In the present study, we propose a novel case-based similar image retrieval (SIR) method for hematoxylin and eosin (H&E) stained histopathological images of malignant lymphoma. When a whole slide image (WSI) is used as an input query, it is desirable to be able to retrieve similar cases by focusing on image patches in pathologically important regions such as tumor cells. To address this problem, we employ attention-based multiple instance learning, which enables us to focus on tumor-specific regions when the similarity between cases is computed. Moreover, we employ contrastive distance metric learning to incorporate immunohistochemical (IHC) staining patterns as useful supervised information for defining appropriate similarity between heterogeneous malignant lymphoma cases. In the experiment with 249 malignant lymphoma patients, we confirmed that the proposed method exhibited higher evaluation measures than the baseline case-based SIR methods. Furthermore, the subjective evaluation by pathologists revealed that our similarity measure using IHC staining patterns is appropriate for representing the similarity of H&E stained tissue images for malignant lymphoma.

We propose a new method for cancer subtype classification from histopathological images, which can automatically detect tumor-specific features in a given whole slide image (WSI). The cancer subtype should be classified by referring to a WSI, i.e., a large-sized image (typically 40,000x40,000 pixels) of an entire pathological tissue slide, which consists of cancer and non-cancer portions. One difficulty arises from the high cost associated with annotating tumor regions in WSIs. Furthermore, both global and local image features must be extracted from the WSI by changing the magnifications of the image. In addition, the image features should be stably detected against the differences of staining conditions among the hospitals/specimens. In this paper, we develop a new CNN-based cancer subtype classification method by effectively combining multiple-instance, domain adversarial, and multi-scale learning frameworks in order to overcome these practical difficulties. When the proposed method was applied to malignant lymphoma subtype classifications of 196 cases collected from multiple hospitals, the classification performance was significantly better than the standard CNN or other conventional methods, and the accuracy compared favorably with that of standard pathologists.

Analyzing 1,000 faulty 1 Mb SRAM chips that were randomly selected from a single manufacture, we found 251 stuck-at cell faults, 5 stuck-at bit-line faults, 1 stuck-at word-line fault, 46 neighborhood-pattern-sensitive faults, and other kinds of faults. Under the condition that I/sub dd/=4.5 I; temperature=70/spl deg/C, and load capacity C/sub L/=30 pF, we detected margin faults in 460 chips. Because the actual fault data for SRAM chips is rarely reported, the data in this manuscript are very useful and should be of practical importance.

To develop better ROM BIST techniques we first experimentally surveyed cell faults, word-line faults, bit-line faults, delay faults and other types of faults occurring in 1,000 faulty mask ROM chips. We found that most of the stuck-at faults were within a single mat. We then theoretically analyzed the aliasing probability for a mask ROM containing a fault or faults within a single mat. To experimentally evaluate BIST aliasing errors we implemented six MISRs on a custom board and observed actual aliasing errors. The experimentally measured aliasing probabilities agreed with the probabilities derived theoretically. No aliasing error occurred for the 16-stage, 8-input MISR.

The aliasing probability is theoretically analyzed for a mask ROM containing a word/bit-line fault or faults within a mat. Analysis of 1000 faulty mask ROM chips revealed cell faults, word-line faults, bit-line faults, delay faults and other types. For these chips, the BIST aliasing errors were experimentally examined. Six MISRs were implemented on a custom board, and aliasing errors were actually observed. The best of the six is shown to be the 16-stage, 8-input MISR with no aliasing error.>