Knowledge Distillation in Video-Based Human Action Recognition: An Intuitive Approach to Efficient and Flexible Model Training.

Training a model to recognize human actions in videos is computationally intensive. While modern strategies employ transfer learning methods to make the process more efficient, they still face challenges regarding flexibility and efficiency. Existing solutions are limited in functionality and rely heavily on pretrained architectures, which can restrict their applicability to diverse scenarios. Our work explores knowledge distillation (KD) for enhancing the training of self-supervised video models in three aspects: improving classification accuracy, accelerating model convergence, and increasing model flexibility under regular and limited-data scenarios. We tested our method on the UCF101 dataset using differently balanced proportions: 100%, 50%, 25%, and 2%. We found that using knowledge distillation to guide the model's training outperforms traditional training without affecting the classification accuracy and while reducing the convergence rate of model training in standard settings and a data-scarce environment. Additionally, knowledge distillation enables cross-architecture flexibility, allowing model customization for various applications: from resource-limited to high-performance scenarios.

A Rare Complication of Henoch-Schönlein Purpura/IgA Vasculitis in an Adult Woman After COVID-19 Infection.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection can affect multiple organs due to activation of an inflammatory response. One of the key components of this response is the activation of immunoglobulin A (IgA), thus causing endothelial injury and inflammation. Henoch-Schönlein purpura (HSP) has been rarely reported in adult patients as a complication of the coronavirus disease 2019 (COVID-19) infection. In this report, we present a case of HSP occurring one week after the diagnosis of COVID-19 in a 23-year-old woman. Her symptoms included nausea, vomiting, diffused abdominal pain, joint pain, hematuria, and palpable purpura of the lower extremities. She was treated with intravenous methylprednisolone sodium succinate, followed by oral prednisone therapy, which resulted in clinical improvement, including resolution of abdominal and joint pain as well as skin rashes, without remaining renal complication.

Real-time instance segmentation of surgical instruments using attention and multi-scale feature fusion.

Precise instrument segmentation aids surgeons to navigate the body more easily and increases patient safety. While accurate tracking of surgical instruments in real-time plays a crucial role in minimally invasive computer-assisted surgeries, it is a challenging task to achieve, mainly due to: (1) a complex surgical environment, and (2) model design trade-off in terms of both optimal accuracy and speed. Deep learning gives us the opportunity to learn complex environment from large surgery scene environments and placements of these instruments in real world scenarios. The Robust Medical Instrument Segmentation 2019 challenge (ROBUST-MIS) provides more than 10,000 frames with surgical tools in different clinical settings. In this paper, we propose a light-weight single stage instance segmentation model complemented with a convolutional block attention module for achieving both faster and accurate inference. We further improve accuracy through data augmentation and optimal anchor localization strategies. To our knowledge, this is the first work that explicitly focuses on both real-time performance and improved accuracy. Our approach out-performed top team performances in the most recent edition of ROBUST-MIS challenge with over 44% improvement on area-based multi-instance dice metric MI_DSC and 39% on distance-based multi-instance normalized surface dice MI_NSD. We also demonstrate real-time performance (>60 frames-per-second) with different but competitive variants of our final approach.

Assessing YOLACT++ for real time and robust instance segmentation of medical instruments in endoscopic procedures.

Image-based tracking of laparoscopic instruments plays a fundamental role in computer and robotic-assisted surgeries by aiding surgeons and increasing patient safety. Computer vision contests, such as the Robust Medical Instrument Segmentation (ROBUST-MIS) Challenge, seek to encourage the development of robust models for such purposes, providing large, diverse, and high-quality datasets. To date, most of the existing models for instance segmentation of medical instruments were based on two-stage detectors, which provide robust results but are nowhere near to the real-time, running at 5 frames-per-second (fps) at most. However, for the method to be clinically applicable, a real-time capability is utmost required along with high accuracy. In this paper, we propose the addition of attention mechanisms to the YOLACT architecture to allow real-time instance segmentation of instruments with improved accuracy on the ROBUST-MIS dataset. Our proposed approach achieves competitive performance compared to the winner of the 2019 ROBUST-MIS challenge in terms of robustness scores, obtaining 0.313 ML_DSC and 0.338 MLNSD while reaching real-time performance at >45 fps.

Utilidad de las Escalas de Gradación en el Tratamiento Quirúrgico de Malformaciones Arteriovenosas Cerebrales / Usefulness of the Gradation Scales in the Surgical Treatment of Cerebral Arteriovenous Malformations

Introducción: Las malformaciones arteriovenosas (MAVs) cerebrales comprenden una compleja patología responsable de hasta el 38% de las hemorragias en pacientes de entre 15-45 años, acarreando cada episodio de sangrado un 25-50% de morbilidad y un 10-20% de mortalidad. La decisión terapéutica en un paciente con una MAV debe tener en cuenta la comparación entre los riesgos propios de la intervención y los de la historia natural de esta enfermedad. Objetivo: Evaluar la utilidad de predecir riesgo quirúrgico de diferentes escalas de gradación de MAV cerebrales según nuestra experiencia en una serie de casos. Material y Métodos: Se realizó un análisis bibliográfico de escalas de gradación de riesgo quirúrgico de MAV cerebrales utilizando como motor de búsqueda Pubmed incluyendo como palabras clave "malformación arteriovenosa cerebral" y "escala de gradación" (brain arteriovenous and malformation grading scale). Se analizaron de forma retrospectiva aquellos pacientes intervenidos quirúrgicamente por MAV en este hospital público, se las clasificó acorde a las escalas analizadas y se compararon los resultados obtenidos con los previstos en ellas. Resultados: Se analizaron 90 pacientes intervenidos quirúrgicamente por MAV, sin tratamiento coadyuvante. De forma retrospectiva se los agrupó acorde a las escalas de Spetzler Martin (SM), Spetzler-Ponce (SP) y suplementaria de Lawton. Las MAV grado 3 se subclasificaron según las escalas de Lawton y de de Oliveira. Considerando buenos resultados aquellos con Rankin modificado (mRs) igual o menor a 2. Con un rango de seguimiento de 12 a 48 meses, encontramos buenos resultados en el 100% de MAV SM grado 1, 91.7% de las grado 2, 80% en grado 3 y 42.9% en grado 4. Utilizando la escala SP, 93.7% de buenos resultados en tipo A, 80% en tipo B y 42.9% en tipo C. Subclasifican-do las MAV SM 3 acorde a las escalas de de Oliveira y Lawton, 84% de buenos resultados en el tipo 3A, 71.3% en las 3B, 92% en MAV tipo 3-, 72.1% en el tipo 3+, 60% en tipo 3. Utilizando la escala suplementaria de Lawton combinada con SM, buen resultado en 100% grados II y III, 85,7% grado IV, 87,6 grado V, 80% grado VI, 75% grado VII y 66,6% grado VIII. Conclusión: Reafirmamos en esta serie, la utilidad de estimar riesgo quirúrgico con las escalas SM, SP, y la subclasificación de las MAV grado 3 propuesta por Lawton. Y principalmente el utilizar la escala suplementaria de Lawton-Young al considerar el tratamiento quirúrgico de los pacientes con MAV que sangraron.

Introduction: Brain arteriovenous malformations (AVM) are a complex disease responsible for up to 38% of hemorrhages in patients between 15-45 years old, carrying every bleeding episode a 25-50% risk of morbidity and a 10-20% of mortality. The therapeutic decision in a patient with an AVM needs to consider both the risks of the intervention and the risks of the natural evolution of the disease. Objective: To assess the effectiveness of different AVM grading scales in predicting surgical risks according to our experience in a case series. Material and Method: a literature review of the AVM grading scales was made, through Pubmed including as key words "brain arteriovenous malformations" and "grading scale". A retrospective analysis was made of patients with AVM who were operated in our institution, they were classified according to the scales and their results were compared. Results: 90 patients were operated in our institution with AVM. Retrospectively, they were classified according to the Spetzler-Martin (SM), Spetzler-Ponce (SP), Lawton supplementary, and the sub-classifications in AVM grade 3, from Lawton and de Oliveira. Good outcome were considered when modified Rankin Scale (mRs) was equal or less than 2. The follow-up ranged from 12-48 months, having good outcome in 100% of AVM SM grade I, 91,7% grade II, 80% in grade III and 42,9% in grade IV. Using the SP scale, 93,7% of good outcome in grade A, 80% in grade B and 42,9% in grade C. In the sub-classification of AVM SM 3, we found 84% of good outcome in type 3A de Oliveira and 71,3% in type 3B. According to the Lawton scale, good outcome were found in 92% in type 3-, 72,1% in type 3+ and 60% in type 3. Using Lawton supplementary scale combined with SM, there were 100% of good outcome in grades II and III, 85,7% in grade IV, 87,6% in grade V, 80% in grade VI, 75% in grade VII, 66,6% in grade VIII. Conclusion: In our series, we reaffirm the effectiveness to predict surgical risk of the following scales: SM, SP and the Lawton's sub-classification of AVM grade 3. Specially, the use of the supplementary Lawton-Young scale in the surgical treatment of bleeding AVMs.

Effective and Generalizable Graph-Based Clustering for Faces in the Wild.

Face clustering is the task of grouping unlabeled face images according to individual identities. Several applications require this type of clustering, for instance, social media, law enforcement, and surveillance applications. In this paper, we propose an effective graph-based method for clustering faces in the wild. The proposed algorithm does not require prior knowledge of the data. This fact increases the pertinence of the proposed method near to market solutions. The experiments conducted on four well-known datasets showed that our proposal achieves state-of-the-art results, regarding the clustering performance, also showing stability for different values of the input parameter. Moreover, in these experiments, it is shown that our proposal discovers a number of identities closer to the real number existing in the data.

[Usefulness of Gradation Scales in the Surgical Treatment of Cerebral Arteriovenous Malformations]. / Utilidad de las Escalas de Gradación en el Tratamiento Quirúrgico de Malformaciones Arteriovenosas Cerebrales.

INTRODUCTION: Brain arteriovenous malformations (AVM) are a complex disease responsible for up to 38% of hemorrhages in patients between 15-45 years old, carrying every bleeding episode a 25-50% risk of morbidity and a 10-20% of mortality. The therapeutic decision in a patient with an AVM needs to consider both the risks of the intervention and the risks of the natural evolution of the disease. OBJECTIVE: To assess the effectiveness of different AVM grading scales in predicting surgical risks according to our experience in a case serie. MATERIAL AND METHOD: A literature review of the AVM grading scales was made, through Pubmed including as key words "brain arteriovenous malformations" and "grading scale". A retrospective analysis was made of patients with AVM who were operated in our institution, they were classified according to the scales and their results were compared. RESULTS: 90 patients were operated in our institution with AVM. Retrospectively, they were classified according to the Spetzler-Martin (SM), Spetzler-Ponce (SP), Lawton supplementary, and the sub-classifications in AVM grade 3, from Lawton and de Oliveira. Good outcome were considered when modified Rankin Scale (mRs) was equal or less than 2. The follow-up ranged from 12-48 months, having good outcome in 100% of AVM SM grade I, 91,7% grade II, 80% in grade III and 42,9% in grade IV. Using the SP scale, 93,7% of good outcome in grade A, 80% in grade B and 42,9% in grade C. In the sub-classification of AVM SM 3, we found 84% of good outcome in type 3A de Oliveira and 71,3% in type 3B. According to the Lawton scale, good outcome were found in 92% in type 3-, 72,1% in type 3+ and 60% in type 3. Using Lawton supplementary scale combined with SM, there were 100% of good outcome in grades II and III, 85,7% in grade IV, 87,6% in grade V, 80% in grade VI, 75% in grade VII, 66,6% in grade VIII. CONCLUSION: In our serie, we reaffirm the effectiveness to predict surgical risk of the following scales: SM, SP and the Lawton's sub-classification of AVM grade 3. Specially, the use of the supplementary Lawton-Young scale in the surgical treatment of bleeding AVMs.