Removal of smoke effects in laparoscopic surgery via adversarial neural network and the dark channel prior.

BACKGROUND: In laparoscopic surgery, image quality can be severely degraded by surgical smoke caused by the use of tissue dissection tools that reduce the visibility of the observed organs and tissues. OBJECTIVE: Improve visibility in laparoscopic surgery by combining image processing techniques based on classical methods and artificial intelligence. METHOD: Development of a hybrid approach to eliminating the effects of surgical smoke, based on the combination of the dark channel prior (DCP) method and a pixel-to-pixel neural network architecture known as a generative adversarial network (GAN). RESULTS: Experimental results have shown that the proposed method achieves better performance than individual DCP and GAN results in terms of restoration quality, obtaining (according to PSNR and SSIM index metrics) better results than some related state-of-the-art methods. CONCLUSIONS: The proposed approach decreases the risks and time of laparoscopic surgery because once the network is trained, the system can improve real-time visibility.

ANTECEDENTES: Durante la cirugía laparoscópica, la calidad de la imagen puede verse gravemente degradada por el humo quirúrgico causado por el uso de herramientas de disección de tejidos que reducen la visibilidad de los órganos y tejidos. OBJETIVO: Mejorar la visibilidad en cirugía laparoscópica mediante la combinación de técnicas de procesamiento de imágenes basadas en técnicas clásicas e inteligencia artificial. MÉTODO: Desarrollo de un enfoque híbrido para la eliminación de los efectos del humo quirúrgico, basado en la combinación del método del principio del canal oscuro (DCP, dark channel prior) y una arquitectura de red neuronal píxel a píxel conocida como red antagónica generativa (GAN, generative adversial network). RESULTADOS: Los resultados experimentales han demostrado que el método propuesto logra un mejor rendimiento que los resultados individuales de DCP y GAN en cuanto a calidad de la restauración, obteniendo (según las métricas de la proporción máxima de señal a ruido [PSNR, Peak Signal-to-Noise Ratio] y el índice de similitud estructural [SSIM, Structural Similarity Index]) mejores resultados que otros métodos relacionados. CONCLUSIONES: El enfoque propuesto disminuye los riesgos y el tiempo de la cirugía laparoscópica, ya que una vez que la red está correctamente entrenada, el sistema puede mejorar la visibilidad en tiempo real.

A New Approach for Motor Imagery Classification Based on Sorted Blind Source Separation, Continuous Wavelet Transform, and Convolutional Neural Network.

Brain-Computer Interfaces (BCI) are systems that allow the interaction of people and devices on the grounds of brain activity. The noninvasive and most viable way to obtain such information is by using electroencephalography (EEG). However, these signals have a low signal-to-noise ratio, as well as a low spatial resolution. This work proposes a new method built from the combination of a Blind Source Separation (BSS) to obtain estimated independent components, a 2D representation of these component signals using the Continuous Wavelet Transform (CWT), and a classification stage using a Convolutional Neural Network (CNN) approach. A criterion based on the spectral correlation with a Movement Related Independent Component (MRIC) is used to sort the estimated sources by BSS, thus reducing the spatial variance. The experimental results of 94.66% using a k-fold cross validation are competitive with techniques recently reported in the state-of-the-art.