Search for: All records

Deep learning models have strong potential for automating breast ultrasound (BUS) image classification to support early cancer detection. However, their vulnerability to small input perturbations poses a challenge for clinical reliability. This study examines how minimal pixel-level changes affect classification performance and predictive uncertainty, using the BUSI dataset and a ResNet-50 classifier. Two perturbation types are evaluated: (1) adversarial perturbations via the One Pixel Attack and (2) non-adversarial, device-related noise simulated by setting a single pixel to black. Robustness is assessed alongside uncertainty estimation using Monte Carlo Dropout, with metrics including Expected Kullback–Leibler divergence (EKL), Predictive Variance (PV), and Mutual Information (MI) for epistemic uncertainty, and Maximum Class Probability (MP) for aleatoric uncertainty. Both perturbations reduced accuracy, producing 17 and 29 “fooled” test samples, defined as cases classified correctly before but incorrectly after perturbation, for the adversarial and non-adversarial settings, respectively. Samples that remained correct are referred to as “unfooled.” Across all metrics, uncertainty increased after perturbation for both groups, and fooled samples had higher uncertainty than unfooled samples even before perturbation. We also identify spatially localized “uncertainty-decreasing” regions, where individual single-pixel blackouts both flipped predictions and reduced uncertainty, creating overconfident errors. These regions represent high-risk vulnerabilities that could be exploited in adversarial attacks or addressed through targeted robustness training and uncertainty-aware safeguards. Overall, combining perturbation analysis with uncertainty quantification provides valuable insights into model weaknesses and can inform the design of safer, more reliable AI systems for BUS diagnosis. 

Breast cancer is the leading cancer affecting women globally. Despite deep learning models making significant strides in diagnosing and treating this disease, ensuring fair outcomes across diverse populations presents a challenge, particularly when certain demographic groups are underrepresented in training datasets. Addressing the fairness of AI models across varied demographic backgrounds is crucial. This study analyzes demographic representation within the publicly accessible Emory Breast Imaging Dataset (EMBED), which includes de-identified mammography and clinical data. We spotlight the data disparities among racial and ethnic groups and assess the biases in mammography image classification models trained on this dataset, specifically ResNet-50 and Swin Transformer V2. Our evaluation of classification accuracies across these groups reveals significant variations in model performance, highlighting concerns regarding the fairness of AI diagnostic tools. This paper emphasizes the imperative need for fairness in AI and suggests directions for future research aimed at increasing the inclusiveness and dependability of these technologies in healthcare settings. Code is available at: https://github.com/kuanhuang0624/EMBEDFairModels. 

Tandem mass spectrometry (MS/MS) is a widely used technology for identifying metabolites. De novo metabolite identification is an identification strategy that does not refer to any spectral or metabolite database. However, this strategy is time-consuming and cannot meet the need for high-throughput metabolite identification. Böcker et al. converted the de novo identification problem into the maximum colorful subtree (MCS) problem. Unfortunately, the MCS problem is NPhard, which indicates there are no existing efficient exact algorithms. To address this issue, we propose to apply quantum computing to accelerate metabolite identification. Quantum computing performs computations on quantum computers. The recent progress in this area has brought the hope of making some computationally intractable areas trackable, although there are still no general approaches to converting regular computer algorithms into quantum algorithms. Specifically, there is no efficient quantum algorithm for the MCS problem. The MCS problem can be considered as the combination of many maximum spanning tree problems that can be converted into minimum spanning tree problems. This work applies a quantum algorithm designed for the minimum spanning problem to speed up de novo metabolite identification. The possible strategy for further improving the performance is also briefly discussed. 

Tandem mass spectrometry (MS/MS) is a popular technology for identifying small molecules involved in metabolism, better known as metabolites. Coupled with liquid chromatography (LC), LC-MS/MS instruments first separate, ionize, and fragment metabolites, then measure mass-to-charge ratios (m/z) and intensities of metabolite fragments. Understanding metabolite fragmentation is crucial to develop computational tools for identifying metabolites based on this spectroscopic data. Metabolite fragmentation patterns have large variations making it especially difficult for computer scientists to design and implement metabolite identification approaches. To address this interdisciplinary challenge, this article presents FragView, a web-based application providing the web service for visualizing metabolite fragmentation. Users can break chemical bonds to produce metabolite fragments and export 3D fragment structures for 3D printing. Developing FragView is an opportunity for exposing student participants to this interdisciplinary bioinformatics project. This paper summarizes the experience of training student participants in bootcamps and designing the implementation plan based on student backgrounds. Students were exposed to project meeting discussions on coding and raw data visualization and visited a lab with an LC-MS/MS instrument. FragView is an open source, freely accessible tool, released under the GPLv3 license. We will continue to improve and update FragView in the future based on feedback.