Search

Identification of brain tumors border and determination of their possible pathology in MR images is an important step in pre-operation analyzing of this serious medical condition. Manual segmentation and classification of brain tumors could be challenge full in neurosurgical practice because of vast differences between brain tumors characteristic such as shape, border irregularity, consistency and etc. as well as interobserver variations. To solve this problem, some automatic methods have been proposed for brain tumors segmentation or classification during recent years, but an intelligence-based method for simultaneous identification of tumor type and tumor border in MR images has not proposed till now. Here, we have planned a unique automatic model includes a common encoder for feature representation, one decoder for segmentation and a multi-layer perceptron for classification of three common primary brain tumors (meningiomas, gliomas and pituitary adenomas) in brain MR images. The proposed model was examined on a brain tumor images dataset and the output were evaluated in both multi-task and single-task learning model. The multi-task learning model gains significant improvement in simultaneous classification and segmentation of brain tumors with promising accuracy of 97% for each task. So, this model could serve as a primary screening tool for early diagnosis of common primary brain tumors in general practice with a high success rate.

craniosynostosis (CSO) is a congenital disorder resulting from early closure of cranial sutures in newborns, while could cause significant cosmetic and neurodevelopmental problems. As a standard method, different craniometric indices are measured directly from child head or from their 3D CT scan of skull for diagnosis or in post-operative follow-up period. We propose a novel telehealth-compatible deep learning neural network-based method for identifying different craniometric indices in non-syndromic CSO patients 2D photographic data.