A Pathology-Based Machine Learning Method to Assist in Epithelial Dysplasia Diagnosis

The Epithelial Dysplasia (ED) is a tissue alteration commonly present in lesions preceding oral cancer, being its presence one of the most important factors in the progression toward carcinoma. This study proposes a method to design a low computational cost classification system to support the detection of dysplastic epithelia, contributing to reduce the variability of pathologist assessments. We employ a multilayer artificial neural network (MLP-ANN) and defining the regions of the epithelium to be assessed based on the knowledge of the pathologist. The performance of the proposed solution was statistically evaluated. The implemented MLP-ANN presented an average accuracy of 87%, with a variability much inferior to that obtained from three trained evaluators. Moreover, the proposed solution led to results which are very close to those obtained using a convolutional neural network (CNN) implemented by transfer learning, with 100 times less computational complexity. In conclusion, our results show that a simple neural network structure can lead to a performance equivalent to that of much more complex structures, which are routinely used in the literature.

Robust parameter design for Wiener-based binaural noise reduction methods in hearing aids

This work presents a method for designing the weighting parameter required by Wiener-based binaural noise reduction methods. This parameter establishes the desired tradeoff between noise reduction and binaural cue preservation in hearing aid applications. The proposed strategy was specially derived for the preservation of interaural level difference, interaural time difference and interaural coherence binaural cues. It is defined as a function of the average input noise power at the microphones, providing robustness against the influence of joint changes in noise and speech power (Lombard effect), as well as to signal to noise ratio (SNR) variations. A theoretical framework, based on the mathematical definition of the homogeneity degree, is presented and applied to a generic augmented Wiener-based cost function. The theoretical insights obtained are supported bycomputational simulations and psychoacoustic experiments using the multichannel Wiener filter with interaural transfer function preservation technique (MWF-ITF), as a case study. Statistical analysis indicates that the proposed dynamic structure for the weighting parameter and the design method of its fixed part provide significant robustness against changes in the original binaural cues of both speech and residual noise, at the cost of a small decrease in the noise reduction performance, as compared to the use of a purely fixed weighting parameter.