Skin cancer is most likely to disseminate to other parts of the human body if it is not detected and treated in a timely manner. Consequently, early detection is crucial for prompt and effective treatment. The evident similarity between skin conditions has complicated medical diagnosis. Although melanoma is the most well-known form of skin cancer, other diseases have caused a significant number of deaths in recent years. Recent advances in computerized methods for mak- ing these diagnoses have made them more accurate and quicker, which is very encouraging. The absence of sizable datasets is one of the greatest obstacles to developing a reliable automatic classification system. For this purpose, the ISIC Skin Lesion Classification Challenge provided 25331 images from eight distinct classifications. This paper presents a CNN-based deep learning model for cuta- neous cancer detection. Its primary objective is to categorize skin lesions based on Dermoscope images. With a sensitivity of 55.32%, a specificity of 88.92%, an accuracy of 90.18%, a precision of 91.01%, a dice accuracy of 90.80%, and a jac- card accuracy of 83.37%, our method has yielded results that are significantly superior to those of existing methods. The proposed technique is significantly superior to the existing methods for recognizing and categorizing skin diseases.
25 January 2018 in Medical & Biological Engineering & Computing
Breast cancer is one of the major causes of death in women. Computer Aided Diagnosis (CAD) systems are being developed to assist radiologists in early diagnosis. Micro-calcifications can be an early symptom of breast cancer. Besides detection, classification of micro-calcification as benign or malignant is essential in a complete CAD system. We have developed a novel method for the classification of benign and malignant micro-calcification using an improved Fisher Linear Discriminant Analysis (LDA) approach for the linear transformation of segmented micro-calcification data in combination with a Support Vector Machine (SVM) variant to classify between the two classes. The results indicate an average accuracy equal to 96% which is comparable to state-of-the art methods in the literature.
In recent years, an increased interest has been seen in the area of medical image processing and, as a consequence, Computer Aided Diagnostic (CAD) systems. The basic purpose of CAD systems is to assist doctors in the process of diagnosis. CAD systems, however, are quite expensive, especially, in most of the developing countries. Our focus is on developing a low-cost CAD system. Today, most of the CAD systems regarding mammogram classification target automatic detection of calcification and abnormal mass. Calcification normally indicates an early symptom of breast cancer if it appears as a small size bright spot in a mammogram image. Based on the observation that calcification appears as small bright spots on a mammogram image, we propose a new scale-specific blob detection technique in which the scale is selected through supervised learning. By computing energy for each pixel at two different scales, a new feature “Ratio Energy” is introduced for efficient blob detection. Due to the imposed simplicity of the feature and post processing, the running time of our algorithm is linear with respect to image size. Two major types of calcification, microcalcification and macrocalcification have been identified and highlighted by drawing a circular boundary outside the area that contains calcification. Results are quite visible and satisfactory, and the radiologists can easily view results through the final detected boundary. CAD systems are designed to help radiologists in verifying their diagnostics. A new way of identifying calcification is proposed based on the property that microcalcification is small in size and appears in clusters. Results are quite visible and encouraging, and can assist radiologists in early detection of breast cancer.
Lung cancer is the world’s second-largest cause of cancer mortality. Patients’ lives can be saved if this malignancy is detected early. Doctors, however, encounter difficulties in detecting cancer in computed tomography (CT) images. In recent years, significant research has been devoted to producing automated lung nodule detection methods that can help radiologists. Most of them use only the lung window in their analysis and generally do not consider the mediastinal windows, which, according to recent research, carry important information. In this paper, we propose a simple yet effective algorithm to analyze multi-window CT images for lung nodules. The algorithm works in three steps. First, the CT image is preprocessed to suppress any noise and improve the image quality. Second, the lungs are extracted from the preprocessed image. Based on the histogram analysis of the lung windows, we propose a multi-Otsu-based approach for lung segmentation in lung windows. The case of mediastinal windows is rather difficult due to irregular patterns in the histograms. To this end, we propose a global–local-mean-based thresholding technique for lung detection. In the final step, the nodule candidates are extracted from the segmented lungs using simple intensity-based thresholding. The radius of the extracted objects is computed to separate the nodule from the bronchioles and blood vessels. The proposed algorithm is evaluated on the benchmark LUNA16 dataset and achieves accuracy of over 94% for lung tumor detection, surpassing that of existing similar methods.