patch size configurations, in order to evaluate their impact on the effectiveness of the histology image classification. Additionally, we provide evidence for the increase in effectiveness gathered through geometric and color data augmentation techniques. We primarily utilize the BACH dataset to train and validate our methods and models, but we also test them on two additional datasets, BRACS and AIDPATH, to verify their generalization capabilities. Our model, developed from a transformer pretrained on ImageNet, achieves an accuracy rate of 0.91 on the BACH dataset, 0.74 on the BRACS dataset, and 0.92 on the AIDPATH dataset. Using a model based on the prostate small and prostate medium HistoEncoder models, we achieve accuracy rates of 0.89 and 0.86, respectively. Our results suggest that pretraining on large-scale general datasets like ImageNet is advantageous. We also show the potential benefits of using domain-specific pretraining datasets, such as extensive histopathological image collections as in HistoEncoder, though not yet with clear advantages.

Dyspnea in Post-Acute COVID-19: A Multi-Parametric Cardiopulmonary Evaluation.

Post-acute COVID-19 is characterized by the persistence of dyspnea, but the pathophysiology is unclear. We evaluated the prevalence of dyspnea during follow-up and factors at admission and follow-up associated with dyspnea persistence. After five months from discharge, 225 consecutive patients hospitalized for moderate to severe COVID-19 pneumonia were assessed clinically and by laboratory tests, echocardiography, six-minute walking test (6MWT), and pulmonary function tests. Fifty-one patients reported persistent dyspnea. C-reactive protein (p = 0.025, OR 1.01 (95% CI 1.00-1.02)) at admission, longer duration of hospitalization (p = 0.005, OR 1.05 (95% CI 1.01-1.10)) and higher body mass index (p = 0.001, OR 1.15 (95% CI 1.06-1.28)) were independent predictors of dyspnea. Absolute drop in SpO2 at 6MWT (p = 0.001, OR 1.37 (95% CI 1.13-1.69)), right ventricular (RV) global longitudinal strain (p = 0.016, OR 1.12 (95% CI 1.02-1.25)) and RV global longitudinal strain/systolic pulmonary artery pressure ratio (p = 0.034, OR 0.14 (95% CI 0.02-0.86)) were independently associated with post-acute COVID-19 dyspnea. In conclusion, dyspnea is present in many patients during follow-up after hospitalization for COVID-19 pneumonia. While higher body mass index, C-reactive protein at admission, and duration of hospitalization are predictors of persistent dyspnea, desaturation at 6MWT, and echocardiographic RV dysfunction are associated with this symptom during the follow-up period.

Subclinical Myocardial Injury in Patients Recovered from COVID-19 Pneumonia: Predictors and Longitudinal Assessment.

(1) Background: Emerging data regarding patients recovered from COVID-19 are reported in the literature, but cardiac sequelae have not yet been clarified. To quickly detect any cardiac involvement at follow-up, the aims of the research were to identify: elements at admission predisposing subclinical myocardial injury at follow up; the relationship between subclinical myocardial injury and multiparametric evaluation at follow-up; and subclinical myocardial injury longitudinal evolution. (2) Methods and Results: A total of 229 consecutive patients hospitalised for moderate to severe COVID-19 pneumonia were initially enrolled, of which 225 were available for follow-up. All patients underwent a first follow-up visit, which included a clinical evaluation, a laboratory test, echocardiography, a six-minute walking test (6MWT), and a pulmonary functional test. Of the 225 patients, 43 (19%) underwent a second follow-up visit. The median time to the first follow-up after discharge was 5 months, and the median time to the second follow-up after discharge was 12 months. Left ventricular global longitudinal strain (LVGLS) and right ventricular free wall strain (RVFWS) were reduced in 36% (n = 81) and 7.2% (n = 16) of the patients, respectively, at first the follow-up visit. LVGLS impairment showed correlations with patients of male gender (p 0.008, OR 2.32 (95% CI 1.24-4.42)), the presence of at least one cardiovascular risk factor (p < 0.001, OR 6.44 (95% CI 3.07-14.9)), and final oxygen saturation (p 0.002, OR 0.99 (95% CI 0.98-1)) for the 6MWTs. Subclinical myocardial dysfunction had not significantly improved at the 12-month follow-ups. (3) Conclusions: in patients recovered from COVID-19 pneumonia, left ventricular subclinical myocardial injury was related to cardiovascular risk factors and appeared stable during follow-up.

Role of Cardiac Imaging Modalities in the Evaluation of COVID-19-Related Cardiomyopathy.

Cardiac involvement has been described during the course of SARS-CoV-2 disease (COVID-19), with different manifestations. Several series have reported only increased cardiac troponin without ventricular dysfunction, others the acute development of left or right ventricular dysfunction, and others myocarditis. Ventricular dysfunction can be of varying degrees and may recover completely in some cases. Generally, conventional echocardiography is used as a first approach to evaluate cardiac dysfunction in patients with COVID-19, but, in some cases, this approach may be silent and more advanced cardiac imaging techniques, such as myocardial strain imaging or cardiac magnetic resonance, are necessary to document alterations in cardiac structure or function. In this review we sought to discuss the information provided by different cardiac imaging techniques in patients with COVID-19, both in the acute phase of the disease and after discharge from hospital, and their diagnostic and prognostic role. We also aimed at verifying whether a specific form of cardiac disease due to the SARS-CoV-2 can be identified.

Visual versus auditory Simon effect: A behavioural and physiological investigation.

This study investigated whether the visual and auditory Simon effects could be accounted for by the same mechanism. In a single experiment, we performed a detailed comparison of the visual and the auditory Simon effects arising in behavioural responses and in pupil dilation, a psychophysiological measure considered as a marker of the cognitive effort induced by conflict processing. To address our question, we performed sequential and distributional analyses on both reaction times and pupil dilation. Results confirmed that the mechanisms underlying the visual and auditory Simon effects are functionally equivalent in terms of the interaction between unconditional and conditional response processes. The two modalities, however, differ with respect to the strength of their activation and inhibition. Importantly, pupillary data mirrored the pattern observed in behavioural data for both tasks, adding physiological evidence to the current literature on the processing of visual and auditory information in a conflict task.

SlideJ: An ImageJ plugin for automated processing of whole slide images.

The digital slide, or Whole Slide Image, is a digital image, acquired with specific scanners, that represents a complete tissue sample or cytological specimen at microscopic level. While Whole Slide image analysis is recognized among the most interesting opportunities, the typical size of such images-up to Gpixels- can be very demanding in terms of memory requirements. Thus, while algorithms and tools for processing and analysis of single microscopic field images are available, Whole Slide images size makes the direct use of such tools prohibitive or impossible. In this work a plugin for ImageJ, named SlideJ, is proposed with the objective to seamlessly extend the application of image analysis algorithms implemented in ImageJ for single microscopic field images to a whole digital slide analysis. The plugin has been complemented by examples of macro in the ImageJ scripting language to demonstrate its use in concrete situations.

The Simon Effect With Saccadic Eye Movements.

In the Simon effect performance is faster and more accurate when the task-irrelevant spatial dimension of the stimulus corresponds to the location of the response, compared to when they do not correspond. In the prosaccade-antisaccade effect the latencies of saccades away from the stimulus location (i.e., antisaccades) are slower than the latencies of saccades toward the stimulus location (i.e., prosaccades). Because these two effects share a similar basis, the study of the Simon effect with saccadic eye movements needs to be decoupled from the prosaccade-antisaccade effect. A standard Simon task (Experiment 1) and a Simon task in which a distractor stimulus was also presented (Experiment 2) were implemented. In Experiment 1, results showed an effect likely attributable to the sum of the Simon effect and the prosaccade-antisaccade effect. In Experiment 2, in which the difference between the prosaccade and antisaccade was eliminated, only a Simon effect, cognitive in nature, manifested itself.

Novelty and emotion: Pupillary and cortical responses during viewing of natural scenes.

Given the remarkable similarities in the antecedent conditions-stimulus motivational relevance and novelty (i.e., probability of occurrence)-that elicit amplitude modulation of the late positive potential (LPP) and the pupillary dilation response, the present study examines whether these two indexes of orienting response reflect common processes that are responsible for modulatory patterns in motivationally relevant contexts. In the present study, the LPP and the pupillary dilation response were co-registered in a free-picture viewing context in which stimulus novelty was manipulated through repeated presentation of the same picture exemplar. More specifically, pictures depicting both emotional and neutral contents could be novel, that is never seen before in the course of the study, or repeated 4-8 times in a row (i.e., massed repetition). Results showed that, despite massed repetitions, the late positive potential amplitude continued to be highly modulated by picture content, whereas affective modulation of pupil dilation decreased with picture repetition. These findings indicate that, although the LPP and pupil dilation are similarly affected by motivational relevance during the viewing of novel pictures, they differ when pictures are highly familiar, possibly reflecting different functional meanings in the context of the orienting response.

The modality-switch effect: visually and aurally presented prime sentences activate our senses.

Verifying different sensory modality properties for concepts results in a processing cost known as the modality-switch effect. It has been argued that this cognitive cost is the result of a perceptual simulation. This paper extends this argument and reports an experiment investigating whether the effect is the result of an activation of sensory information which can also be triggered by perceptual linguistically described stimuli. Participants were first exposed to a prime sentence describing a light or a sound's perceptual property (e.g., "The light is flickering", "The sound is echoing"), then required to perform a property-verification task on a target sentence (e.g., "Butter is yellowish", "Leaves rustle"). The content modalities of the prime and target sentences could be compatible (i.e., in the same modality: e.g., visual-visual) or not (i.e., in different modalities). Crucially, we manipulated the stimuli's presentation modality such that half of the participants was faced with written sentences while the other half was faced with aurally presented sentences. Results show a cost when two different modalities alternate, compared to when the same modality is repeated with both visual and aural stimuli presentations. This result supports the embodied and grounded cognition view which claims that conceptual knowledge is grounded into the perceptual system. Specifically, this evidence suggests that sensory modalities can be pre-activated through the simulation of either read or listened linguistic stimuli describing visual or acoustic perceptual properties.

Walking boosts your performance in making additions and subtractions.

Previous research demonstrates that the processing of spatial information and numerical magnitude are strictly interwoven. Recent studies also provide converging evidence that number processing is influenced by body movements. In the present study we further investigate this issue by focusing on whether and how motions experienced with the whole body can influence arithmetical calculations. We asked participants to make additions or subtractions while experiencing leftward and rightward motions. Data revealed the emergence of a congruency effect between the orientation inferred by the type of arithmetical calculations and the type of motions experienced along an horizontal axis.

The impact of social context and language comprehension on behaviour: a kinematic investigation.

We investigated whether and how comprehending sentences that describe a social context influences our motor behaviour. Our stimuli were sentences that referred to objects having different connotations (e.g., attractive/ugly vs. smooth/prickly) and that could be directed towards the self or towards "another person" target (e.g., "The object is ugly/smooth. Bring it to you/Give it to another person"). Participants judged whether each sentence was sensible or non-sensible by moving the mouse towards or away from their body. Mouse movements were analysed according to behavioral and kinematics parameters. In order to enhance the social meaning of the linguistic stimuli, participants performed the task either individually (Individual condition) or in a social setting, in co-presence with the experimenter. The experimenter could either act as a mere observer (Social condition) or as a confederate, interacting with participants in an off-line modality at the end of task execution (Joint condition). Results indicated that the different roles taken by the experimenter affected motor behaviour and are discussed within an embodied approach to language processing and joint actions.

Mechanisms underlying transfer of task-defined rules across feature dimensions.

The Simon effect can be reversed, favoring spatially noncorresponding responses, when people respond to stimulus colors (e.g., green) by pressing a key labeled with the alternative color (i.e., red). This Hedge and Marsh reversal is most often attributed to transfer of logical recoding rules from the color dimension to the location dimension. A recent study showed that this transfer of logical recoding rules can occur not only within a single task but also across two separate tasks that are intermixed. The present study investigated the conditions that determine the transfer of logical recoding rules across tasks. Experiment 1 examined whether it occurs in a transfer paradigm, that is when the two tasks are performed separately, but provided little support for this possibility. Experiment 2 investigated the role of task-set readiness, using a mixed-task paradigm with a predictable trials sequence, which indicated that there is no transfer of task-defined rules across tasks even when they are highly active during the Simon task. Finally, Experiments 3 and 4 used a mixed-task paradigm, where trials of the two tasks were mixed randomly and unpredictably, and manipulated the amount of feature overlap between tasks. Results indicated that task similarity is a determining factor for transfer of task-defined rules to occur. Overall, the study provides evidence that transfer of logical recoding rules tends to occur across two tasks when tasks are unpredictably intermixed and use stimuli that are highly similar and confusable.

Counting is easier while experiencing a congruent motion.

Several studies suggest that numerical and spatial representations are intrinsically linked. Recent findings demonstrate that also motor actions interact with number magnitude processing, showing a motor-to-semantic effect. The current study assesses whether calculation processes can be modulated by motions performed with the whole body. Participants were required to make additions or subtractions while performing (on-line condition) or after having experienced (off-line condition) an ascending or descending motion through a passive (i.e., taking the elevator) or an active (i.e., taking the stairs) mode. Results show a congruency effect between the type of calculation and the direction of the motion depending on: a) the off-line or on-line condition, b) the passive or active mode and c) the real or imagined task. Implications of the results for an embodied and grounded perspective view will be discussed.

Self, others, objects: how this triadic interaction modulates our behavior.

Two experiments investigated whether the triadic interaction between objects, ourselves and other persons modulates motor system activation during language comprehension. Participants were faced with sentences formed by a descriptive part referring to a positive or negative emotively connoted object and an action part composed of an imperative verb implying a motion toward the self or toward other persons (e.g., "The object is attractive/ugly. Bring it toward you/Give it to another person/Give it to a friend"). Participants judged whether each sentence was sensible or not by moving the mouse toward or away from their body. Findings showed that the simulation of a social context influenced both (1) the motor system and (2) the coding of stimulus valence. Implications of the results for theories of embodied and social cognition are discussed.

Is access to the body structural description sensitive to a body part's significance for action and cognition? A study of the sidedness effect using feet.

There is evidence suggesting that viewing hands triggers automatic access to the Body Structural Description, a visual-spatial representation of human body parts configuration. Hands, however, have a special representational status within the brain because of their significance for action and cognition. We tested whether feet, less important in gestural and object-directed action, would similarly show automatic access to the Body Structural Description. Positive evidence of that would be finding a Sidedness effect (Ottoboni et al. J Exp Psychol Hum Percept Perform 31:778-789, 2005), a Simon-like paradigm previously used to study automatic hand recognition. This effect demonstrates that processing hands generates spatial codes corresponding to the side of the body on which the hand would be located within the Body Structural Description map. Feet were shown with toes pointing upwards (Experiment 1), without any connection to the ankle and the leg (Experiment 2) and with toes pointing downwards (Experiment 3). Results revealed a Sidedness effect in both Experiments 1 and 3: spatial compatibility occurred according to the side of the body that each foot would assume within the Body Structural Description. In Experiment 2, as already found in stimuli similarly featured, no effect emerged, due to the lack of the necessary anatomical links connecting the foot to a body. Results suggest that body parts with variable degrees of significance for action and cognition can access automatically the Body Structural Description hence reinforcing the hypothesis of its pure visuo-spatial nature.

Influence of temporal overlap on time course of the Simon effect.

Two experiments are reported in which we manipulated relevant and irrelevant stimulus dimensions to assess whether an increase in temporal overlap would influence the time course of a "standard" Simon effect (obtained when visual stimuli are presented on the left/right of the screen and left/right responses are performed with uncrossed hands). This procedure is new in two ways: First, the manipulations were intended to reduce, instead of increase, the distance between conditional and unconditional response-activation processes. Second, we manipulated the relevant and irrelevant stimulus dimensions in a manner that did not vary stimulus onset asynchronies, precues, or go/no go trials, or alter the stimulus quality. Results were consistent with the hypothesis that when the two response processes are shifted closer to each other, the Simon effect would be sustained across time, instead of decreasing as typically found. These findings are discussed in line with the temporal overlap hypothesis and with an automatic activation account.

Reaction time distribution analysis of spatial correspondence effects.

Since 1994, group reaction time (RT) distribution analyses of spatial correspondence effects have been used to evaluate the dynamics of the spatial Simon effect, a benefit of correspondence of stimulus location information with response location for tasks in which stimulus location is irrelevant. We review the history and justification for analyzing group RT distributions and clarify which conditions result in the Simon effect decreasing across the distribution and which lead to flat or increasing functions. Although the standard left-right Simon effect typically yields a function for which the effect decreases as RT increases, in most other task variations, the Simon effect remains stable or increases across the RT distribution. Studies that have used other means of evaluating the temporal dynamics of the Simon effect provide converging evidence that the changes in the Simon effect across the distribution are due mainly to temporal activation properties, an issue that has been a matter of some dispute.

Attention control and susceptibility to hypnosis.

The present work aimed at assessing whether the interference exerted by task-irrelevant spatial information is comparable in high- and low-susceptible individuals and whether it may be eliminated by means of a specific posthypnotic suggestion. To this purpose high- and low-susceptible participants were tested using a Simon-like interference task after the administration of a suggestion aimed at preventing the processing of the irrelevant spatial information conveyed by the stimuli. The suggestion could be administered either in the absence or following a standard hypnotic induction. We showed that, outside from the hypnotic context, the Simon effect was similar in high and low-susceptible participants and it was significantly reduced following the posthypnotic suggestion in high-susceptible participants only. These results show that a specific posthypnotic suggestion can alter information processing in high-susceptible individuals and reduce the interfering effect exerted by arrow stimuli.

Emotive concept nouns and motor responses: attraction or repulsion?

We carried out three experiments aimed at testing whether hand posture affects the compatibility effect that Chen and Bargh (1999) found between a word's emotional connotation and arm movement direction. In the present study, participants responded by pressing two buttons: one placed near their body, the other far away. In Experiment 1, in which they pressed the response button with their hand open, RTs were shorter when participants pressed the far button for positive words and the near button for negative words, as if they simulated reaching for something good and avoiding something bad. However, in Experiments 2 and 3, in which participants pressed the response button with a tennis ball in their hand, RTs were shorter when participants pressed the near button for positive words and the far button for negative words, as if they simulated drawing a good thing closer and pushing a bad thing away. Results are discussed within the framework of theories on concept grounding in emotion and action systems.