Min Xian, Ph.D.
Assistant Professor
Department of Computer Science
University of Idaho

1776 Science Center Drive
Idaho Falls, ID 83402
Office: TAB 309 | +1-208-757-5425
Email: mxian@uidaho.edu

We are transforming the world by developing groundbreaking AI/ML methodologies that enable future dependable, trustworthy, ethical, and efficient machine intelligence. 

Join us:

  Recent Research Highlights [publication list]
AI for Clean Energy Future
kewords: AI, Machine Learning, Advanced Materials Characterization

"Developing the nuclear power systems of the future requires innovative thinking and new approaches to solving complex challenges. For the first time, a team of Idaho National Laboratory (INL) and University of Idaho researchers has successfully applied machine learning to characterizing the microstructure of metallic nuclear fuel, the fine details only visible under powerful magnification. The data collected through this technique will be used by engineers to predict fuel performance more accurately as they develop fuel for the next generation of nuclear power reactors. "

[AI for a clean energy future: Researchers use machine learning for advanced fuel development - INL]

MARST/MIRST-DM: Multi-Instance RST with Drop-Max Layer for Robust Classification
keywords: Adversarial robustness, Robust self-training, breast cancer detection

Robust self-training (RST) can augment the adversarial robustness of image classification models without significantly sacrificing models’generaliza-bility. However, RST and other state-of-the-art defense approachesfailed to pre-serve  the generalizability  and  reproduce their good adversarial robustness on small medical image sets. In this work, we propose the Multi-instance RST with drop-maxlayer, namely MIRST-DM, which involves asequence of  iteratively generated adversarial instances during training to learn smoother decision bound-aries on small datasets.The proposeddrop-max layereliminates unstable features and helps learn representations that are robust to image perturbations.The pro-posed approach was validated usinga small breast ultrasound dataset with 1,190 images. The  results demonstratethat  the  proposed  approach  achieves  state-of-the-art adversarial robustness against three prevalent attacks.


BI-RADS-Net: An Explainable Multitask Learning Approach for Cancer Diagnosis in Breast Ultrasound Images
Explainable  machine learning, computer-aided diagnosis, breast cancer, ultrasound

In healthcare, it is essential to explain the decision-making process of machine learning models to establish the trustworthiness of clinicians. This paper introduces BI-RADS-Net, a novel explainable deep learning approach for cancer detection in breast ultrasound images. The proposed approach incorporates tasks for explaining and classifying breast tumors, by learning feature representations relevant to clinical diagnosis. Explanations of the predictions (benign or malignant) are provided in terms of morphological features that are used by clinicians for diagnosis and reporting in medical practice. The employed features include the BI-RADS descriptors of shape, orientation, margin, echo pattern, and posterior features. Additionally, our approach predicts the likelihood of malignancy of the findings, which relates to the BI-RADS assessment category reported by clinicians. Experimental validation on a dataset consisting of 1,192 images indicates improved model accuracy, supported by explanations in clinical terms using the BI-RADS lexicon.


BUSIS: A Benchmark for Breast Ultrasound Image Segmentation
keywords: benchmark, breast ultrasound, segmentation

In this work, we publish a breast ultrasound benchmark for tumor segmentation. It consists of 562 images and labeled ground truths generated by four radiologists. We also evaluated the performance of fourteen state-of-the-art BUS segmentation methods quantitatively.

Background: Breast cancer is one of the leading causes of cancer death among women, and one in eight women in the United States will develop breast cancer during their lifetime. In clinical routine, tumor segmentation is a critical but quite challenging step for further cancer diagnosis and treatment planning. Many BUS segmentation approaches have been proposed in the last two decades, but the performances of most approaches have been assessed using relatively small private datasets with different quantitative metrics, which result in the discrepancy in performance comparison. Therefore, there is a pressing need for building a benchmark to compare existing methods using a public dataset objectively, and to determine the performance of the best breast tumor segmentation algorithm available today and to investigate what segmentation strategies are valuable in clinical practice and theoretical study.

[Download, Open Access at Heathcare]

STAN: Small Tumor-Aware Network
keywords: small tumor segmentation

Recently, deep learning-based approaches have achieved great success for biomedical image analysis, but current state-of-the-art approaches achieve poor performance for segmenting small breast tumors. In this work, we propose a novel deep neural network architecture, namely Small Tumor-Aware Network (STAN), to accurately and robustly segment breast tumors. STAN introduces two encoders to extract and fuse image context information at different scales. We validate the proposed approach and compare it to nine state-of-the-art approaches on three public breast ultrasound datasets using seven quantitative metrics. The results demonstrate that the proposed approach achieves the best overall performance and outperforms all other approaches on small tumor segmentation.


BendNet: Bending Loss Regularized Network for Nuclei Segmentation
keywords; nuclei segmentation, histopathology image analysis

Separating overlapped nuclei is a major challenge in histopathology image analysis. Recently published approaches have achieved promising overall performance on public datasets; however, their performance in segmenting overlapped nuclei are limited. To address the issue, we propose the bending loss regularized network for nuclei segmentation. The proposed approach is validated on the MoNuSeg dataset using five quantitative metrics. It outperforms six state-of-the-art approaches.

[BendNet PDF]

Attention-Enriched Deep Learning
keywords: Attention-enriched deep learning, breast tumor segmentation

This work proposes a new approach to integrating visual saliency into a deep learning model for breast tumor segmentation in ultrasound images. Visual saliency refers to image maps containing regions that are more likely to attract radiologists’ visual attention. The proposed approach introduces attention blocks into a U-Net architecture and learns feature representations that prioritize spatial regions with high saliency levels. The validation results indicate increased accuracy for tumor segmentation relative to models without salient attention layers. The approach achieved a Dice similarity coefficient (DSC) of 90.5% on a data set of 510 images. The salient attention model has the potential to enhance accuracy and robustness in processing medical images of other organs, by providing a means to incorporate task-specific knowledge into deep learning architectures.


Locations of Site Visitors