Rahnamayan, Shahryar

Abstract Binary encoding of features has proven to be a versatile tool for optimizing data processing and memory efficiency in various machine learning applications. This includes deep barcoding of images for retrieval of similar cases among millions of indexed images. Despite the recent advancement in barcode generation methods, converting hyperdimensional feature vectors (e.g., deep features) to compact and discriminative binary barcodes is still an urgent necessity and widely an unresolved problem. MinMax barcoding of features is one of the efficient binarization methods, where the accuracy of the generated barcodes is inherently dependent on the feature order due to its computation process. In this paper, we propose a combinatorial evolutionary framework for optimizing barcodes. The primary goal is to determine an optimal permutation of extracted deep features, leading to more accurate barcodes. The performance of the proposed model has been assessed in downstream image retrieval tasks using a variety of datasets. This evaluation includes medical images from The Cancer Genome Atlas (TCGA), a comprehensive publicly available dataset, as well as images from a COVID-19 dataset. In addition, the model's performance was tested on diverse non-medical image collections, such as CIFAR-10 and Fashion-MNIST. Our findings demonstrate that optimizing barcodes significantly enhances retrieval accuracy across a wide range of applications, highlighting the applicability and potential of barcodes in various domains.

Abstract The exponential growth of data across various medical domains has generated a substantial demand for techniques to analyze multimodal big data. This demand is particularly pronounced in fields such as computational pathology due to the diverse nature of the tissue. Cross-modal retrieval aims to identify a common latent space where different modalities, such as image-text pairs, exhibit close alignment. The primary challenge, however, often lies in the representation of tissue features. While language models can be trained relatively easily, visual models frequently struggle due to the scarcity of labeled data. To address this issue, the innovative concept of harmonization has been introduced, extending the learning scheme distillation without supervision, known as DINO. The harmonization of scale refines the DINO paradigm through a novel patching approach, overcoming the complexities posed by gigapixel whole slide images in digital pathology. Experiments conducted on diverse datasets have demonstrated that the proposed approach significantly enhances cross-modal retrieval in tissue imaging. Moreover, it exhibits vast potential for other fields that rely on gigapixel imaging.

Abstract Appearing traces of bias in deep networks is a serious issue that can play a significant role in ethics and generalization. Recent studies report that the deep features extracted from the histopathology images of The Cancer Genome Atlas (TCGA), the largest publicly available archive of 11,000 patients covering 25 organs and 32 cancer subtypes, are surprisingly able to accurately classify the whole slide images (WSIs) based on their acquisition site. This is clear evidence that the utilized Deep Neural Networks (DNNs) unexpectedly detect the specific patterns of the source site rather than histomorphologic patterns, biased behaviour resulting in degraded generalization. This observation motivated us to propose a method to alleviate the destructive impact of hospital bias through a novel feature selection process. To this effect, we have proposed an evolutionary strategy select a small set of optimal features to not only accurately represent the histological patterns of tissue samples but also to eliminate the features leading to internal bias toward the institution. The performance of the proposed method has been assessed using the features extracted from TCGA images made available by NIH (National Institute of Health). The selected features extracted by a state-of-the-art network trained on TCGA images (i.e., the KimiaNet), considerably decreased the institutional bias. The proposed scheme is not limited to DNNs and/or specific types of biases; it can be employed to reduce various kinds of bias in a much more comprehensive range of data-driven feature extraction models. The conducted experiments, the external validation in specific, clearly demonstrate that the bias cannot be fully controlled just during training a model. Therefore, a feature selection for downstream tasks plays a crucial role in significantly reducing the bias.