There are several different approaches used to treat prostate cancer, depending on age and general health conditions of the patient but also how severe the cancer is. To determine the latter, Gleason grading is used. The grade is determined by a pathologist, based on structures in histology samples from prostate biopsies. To determine the diagnosis, both the most common Gleason grade but also the highest Gleason grade occurring is used. Since the tumours typically split up the more malignant they are, single cells of Gleason grade 5, the highest and most malignant Gleason grade, can occur intermingled with benign tissue. Therefore, it is of great importance to fid even very small areas of the highest grade. This is what we aim to automatically do in this work. We have trained a convolutional neural network, with a ResNet design, to classify small areas of tissue in high magnification as either Gleason 5 or non-Gleason 5. The dataset used is generated from whole slide images from Skåne University Hospital, and consists in total of 19680 small images with the size 128×128 pixels in 40X. We try to make the algorithm more robust to stain variations, which is a common issue for this type of data, by using colour augmentation. The best accuracy we achieve for classification of Gleason 5 versus non-Gleason 5 images is 92%.
Prostate cancer is the most diagnosed cancer in men. The diagnosis is confirmed by pathologists based on ocular inspection of prostate biopsies in order to classify them according to Gleason score. The main goal of this paper is to automate the classification using convolutional neural networks (CNNs). The introduction of CNNs has broadened the field of pattern recognition. It replaces the classical way of designing and extracting hand-made features used for classification with the substantially different strategy of letting the computer itself decide which features are of importance.
For automated prostate cancer classification into the classes: Benign, Gleason grade 3, 4 and 5 we propose a CNN with small convolutional filters that has been trained from scratch using stochastic gradient descent with momentum. The input consists of microscopic images of haematoxylin and eosin stained tissue, the output is a coarse segmentation into regions of the four different classes. The dataset used consists of 213 images, each considered to be of one class only. Using four-fold cross-validation we obtained an error rate of 7.3%, which is significantly better than previous state of the art using the same dataset. Although the dataset was rather small, good results were obtained. From this we conclude that CNN is a promising method for this problem. Future work includes obtaining a larger dataset, which potentially could diminish the error margin.
Asphalt is made of a mixture of stones of different sizes and a binder called bitumen, the size distribution of the stones is determined by the recipe of the asphalt. One quality check of asphalt is to see if the real size distribution of asphalt samples is consistent with the recipe. This is usually done by first extracting the binder using methylenchloride and the sieving the stones and see how much that pass every sieve size. Methylenchloride is highly toxic and it is desirable to find the size distribution in some other way. In this paper we find the size distribution by slicing up the asphalt sample and using image analysis techniques to analyze the cross-sections. First the stones are segmented from the background, bitumen, and then rectangles are fit to the detected stones. We then estimate the sizes of the stones by using the width of the rectangle. The result is compared with both the recipe for the asphalt and with the result from the standard analysis method, and our method shows good correlation with those.
Ultrasound imaging of the heart is a non-invasive method widely used for different applications. One of them is
to measure the blood volume in the left ventricle at different stages of the heart cycle. This demands a proper
segmentation of the left ventricle and a (semi-) automated method would decrease intra-variability as well as
workload. This paper presents a semi-automated segmentation method that uses a region based snake. To avoid
any unwanted concavities in the segmentations due to the cardiac valve we use two anchor points in the snake
that are located to the left and to the right of the cardiac valve respectively. For the possibility of segmentations
in different stages of the heart cycle these anchor points are tracked through the cycle. This tracking is based both
on the resemblance of a region around the anchor points and a prior model of the movement in the y-direction
of the anchor points. The region based snake functional is the sum of two terms, a regularizing term and a data
term. It is our data term that is region based since it involves the integration of a two-dimensional subdomain of
the image plane. A segmentation of the left ventricle is obtained by minimizing the functional which is done by
continuously reshaping the contour until the optimal shape and size is obtained. The developed method shows