Interpretable Artificial Intelligence: Why and When

Ghosh, Adarsh; Kandasamy, Devasenathipathy

doi:10.2214/ajr.19.22145

Cited by 41 publications

(24 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The opacity of these algorithms makes their clinical implementation a dilemma. Moreover, large amounts of standardized image data are required to develop the DL algorithm, which is difficult to acquire in the medical field (29,30). Considering these concerning issues of the DL algorithm, we elected interpretable ML methods that can take advantage of obtainable sample sizes of the data to study for the risk stratification of thyroid lesions.…”

Section: Discussionmentioning

confidence: 99%

A Comparative Analysis of Two Machine Learning-Based Diagnostic Patterns with Thyroid Imaging Reporting and Data System for Thyroid Nodules: Diagnostic Performance and Unnecessary Biopsy Rate

Zhao

Ren

Yin

et al. 2021

Thyroid

View full text Add to dashboard Cite

Background: The risk stratification system of the American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) for thyroid nodules is affected by low diagnostic specificity. Machine learning (ML) methods can optimize the diagnostic performance in medical image analysis. However, it is unknown which ML-based diagnostic pattern is more effective in improving diagnostic performance for thyroid nodules and reducing nodule biopsies. Therefore, we compared ML-assisted visual approaches and radiomics approaches with ACR TI-RADS in diagnostic performance and unnecessary fine-needle aspiration biopsy (FNAB) rate for thyroid nodules. Methods: This retrospective study evaluated a data set of ultrasound (US) and shear wave elastography (SWE) images in patients with biopsy-proven thyroid nodules (‡1 cm) from the Shanghai Tenth People's Hospital (743 nodules in 720 patients from September 2017 to January 2019) and an independent test data set from the Ma'anshan People's Hospital (106 nodules in 102 patients from February 2019 to April 2019). Six US features and five SWE parameters from the radiologists' interpretation were used for building the ML-assisted visual approaches. The radiomics features extracted from the US and SWE images were used with ML methods for developing the radiomics approaches. The diagnostic performance for differentiating thyroid nodules and the unnecessary FNAB rate of the ML-assisted visual approaches and the radiomics approaches were compared with ACR TI-RADS. Results: The ML-assisted US visual approach had the best diagnostic performance than the US radiomics approach and ACR TI-RADS (area under the curve [AUC]: 0.900 vs. 0.789 vs. 0.689 for the validation data set, 0.917 vs. 0.770 vs. 0.681 for the test data set). After adding SWE, the ML-assisted visual approach had a better diagnostic performance than US alone (AUC: 0.951 vs. 0.900 for the validation data set, 0.953 vs. 0.917 for the test data set). When applying the ML-assisted US+SWE visual approach, the unnecessary FNAB rate decreased from 30.0% to 4.5% in the validation data set and from 37.7% to 4.7% in the test data set in comparison to ACR TI-RADS. Conclusions: The ML-assisted dual modalities visual approach can assist radiologists to diagnose thyroid nodules more effectively and considerably reduce the unnecessary FNAB rate in the clinical management of thyroid nodules.

show abstract

Section: Discussionmentioning

confidence: 99%

A Comparative Analysis of Two Machine Learning-Based Diagnostic Patterns with Thyroid Imaging Reporting and Data System for Thyroid Nodules: Diagnostic Performance and Unnecessary Biopsy Rate

Zhao

Ren

Yin

et al. 2021

Thyroid

View full text Add to dashboard Cite

show abstract

“…In addition to the successful results, there are several aspects in which scGCN can be improved. First, as an artificial intelligence (AI) model, scGCN shows not only the merits of its kind, but also some limitations including the black-box nature of AI models [47][48][49], which can be addressed through downstream analysis such as differential gene identification and enrichment analysis that can ameliorate some of the problems and bring insights into the labeled cells. Second, as a graph model, improving the graph construction can further boost the model performance.…”

Section: Discussionmentioning

confidence: 99%

“…As 42 more and more single-cell data becomes available, there is an urgent need to leverage existing data 43 with the newly generated data in a reliable and reproducible way, learning from the established 44 single-cell data with well-defined labels as reference, and transferring labels to new datasets to 45 assign cell-level annotations [10,11]. However, existing datasets and new datasets are often 46 collected from different tissues and species [14,15], under various experimental conditions, 47 generated by different platforms [16,17], and in the form of different omics types [18]. Thus a 48 reliable and accurate knowledge transfer method must overcome the following challenges: 1) the 49 unique technical issues of single-cell data (e.g., dropouts and dispersion) [19][20][21][22]; 2) batch effects 50 arisen from different operators, experimental protocols [23], and technical variation (e.g., mRNA quality, pre-amplification efficiency, technical settings during data generation) [24][25][26];…”

Section: Introductionmentioning

confidence: 99%

scGCN: a Graph Convolutional Networks Algorithm for Knowledge Transfer in Single Cell Omics

Song

Zhang

2020

Preprint

View full text Add to dashboard Cite

Single-cell omics represent the fastest-growing genomics data type in the literature and the public genomics repositories. Leveraging the growing repository of labeled datasets and transferring labels from existing datasets to newly generated datasets will empower the exploration of the single-cell omics. The current label transfer methods have limited performance, largely due to the intrinsic heterogeneity and extrinsic differences between datasets. Here, we present a robust graph-based artificial intelligence model, single-cell Graph Convolutional Network (scGCN), to achieve effective knowledge transfer across disparate datasets. Benchmarked with other label transfer methods on totally 30 single cell omics datasets, scGCN has consistently demonstrated superior accuracy on leveraging cells from different tissues, platforms, and species, as well as cells profiled at different molecular layers. scGCN is implemented as an integrated workflow as a python software, which is available at https://github.com/QSong-github/scGCN.

show abstract

“…In addition to the successful results, there are several aspects that DSTG can be improved. First, as an artificial intelligence (AI) model, DSTG shows not only the merits of its kind, but also some limitations including the black-box nature of AI models [34][35][36], which can be addressed through downstream analysis that can ameliorate some of the problems and bring insights into the learned cellular compositions. Second, as a graph model, improving the built graph can further boost the model performance.…”

Section: Discussionmentioning

confidence: 99%

DSTG: Deconvoluting Spatial Transcriptomics Data through Graph-based Artificial Intelligence

Song

2020

Preprint

View full text Add to dashboard Cite

Recent development of spatial transcriptomics (ST) is capable of associating spatial information at different spots in the tissue section with RNA abundance of cells within each spot, which is particularly important to understand tissue cytoarchitectures and functions. However, for such ST data, since a spot is usually larger than an individual cell, gene expressions measured at each spot are from a mixture of cells with heterogenous cell types. Therefore, ST data at each spot needs to be disentangled so as to reveal the cell compositions at that spatial spot. In this study, we propose a novel method, named DSTG, to accurately deconvolute the observed gene expressions at each spot and recover its cell constitutions, thus achieve high-level segmentation and reveal spatial architecture of cellular heterogeneity within tissues. DSTG not only demonstrates superior performance on synthetic spatial data generated from different protocols, but also effectively identifies spatial compositions of cells in mouse cortex layer, hippocampus slice, and pancreatic tumor tissues. In conclusion, DSTG accurately uncovers the cell states and subpopulations based on spatial localization.

show abstract

Interpretable Artificial Intelligence: Why and When

Cited by 41 publications

References 4 publications

A Comparative Analysis of Two Machine Learning-Based Diagnostic Patterns with Thyroid Imaging Reporting and Data System for Thyroid Nodules: Diagnostic Performance and Unnecessary Biopsy Rate

A Comparative Analysis of Two Machine Learning-Based Diagnostic Patterns with Thyroid Imaging Reporting and Data System for Thyroid Nodules: Diagnostic Performance and Unnecessary Biopsy Rate

scGCN: a Graph Convolutional Networks Algorithm for Knowledge Transfer in Single Cell Omics

DSTG: Deconvoluting Spatial Transcriptomics Data through Graph-based Artificial Intelligence

Contact Info

Product

Resources

About