Assessing the Trustworthiness of Saliency Maps for Localizing Abnormalities in Medical Imaging

Machine Learning and Deep Learning in Cardiothoracic Imaging: A Scoping Review

Machine-learning (ML) and deep-learning (DL) algorithms are part of a group of modeling algorithms that grasp the hidden patterns in data based on a training process, enabling them to extract complex information from the input data. In the past decade, these algorithms have been increasingly used for image processing, specifically in the medical domain. Cardiothoracic imaging is one of the early adopters of ML/DL research, and the COVID-19 pandemic resulted in more research focus on the feasibility and applications of ML/DL in cardiothoracic imaging. In this scoping review, we systematically searched available peer-reviewed medical literature on cardiothoracic imaging and quantitatively extracted key data elements in order to get a big picture of how ML/DL have been used in the rapidly evolving cardiothoracic imaging field. During this report, we provide insights on different applications of ML/DL and some nuances pertaining to this specific field of research. Finally, we provide general suggestions on how researchers can make their research more than just a proof-of-concept and move toward clinical adoption.

Benchmarking saliency methods for chest X-ray interpretation

Saliency methods, which produce heat maps that highlight the areas of the medical image that influence model prediction, are often presented to clinicians as an aid in diagnostic decision-making. However, rigorous investigation of the accuracy and reliability of these strategies is necessary before they are integrated into the clinical setting. In this work, we quantitatively evaluate seven saliency methods, including Grad-CAM, across multiple neural network architectures using two evaluation metrics. We establish the first human benchmark for chest X-ray segmentation in a multilabel classification set-up, and examine under what clinical conditions saliency maps might be more prone to failure in localizing important pathologies compared with a human expert benchmark. We find that (1) while Grad-CAM generally localized pathologies better than the other evaluated saliency methods, all seven performed significantly worse compared with the human benchmark, (2) the gap in localization performance between Grad-CAM and the human benchmark was largest for pathologies that were smaller in size and had shapes that were more complex, and (3) model confidence was positively correlated with Grad-CAM localization performance. Our work demonstrates that several important limitations of saliency methods must be addressed before we can rely on them for deep learning explainability in medical imaging.

CIRDataset: A large-scale Dataset for Clinically-Interpretable lung nodule Radiomics and malignancy prediction

Spiculations/lobulations, sharp/curved spikes on the surface of lung nodules, are good predictors of lung cancer malignancy and hence, are routinely assessed and reported by radiologists as part of the standardized Lung-RADS clinical scoring criteria. Given the 3D geometry of the nodule and 2D slice-by-slice assessment by radiologists, manual spiculation/lobulation annotation is a tedious task and thus no public datasets exist to date for probing the importance of these clinically-reported features in the SOTA malignancy prediction algorithms. As part of this paper, we release a large-scale Clinically-Interpretable Radiomics Dataset, CIRDataset, containing 956 radiologist QA/QC'ed spiculation/lobulation annotations on segmented lung nodules from two public datasets, LIDC-IDRI (N=883) and LUNGx (N=73). We also present an end-to-end deep learning model based on multi-class Voxel2Mesh extension to segment nodules (while preserving spikes), classify spikes (sharp/spiculation and curved/lobulation), and perform malignancy prediction. Previous methods have performed malignancy prediction for LIDC and LUNGx datasets but without robust attribution to any clinically reported/actionable features (due to known hyperparameter sensitivity issues with general attribution schemes). With the release of this comprehensively-annotated CIRDataset and end-to-end deep learning baseline, we hope that malignancy prediction methods can validate their explanations, benchmark against our baseline, and provide clinically-actionable insights. Dataset, code, pretrained models, and docker containers are available at https://github.com/nadeemlab/CIR.

Mitigating Bias in Radiology Machine Learning: 3. Performance Metrics

The increasing use of machine learning (ML) algorithms in clinical settings raises concerns about bias in ML models. Bias can arise at any step of ML creation, including data handling, model development, and performance evaluation. Potential biases in the ML model can be minimized by implementing these steps correctly. This report focuses on performance evaluation and discusses model fitness, as well as a set of performance evaluation toolboxes: namely, performance metrics, performance interpretation maps, and uncertainty quantification. By discussing the strengths and limitations of each toolbox, our report highlights strategies and considerations to mitigate and detect biases during performance evaluations of radiology artificial intelligence models. Keywords: Segmentation, Diagnosis, Convolutional Neural Network (CNN) © RSNA, 2022.

Image Embeddings Extracted from CNNs Outperform Other Transfer Learning Approaches in Classification of Chest Radiographs

To identify the best transfer learning approach for the identification of the most frequent abnormalities on chest radiographs (CXRs), we used embeddings extracted from pretrained convolutional neural networks (CNNs). An explainable AI (XAI) model was applied to interpret black-box model predictions and assess its performance. Seven CNNs were trained on CheXpert. Three transfer learning approaches were thereafter applied to a local dataset. The classification results were ensembled using simple and entropy-weighted averaging. We applied Grad-CAM (an XAI model) to produce a saliency map. Grad-CAM maps were compared to manually extracted regions of interest, and the training time was recorded. The best transfer learning model was that which used image embeddings and random forest with simple averaging, with an average AUC of 0.856. Grad-CAM maps showed that the models focused on specific features of each CXR. CNNs pretrained on a large public dataset of medical images can be exploited as feature extractors for tasks of interest. The extracted image embeddings contain relevant information that can be used to train an additional classifier with satisfactory performance on an independent dataset, demonstrating it to be the optimal transfer learning strategy and overcoming the need for large private datasets, extensive computational resources, and long training times.

Protocol for a qualitative study to explore acceptability, barriers and facilitators of the implementation of new teleophthalmology technologies between community optometry practices and hospital eye services

INTRODUCTION

Novel teleophthalmology technologies have the potential to reduce unnecessary and inaccurate referrals between community optometry practices and hospital eye services and as a result improve patients' access to appropriate and timely eye care. However, little is known about the acceptability and facilitators and barriers to the implementations of these technologies in real life.

METHODS AND ANALYSIS

A theoretically informed, qualitative study will explore patients' and healthcare professionals' perspectives on teleophthalmology and Artificial Intelligence Decision Support System models of care. A combination of situated observations in community optometry practices and hospital eye services, semistructured qualitative interviews with patients and healthcare professionals and self-audiorecordings of healthcare professionals will be conducted. Participants will be purposively selected from 4 to 5 hospital eye services and 6-8 affiliated community optometry practices. The aim will be to recruit 30-36 patients and 30 healthcare professionals from hospital eye services and community optometry practices. All interviews will be audiorecorded, with participants' permission, and transcribed verbatim. Data from interviews, observations and self-audiorecordings will be analysed thematically and will be informed by normalisation process theory and an inductive approach.

ETHICS AND DISSEMINATION

Ethical approval has been received from London-Bromley research ethics committee. Findings will be reported through academic journals and conferences in ophthalmology, health services research, management studies and human-computer interaction.

Radiomics and deep learning for myocardial scar screening in hypertrophic cardiomyopathy

BACKGROUND

Myocardial scar burden quantified using late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR), has important prognostic value in hypertrophic cardiomyopathy (HCM). However, nearly 50% of HCM patients have no scar but undergo repeated gadolinium-based CMR over their life span. We sought to develop an artificial intelligence (AI)-based screening model using radiomics and deep learning (DL) features extracted from balanced steady state free precession (bSSFP) cine sequences to identify HCM patients without scar.

METHODS

We evaluated three AI-based screening models using bSSFP cine image features extracted by radiomics, DL, or combined DL-Radiomics. Images for 759 HCM patients (50 ± 16 years, 66% men) in a multi-center/vendor study were used to develop and test model performance. An external dataset of 100 HCM patients (53 ± 14 years, 70% men) was used to assess model generalizability. Model performance was evaluated using area-under-receiver-operating curve (AUC).

RESULTS

The DL-Radiomics model demonstrated higher AUC compared to DL and Radiomics in the internal (0.83 vs 0.77, p = 0.006 and 0.78, p = 0.05; n = 159) and external (0.74 vs 0.64, p = 0.006 and 0.71, p = 0.27; n = 100) datasets. The DL-Radiomics model correctly identified 43% and 28% of patients without scar in the internal and external datasets compared to 42% and 16% by Radiomics model and 42% and 23% by DL model, respectively.

CONCLUSIONS

A DL-Radiomics AI model using bSSFP cine images outperforms DL or Radiomics models alone as a scar screening tool prior to gadolinium administration. Despite its potential, the clinical utility of the model remains limited and further investigation is needed to improve the accuracy and generalizability.

Uncertainty-Aware and Lesion-Specific Image Synthesis in Multiple Sclerosis Magnetic Resonance Imaging: A Multicentric Validation Study

Generative adversarial networks (GANs) can synthesize high-contrast MRI from lower-contrast input. Targeted translation of parenchymal lesions in multiple sclerosis (MS), as well as visualization of model confidence further augment their utility, provided that the GAN generalizes reliably across different scanners. We here investigate the generalizability of a refined GAN for synthesizing high-contrast double inversion recovery (DIR) images and propose the use of uncertainty maps to further enhance its clinical utility and trustworthiness. A GAN was trained to synthesize DIR from input fluid-attenuated inversion recovery (FLAIR) and T1w of 50 MS patients (training data). In another 50 patients (test data), two blinded readers (R1 and R2) independently quantified lesions in synthetic DIR (synthDIR), acquired DIR (trueDIR) and FLAIR. Of the 50 test patients, 20 were acquired on the same scanner as training data (internal data), while 30 were scanned at different scanners with heterogeneous field strengths and protocols (external data). Lesion-to-Background ratios (LBR) for MS-lesions vs. normal appearing white matter, as well as image quality parameters were calculated. Uncertainty maps were generated to visualize model confidence. Significantly more MS-specific lesions were found in synthDIR compared to FLAIR (R1: 26.7 ± 2.6 vs. 22.5 ± 2.2 p < 0.0001; R2: 22.8 ± 2.2 vs. 19.9 ± 2.0, p = 0.0005). While trueDIR remained superior to synthDIR in R1 [28.6 ± 2.9 vs. 26.7 ± 2.6 (p = 0.0021)], both sequences showed comparable lesion conspicuity in R2 [23.3 ± 2.4 vs. 22.8 ± 2.2 (p = 0.98)]. Importantly, improvements in lesion counts were similar in internal and external data. Measurements of LBR confirmed that lesion-focused GAN training significantly improved lesion conspicuity. The use of uncertainty maps furthermore helped discriminate between MS lesions and artifacts. In conclusion, this multicentric study confirms the external validity of a lesion-focused Deep-Learning tool aimed at MS imaging. When implemented, uncertainty maps are promising to increase the trustworthiness of synthetic MRI.

Accurate auto-labeling of chest X-ray images based on quantitative similarity to an explainable AI model

The inability to accurately, efficiently label large, open-access medical imaging datasets limits the widespread implementation of artificial intelligence models in healthcare. There have been few attempts, however, to automate the annotation of such public databases; one approach, for example, focused on labor-intensive, manual labeling of subsets of these datasets to be used to train new models. In this study, we describe a method for standardized, automated labeling based on similarity to a previously validated, explainable AI (xAI) model-derived-atlas, for which the user can specify a quantitative threshold for a desired level of accuracy (the probability-of-similarity, pSim metric). We show that our xAI model, by calculating the pSim values for each clinical output label based on comparison to its training-set derived reference atlas, can automatically label the external datasets to a user-selected, high level of accuracy, equaling or exceeding that of human experts. We additionally show that, by fine-tuning the original model using the automatically labelled exams for retraining, performance can be preserved or improved, resulting in a highly accurate, more generalized model.

Multimodal biological brain age prediction using magnetic resonance imaging and angiography with the identification of predictive regions

Biological brain age predicted using machine learning models based on high-resolution imaging data has been suggested as a potential biomarker for neurological and cerebrovascular diseases. In this work, we aimed to develop deep learning models to predict the biological brain age using structural magnetic resonance imaging and angiography datasets from a large database of 2074 adults (21-81 years). Since different imaging modalities can provide complementary information, combining them might allow to identify more complex aging patterns, with angiography data, for instance, showing vascular aging effects complementary to the atrophic brain tissue changes seen in T1-weighted MRI sequences. We used saliency maps to investigate the contribution of cortical, subcortical, and arterial structures to the prediction. Our results show that combining T1-weighted and angiography MR data led to a significantly improved brain age prediction accuracy, with a mean absolute error of 3.85 years comparing the predicted and chronological age. The most predictive brain regions included the lateral sulcus, the fourth ventricle, and the amygdala, while the brain arteries contributing the most to the prediction included the basilar artery, the middle cerebral artery M2 segments, and the left posterior cerebral artery. Our study proposes a framework for brain age prediction using multimodal imaging, which gives accurate predictions and allows identifying the most predictive regions for this task, which can serve as a surrogate for the brain regions that are most affected by aging.