Biomedical Engineering: Applications, Basis and Communications

In the paper, we proposed a pyramid-based mass detection method based on texture analysis and neural classifier for digital mammograms. The proposed mass detection method is composed of four parts: pyramid decomposition, region of interest (ROI) selection, feature extraction and neural classifier. Based on pyramid decomposition, a coarse-to-fine approach was utilized to achieve mass detection for reducing computational complexity in the proposed scheme. For decreasing computational complexity, ROI selection where a thresholding algorithm and polynomial function fitting were to find the breast area is also exploited to remove nonbreast regions in the proposed scheme. In the texture analysis, the intensity and texture information extracted from spatial and wavelet domains are utilized to analyze each pixel within the ROI. After feature extraction, these extracted texture features are combined with a supervised neural network to detect masses in the ROI. To evaluate the performance of the proposed scheme, the mammograms of 19 patients captured in Taiwan are used for testing. The experimental result shows that ROI selection can localize breast regions well for further analysis. In addition, the average recall rate of our proposed scheme is more than 86%. Therefore, these experimental results demonstrate that the proposed pyramid-based scheme can achieve mass detection.

Breast cancer is the second most common cancer in females, after lung cancer in the world. In Taiwan, there are about 8500 female suffering from breast cancer every year. The incidence of breast cancer has exceeded cervical cancer and has become the most common female cancer. Immunohistochemistry (IHC) image is widely applied to the diagnosis of breast cancer, but it requires a great deal of manpower and time. The IHC images are scoring as {0+, 1+, 2+ and 3+} corresponding to no staining, weak, moderate and strong staining, respectively. With the growing of image processing techniques, computer-assisted technologies are the best solution to reduce the variability of pathologists evaluation and provide highly specific per-cell information. Therefore, in this paper, we proposed an automatic method to assess the grade of breast cancer in IHC images. The proposed method consists of four steps, including ROI extraction, feature extraction, feature selection and a hierarchical SVM classifier. The hierarchical SVM classifier is utilized to score the IHC images into 0+ (no staining), 1+ (weak), 2+ (moderate) and 3+ (strong staining). According to the experimental results, the proposed method can automatically and effectively asses the score of IHC images; it provides important information to help physicians treat breast cancer.

The severity and outcome of mastocytosis are known to be related to the number and size of mast cells. The standard treatment of this rare dermatological disorder is phototherapy with UVA1 irradiation. Here, the treatment effects were evaluated during the course of phototherapy on a patient with mastocytosis (urticaria pigmentosa). Photomicrographs of CD117 antibody-stained histological sections of the dermis were analyzed using image software programmed with artificial neural networks. The artificial neural networks analyzed the color feature of epidermis, dermis and mast cells. First, mast cells were isolated from epidermis and dermis, and RGB pixels were converted to HSV color space for artificial neural networks training. Secondly, morphology and connective component labeling were used in epidermal pixels classification to determine the scope of dermal areas. Thirdly, regional separation isolated the dermis for the mast cell analysis. Specifically, the number and size of mast cells were quantified. The method has good reproducibility compared with other methods which had larger variance and lower accuracy. We found also that the UVA1 phototherapy was effective as indicated by a post-treatment change in the ratio of mast cell surface to cell number. In addition, we compared similar results based on subjective visual inspection (from 10 human subjects). Both methods of analysis showed a reduction in the number and size of mast cells after phototherapy. But compared with human judgment, the computerized method was superior in accuracy, sensitivity, specificity and reproducibility of the findings.

Personal health management has been an essential issue in recent years. With the mobile information technology, various personalized applications have been developed for the user to manage his/her health anytime, anywhere. With the popularity of smartphones, such applications have been moved to smartphones due to their great computing and networking power. This paper introduces a health management application (app) run on iPhone. The app uses the QR-Code to input calorie and nutrition information of food, snacks and drinks. This is a novel way to record what the user eats/drinks. It is well known that the daily food intake of a person has a great influence on his/her health. But it has been very difficult to have such a record through manual input. Nevertheless, the data recording can be made easy with the QR-Code reading mechanism on smartphone. In addition, a user can also record his/her exercise activities through the app. Finally, a set of rules is built along with an inference engine within the app. With the personal dietary and exercise record, the app can provide personalized recommendations to the user following the rules. The app aims to provide useful recommendations to users who need to monitor his/her diet due to specific diseases or weight control. With this app, we hope to utilize the ease-of-use of the smartphone to achieve personalized health management.

In order to detect early prostate cancer, physicians usually apply the ultrasound imaging to visualize prostate for diagnosis. However, the traditional gray-scale ultrasound images that obtained from transrectal ultrasound of the prostate are often limited in diagnosing prostate cancer because of cancer heterogeneity, which makes direct diagnosing on the image difficult. Power Doppler ultrasound images improve the diagnosis by providing extra information of blood flow. Previous study has shown that increased blood inflow exists in prostate cancer. It is not difficult to differentiate the malignancy from benign prostate diseases by inspecting Doppler spectrum at bilateral neurovascular bundle sites and using resistive index. Although it is effective; however, it is hand-held, time-consuming, and only limited in a small and static region, which might not represent the real status of the whole prostate. Therefore, this study proposed an integrated dynamic approach to estimate the resistive index of blood flow in regions of interest within or all over the whole prostate. Merits of the proposed approach include the follows: resistive index can be calculated in serial prostate frames; the calculation can be performed on integral prostate region; and the proposed approach shortened the time needed to complete the calculation on whole prostate. Experimental results show that dynamic resistive index is better than conventional resistive index, and comparable with Prostate specific antigen (PSA) test in diagnosing prostate cancer, indicating high performance of the proposed approach, which also reveals feasibility and effectively of it.

Breast cancer is the most common threat to the health of women. Breast masses are usually important signs of breast cancer. Therefore, a level set method (LSM) with a shape model is proposed to segment breast masses in magnetic resonance imaging (MRI) images in this paper. Since the SM proposed by Chan and Vese does not work well on breast mass segmentation, this paper adds shape knowledge into the segmentation method. We first apply the Chan–Vese LSM to obtain a pre-segmented breast mass and then the position and size of the pre-segmented breast mass are calculated to establish the initial shape model. This paper uses dilation processing to calculate the distance to the shape model contour since it takes into consideration the need to update the level set function. Finally, the proposed method is applied to segment the breast mass in the MRI image of the breast. In order to eliminate noise interference in other regions of the breast, we also address the concept of region of interest (ROI). In the experiment, the proposed method is compared with the Chan–Vese method to prove that the proposed method can achieve better performance. The experimental results show that the breast mass can be correctly segmented by the above mechanism.

The occurrence of invasive breast cancer and lymphatic infiltration (LI) is highly relevant. Even though medical screening and imaging technology has made considerable progress today, there are many patients with metastasis phenomenon of axillary lymph gland at the first inspection of breast cancer. This is because the early signs of LI are very difficult to detect. To prevent the breast cancer cells spreading to distant organs through the lymphatic system or bloodstream, physicians usually recommend breast cancer patients for lymphadenectomy (lymph node dissecting). However, it is estimated that about 70% of patients with early breast cancer undergo unnecessary lymphadenectomy. Nowadays, LI is examined through a microscope and results are obtained by the doctor's experience and expertise. It is time-consuming, labor-intensive and prone to inter-observer discrepancy. Therefore, in this paper, we present a multistage model of automated system of lymphocyte detection. Cell nuclei, including lymphocyte, from the patients' H&E images are first segmented. Then the lymph nodes are classified morphologically with support vector machine in which the corresponding feature vectors are extracted according to scale-invariant feature transform. The experimental result demonstrates the agreement between the automated lymphocyte detection and physician manual detection. In addition to helping the early breast cancer patients avoid unnecessary lymphadenectomy surgery, the system also confirmed that it would greatly reduce the burden for physicians.

In this paper, we focus on the medical imaging segmentation techniques which are used in the study of spine diseases. In the medical reports, it is shown that common people worry more about the spine diseases caused by the disc degeneration. Because of the complex composition of the spine, which includes the spine bones, cartilage, fat, water and soft tissue, it is hard to correctly and easily find out the position of each cartilage in the spine images. This above problem always causes over-segmentation or unability to extract the cartilages. Thus, we propose an accurate and automated method to detect the abnormal disc. We combine two standard models with the threshold value to accurately identify the cartilage. Among the processing, we also solve the noising problems of spine image through morphological methods, removing the noncartilage areas using our proposed method, and find out the average height of the cartilages. Therefore, we can easily determine whether the disc is degenerated or not. In the experimental result, the segmentation accuracy of the extracted region by the proposed approach is evaluated by two criterions. The first criterion is statistical evaluation indices of image segmentation. It is evaluated by professional physician's manual segmentation, and the results show that our proposed method is easily implemented and has high accuracy, with the highest rate reaching 99.88%. The second criterion is a comparison evaluation index evaluated by our proposed system and other existence system. From this result, we know that our proposed system is better than other existence system.