Disentangling Fact from Grid Cell Fiction in Trained Deep Path Integrators.

Work on deep learning-based models of grid cells suggests that grid cells generically and robustly arise from optimizing networks to path integrate, i.e., track one's spatial position by integrating self-velocity signals. In previous work [27], we challenged this path integration hypothesis by showing that deep neural networks trained to path integrate almost always do so, but almost never learn grid-like tuning unless separately inserted by researchers via mechanisms unrelated to path integration. In this work, we restate the key evidence substantiating these insights, then address a response to [27] by authors of one of the path integration hypothesis papers [32]. First, we show that the response misinterprets our work, indirectly confirming our points. Second, we evaluate the response's preferred "unified theory for the origin of grid cells" in trained deep path integrators [31, 33, 34] and show that it is at best "occasionally suggestive," not exact or comprehensive. We finish by considering why assessing model quality through prediction of biological neural activity by regression of activity in deep networks [23] can lead to the wrong conclusions.

Fast Optical Signals for Real-Time Retinotopy and Brain Computer Interface.

A brain-computer interface (BCI) allows users to control external devices through brain activity. Portable neuroimaging techniques, such as near-infrared (NIR) imaging, are suitable for this goal. NIR imaging has been used to measure rapid changes in brain optical properties associated with neuronal activation, namely fast optical signals (FOS) with good spatiotemporal resolution. However, FOS have a low signal-to-noise ratio, limiting their BCI application. Here FOS were acquired with a frequency-domain optical system from the visual cortex during visual stimulation consisting of a rotating checkerboard wedge, flickering at 5 Hz. We used measures of photon count (Direct Current, DC light intensity) and time of flight (phase) at two NIR wavelengths (690 nm and 830 nm) combined with a machine learning approach for fast estimation of visual-field quadrant stimulation. The input features of a cross-validated support vector machine classifier were computed as the average modulus of the wavelet coherence between each channel and the average response among all channels in 512 ms time windows. An above chance performance was obtained when differentiating visual stimulation quadrants (left vs. right or top vs. bottom) with the best classification accuracy of ~63% (information transfer rate of ~6 bits/min) when classifying the superior and inferior stimulation quadrants using DC at 830 nm. The method is the first attempt to provide generalizable retinotopy classification relying on FOS, paving the way for the use of FOS in real-time BCI.

Deep Learning-Based Digitally Reconstructed Tomography of the Chest in the Evaluation of Solitary Pulmonary Nodules: A Feasibility Study.

RATIONALE AND OBJECTIVES: Computed tomography (CT) is preferred for evaluating solitary pulmonary nodules (SPNs) but access or availability may be lacking, in addition, overlapping anatomy can hinder detection of SPNs on chest radiographs. We developed and evaluated the clinical feasibility of a deep learning algorithm to generate digitally reconstructed tomography (DRT) images of the chest from digitally reconstructed frontal and lateral radiographs (DRRs) and use them to detect SPNs. METHODS: This single-institution retrospective study included 637 patients with noncontrast helical CT of the chest (mean age 68 years, median age 69 years, standard deviation 11.7 years; 355 women) between 11/2012 and 12/2020, with SPNs measuring 10-30 mm. A deep learning model was trained on 562 patients, validated on 60 patients, and tested on the remaining 15 patients. Diagnostic performance (SPN detection) from planar radiography (DRRs and CT scanograms, PR) alone or with DRT was evaluated by two radiologists in an independent blinded fashion. The quality of the DRT SPN image in terms of nodule size and location, morphology, and opacity was also evaluated, and compared to the ground-truth CT images RESULTS: Diagnostic performance was higher from DRT plus PR than from PR alone (area under the receiver operating characteristic curve 0.95-0.98 versus 0.80-0.85; p < 0.05). DRT plus PR enabled diagnosis of SPNs in 11 more patients than PR alone. Interobserver agreement was 0.82 for DRT plus PR and 0.89 for PR alone; and interobserver agreement for size and location, morphology, and opacity of the DRT SPN was 0.94, 0.68, and 0.38, respectively. CONCLUSION: For SPN detection, DRT plus PR showed better diagnostic performance than PR alone. Deep learning can be used to generate DRT images and improve detection of SPNs.

Longitudinal assessment of demographic representativeness in the Medical Imaging and Data Resource Center open data commons.

Purpose: The Medical Imaging and Data Resource Center (MIDRC) open data commons was launched to accelerate the development of artificial intelligence (AI) algorithms to help address the COVID-19 pandemic. The purpose of this study was to quantify longitudinal representativeness of the demographic characteristics of the primary MIDRC dataset compared to the United States general population (US Census) and COVID-19 positive case counts from the Centers for Disease Control and Prevention (CDC). Approach: The Jensen-Shannon distance (JSD), a measure of similarity of two distributions, was used to longitudinally measure the representativeness of the distribution of (1) all unique patients in the MIDRC data to the 2020 US Census and (2) all unique COVID-19 positive patients in the MIDRC data to the case counts reported by the CDC. The distributions were evaluated in the demographic categories of age at index, sex, race, ethnicity, and the combination of race and ethnicity. Results: Representativeness of the MIDRC data by ethnicity and the combination of race and ethnicity was impacted by the percentage of CDC case counts for which this was not reported. The distributions by sex and race have retained their level of representativeness over time. Conclusion: The representativeness of the open medical imaging datasets in the curated public data commons at MIDRC has evolved over time as the number of contributing institutions and overall number of subjects have grown. The use of metrics, such as the JSD support measurement of representativeness, is one step needed for fair and generalizable AI algorithm development.

Toward fairness in artificial intelligence for medical image analysis: identification and mitigation of potential biases in the roadmap from data collection to model deployment.

Purpose: To recognize and address various sources of bias essential for algorithmic fairness and trustworthiness and to contribute to a just and equitable deployment of AI in medical imaging, there is an increasing interest in developing medical imaging-based machine learning methods, also known as medical imaging artificial intelligence (AI), for the detection, diagnosis, prognosis, and risk assessment of disease with the goal of clinical implementation. These tools are intended to help improve traditional human decision-making in medical imaging. However, biases introduced in the steps toward clinical deployment may impede their intended function, potentially exacerbating inequities. Specifically, medical imaging AI can propagate or amplify biases introduced in the many steps from model inception to deployment, resulting in a systematic difference in the treatment of different groups. Approach: Our multi-institutional team included medical physicists, medical imaging artificial intelligence/machine learning (AI/ML) researchers, experts in AI/ML bias, statisticians, physicians, and scientists from regulatory bodies. We identified sources of bias in AI/ML, mitigation strategies for these biases, and developed recommendations for best practices in medical imaging AI/ML development. Results: Five main steps along the roadmap of medical imaging AI/ML were identified: (1) data collection, (2) data preparation and annotation, (3) model development, (4) model evaluation, and (5) model deployment. Within these steps, or bias categories, we identified 29 sources of potential bias, many of which can impact multiple steps, as well as mitigation strategies. Conclusions: Our findings provide a valuable resource to researchers, clinicians, and the public at large.

Opportunistic detection of type 2 diabetes using deep learning from frontal chest radiographs.

Deep learning (DL) models can harness electronic health records (EHRs) to predict diseases and extract radiologic findings for diagnosis. With ambulatory chest radiographs (CXRs) frequently ordered, we investigated detecting type 2 diabetes (T2D) by combining radiographic and EHR data using a DL model. Our model, developed from 271,065 CXRs and 160,244 patients, was tested on a prospective dataset of 9,943 CXRs. Here we show the model effectively detected T2D with a ROC AUC of 0.84 and a 16% prevalence. The algorithm flagged 1,381 cases (14%) as suspicious for T2D. External validation at a distinct institution yielded a ROC AUC of 0.77, with 5% of patients subsequently diagnosed with T2D. Explainable AI techniques revealed correlations between specific adiposity measures and high predictivity, suggesting CXRs' potential for enhanced T2D screening.

Validation of a deep learning, value-based care model to predict mortality and comorbidities from chest radiographs in COVID-19.

We validate a deep learning model predicting comorbidities from frontal chest radiographs (CXRs) in patients with coronavirus disease 2019 (COVID-19) and compare the model's performance with hierarchical condition category (HCC) and mortality outcomes in COVID-19. The model was trained and tested on 14,121 ambulatory frontal CXRs from 2010 to 2019 at a single institution, modeling select comorbidities using the value-based Medicare Advantage HCC Risk Adjustment Model. Sex, age, HCC codes, and risk adjustment factor (RAF) score were used. The model was validated on frontal CXRs from 413 ambulatory patients with COVID-19 (internal cohort) and on initial frontal CXRs from 487 COVID-19 hospitalized patients (external cohort). The discriminatory ability of the model was assessed using receiver operating characteristic (ROC) curves compared to the HCC data from electronic health records, and predicted age and RAF score were compared using correlation coefficient and absolute mean error. The model predictions were used as covariables in logistic regression models to evaluate the prediction of mortality in the external cohort. Predicted comorbidities from frontal CXRs, including diabetes with chronic complications, obesity, congestive heart failure, arrhythmias, vascular disease, and chronic obstructive pulmonary disease, had a total area under ROC curve (AUC) of 0.85 (95% CI: 0.85-0.86). The ROC AUC of predicted mortality for the model was 0.84 (95% CI,0.79-0.88) for the combined cohorts. This model using only frontal CXRs predicted select comorbidities and RAF score in both internal ambulatory and external hospitalized COVID-19 cohorts and was discriminatory of mortality, supporting its potential use in clinical decision making.

Detecting Racial/Ethnic Health Disparities Using Deep Learning From Frontal Chest Radiography.

PURPOSE: The aim of this study was to assess racial/ethnic and socioeconomic disparities in the difference between atherosclerotic vascular disease prevalence measured by a multitask convolutional neural network (CNN) deep learning model using frontal chest radiographs (CXRs) and the prevalence reflected by administrative hierarchical condition category codes in two cohorts of patients with coronavirus disease 2019 (COVID-19). METHODS: A CNN model, previously published, was trained to predict atherosclerotic disease from ambulatory frontal CXRs. The model was then validated on two cohorts of patients with COVID-19: 814 ambulatory patients from a suburban location (presenting from March 14, 2020, to October 24, 2020, the internal ambulatory cohort) and 485 hospitalized patients from an inner-city location (hospitalized from March 14, 2020, to August 12, 2020, the external hospitalized cohort). The CNN model predictions were validated against electronic health record administrative codes in both cohorts and assessed using the area under the receiver operating characteristic curve (AUC). The CXRs from the ambulatory cohort were also reviewed by two board-certified radiologists and compared with the CNN-predicted values for the same cohort to produce a receiver operating characteristic curve and the AUC. The atherosclerosis diagnosis discrepancy, Δvasc, referring to the difference between the predicted value and presence or absence of the vascular disease HCC categorical code, was calculated. Linear regression was performed to determine the association of Δvasc with the covariates of age, sex, race/ethnicity, language preference, and social deprivation index. Logistic regression was used to look for an association between the presence of any hierarchical condition category codes with Δvasc and other covariates. RESULTS: The CNN prediction for vascular disease from frontal CXRs in the ambulatory cohort had an AUC of 0.85 (95% confidence interval, 0.82-0.89) and in the hospitalized cohort had an AUC of 0.69 (95% confidence interval, 0.64-0.75) against the electronic health record data. In the ambulatory cohort, the consensus radiologists' reading had an AUC of 0.89 (95% confidence interval, 0.86-0.92) relative to the CNN. Multivariate linear regression of Δvasc in the ambulatory cohort demonstrated significant negative associations with non-English-language preference (ß = -0.083, P < .05) and Black or Hispanic race/ethnicity (ß = -0.048, P < .05) and positive associations with age (ß = 0.005, P < .001) and sex (ß = 0.044, P < .05). For the hospitalized cohort, age was also significant (ß = 0.003, P < .01), as was social deprivation index (ß = 0.002, P < .05). The Δvasc variable (odds ratio [OR], 0.34), Black or Hispanic race/ethnicity (OR, 1.58), non-English-language preference (OR, 1.74), and site (OR, 0.22) were independent predictors of having one or more hierarchical condition category codes (P < .01 for all) in the combined patient cohort. CONCLUSIONS: A CNN model was predictive of aortic atherosclerosis in two cohorts (one ambulatory and one hospitalized) with COVID-19. The discrepancy between the CNN model and the administrative code, Δvasc, was associated with language preference in the ambulatory cohort; in the hospitalized cohort, this discrepancy was associated with social deprivation index. The absence of administrative code(s) was associated with Δvasc in the combined cohorts, suggesting that Δvasc is an independent predictor of health disparities. This may suggest that biomarkers extracted from routine imaging studies and compared with electronic health record data could play a role in enhancing value-based health care for traditionally underserved or disadvantaged patients for whom barriers to care exist.