The objective of this research is to build a "Whole Slide Images" classification system using Convolutional Neural Network (CNN). This system is capable of classifying Thyroid tumors into three types: Follicular adenoma, follicular carcinoma, and papillary carcinoma. Furthermore, the cascaded CNN technique is additionally employed to classify the classified follicular carcinoma into four subclasses: follicular carcinoma, papillary follicular variant, well-differentiated follicular carcinoma, and Poorlydifferentiated follicular carcinoma. Results of the proposed CNN architecture showed effective classification of Thyroid carcinoma in the whole slide images with an overall accuracy of 94.69%. In the first classification stage, the images are classified into either one of three main types with an overall accuracy of 98.74%, while in the second classification stage, using the cascaded CNN, accuracy was 95.90% for further subclassification into four sub-classes. Our cascaded CNN outperformed the accuracy of other studies due to splitting classification process of the thyroid into two stages which reduces the number of classes in each stage.INDEX TERMS Whole slide images, Convolutional neural network, Thyroid carcinoma, Data augmentation, Histopathology whole slide imaging.

No abstract

Rehabilitation assistive devices for head/neck pain treatment cannot allow dynamic changes in position and orientation of the head/neck. Moreover, such devices can neither be used simultaneously nor can they assess the patients’ head/neck conditions. This paper aims at designing and implementing a novel dynamic head/neck brace that provides static and dynamic support and/or traction at symmetric and asymmetric positions. This device also provides assessments of the head/neck stiffness for the purpose of fulfilling diagnoses of the head/neck disorders. The device was used and evaluated for its range of motion and its symmetric traction capability using two control modalities. In addition, it was also evaluated in determining the stiffness of the head/neck throughout a simulating mechanical model involved in a set of springs. The device could apply right/left lateral bending to the head/neck ranged −6.97 ± 0.01° to 7.02 ± 0.01° with accuracies of 99.89% and 99.48%, and flexion/extension ranged −8.10 ± 0.02° to 8.12 ± 0.01° with accuracies of 99.57% and 99.42%, respectively, throughout a traction phase of 20 mm. The practical measurements through the symmetric traction tests showed some deviations as compared to that being calculated. Such deviations were greater in flexion/extension rather than the right/left lateral bending. The mean of the obtained error was less than 0.34° for all situations of tests. The accuracies of stiffness measurement of the mechanical model were 99.78% and 99.96%, respectively, throughout performing stair and step tests. The paper presented a novel design of a dynamic head/neck brace that provides support and/or traction to any head/neck positions and capable of evaluating the head/neck stiffness during cervical traction.

No abstract