Fahrenholtz, Samuel J.

Tissue harmonic and spatial compound imaging, widely used US techniques, improve image quality by reducing noise and artifacts through complementary mechanisms.

Abstract Background Despite the high precision of deep learning models on internal tests on CT, their effectiveness often drops on external validation due to artifacts caused by patient motion and implants like metal or silicone that were not accounted for in the carefully curated training data. Potential wide categories of “unknown” anomalies within a CT exam makes training a supervised model for out‐of‐distribution (OOD) identification impractical, especially when considering unseen external data. Purpose To develop an artificial intelligence (AI) model to detect and identify anomalies/OOD data in abdominal–pelvis CT exams for the purpose of improving the performance of downstream applications. Methods Our proposed 2D and 3D generative architecture receives the third lumbar vertebra (L3) slice (slice‐level model) or all the slices from the series (series‐level model), generates a reconstruction and the secondary part of our architecture—anomaly score computation block, computes the anomalies pixels/voxels (slice‐level/series‐level) to identify anomalous L3‐slices/volumes (slice‐level/series‐level). We trained on data from over 2850 abdominal–pelvis CT volumes from adults over age 50 years collected throughout multiple Mayo Clinic campuses (60% female; mean age: 66.9, 92.4% non‐Hispanic White) and tested on a prospective test set of 544 CTs from July 2024 (47.3% female; mean age: 70.9, 94% non‐Hispanic White) as well as an external test set. Results We found that while traditional methods show moderate success, our generative models—Vector Quantized Variational Autoencoder (VQVAE) and Vision Transformer‐Masked Autoencoder (VIT‐MAE)—deliver excellent results with negligible false positives (FPs) and are also superior in identifying varied types of OOD samples. Prospective analysis showed the model was able to handle the under‐documentation of anomaly in radiology reports with 86.11% true positive (TP) rate. We also performed external validation using the publicly available AbdominalCT‐1k dataset, which contains 1062 CT scans compiled from several existing benchmark datasets. The model achieved a 75.26% TP rate, while the 24.7% FP rate was primarily triggered by anomalies located outside the body. Conclusions The proposed method can be leveraged to detect both intra‐ and interclass OOD data from abdominal CT images and can assess the quality of CT datasets to provide actionable insights. This workflow is particularly valuable for nonshareable healthcare collaborations, where it can be deployed as a service within local firewalls for automated dataset curation without prior knowledge about the OOD types. The implementation of the algorithm is available in the GitHub: https://github.com/gokul‐ramasamy/implant_detection.git .

AbstractBackgroundWhile low bone density is a major burden on US health system, current osteoporosis screening guidelines by the US Preventive Services Task Force are limited to women aged ≥65 and all postmenopausal women with certain risk factors. Even within recommended screening groups, actual screening rates are low (<26%) and vary across socioeconomic groups. The proposed model can opportunistically screen patients using abdominal CT studies for low bone density who may otherwise go undiagnosed.PurposeTo develop an artificial intelligence (AI) model for opportunistic screening of low bone density using both contrast and non‐contrast abdominopelvic computed tomography (CT) exams, for the purpose of referral to traditional bone health management, which typically begins with dual energy X‐ray absorptiometry (DXA).MethodsWe collected 6083 contrast‐enhanced CT imaging exams paired with DXA exams within ±6 months documented between May 2015 and August 2021 in a single institution with four major healthcare practice regions. Our fusion AI pipeline receives the coronal and axial plane images of a contrast enhanced abdominopelvic CT exam and basic patient demographics (age, gender, body cross section lengths) to predict risk of low bone mass. The models were trained on lumbar spine T‐scores from DXA exams and tested on multi‐site imaging exams. The model was again tested in a prospective group (N = 344) contrast‐enhanced and non‐contrast‐enhanced studies.ResultsThe models were evaluated on the same test set (1208 exams)—(1) Baseline model using demographic factors from electronic medical records (EMR) ‐ 0.7 area under the curve of receiver operator characteristic (AUROC); Imaging based models: (2) axial view ‐ 0.83 AUROC; (3) coronal view‐ 0.83 AUROC; (4) Fusion model—Imaging + demographic factors ‐ 0.86 AUROC. The prospective test yielded one missed positive DXA case with a hip prosthesis among 23 positive contrast‐enhanced CT exams and 0% false positive rate for non‐contrast studies. Both positive cases among non‐contrast enhanced CT exams were successfully detected. While only about 8% patients from prospective study received a DXA exam within 2 years, about 30% were detected with low bone mass by the fusion model, highlighting the need for opportunistic screening.ConclusionsThe fusion model, which combines two planes of CT images and EMRs data, outperformed individual models and provided a high, robust diagnostic performance for opportunistic screening of low bone density using contrast and non‐contrast CT exams. This model could potentially improve bone health risk assessment with no additional cost. The model's handling of metal implants is an ongoing effort.

Abstract Background Dual‐energy X‐ray absorptiometry (DXA) is an exam that measures areal bone mineral density (aBMD) and is regularly used to diagnose and monitor osteoporosis. Except for exam quality issues such as operator error, the quantitative results of an exam are not modified by a radiologist or other physician. DXA cross‐calibration errors can shift diagnoses, conceivably leading to alternate intervention decisions and patient outcomes. Purpose After identifying and correcting a cross‐calibration bias of 3.8% in our two DXA scanners’ aBMD measurements, we investigated misdiagnosis rates for given cross‐calibration errors in a single patient cohort to determine the impact on patient care and the value of cross‐calibration quality control. Methods The studied cohort was 8012 patients of all ages and sexes with femoral neck exams that were scanned on a single DXA unit from October 1, 2018 to March 31, 2021. There were six subcohorts delineated by age and sex, three female groups and three male groups. Data reporting focused on the highest risk subcohort of 2840 females aged 65 or older. The DXA unit had no calibration changes during that time. Only one femoral neck—left or right—was randomly chosen for analysis. Patients with multiple qualifying exams within the time interval had one exam randomly chosen. The proof‐of‐principle simulation shifted the aBMD values within a range of ±10%, ±8%, ±6%, ±4%, ±3.5%, ±3%, ±2.5%, ±2%, ±1.5%, ±1%, ±0.5%, and 0 (no shift); the cross‐calibration shifts were informed by published results and institutional experience. Measurement precision was modeled by randomly sampling a Gaussian distribution characterized by the worst acceptable least significant change (LSC) of 6.9%, with 100 000 samplings for each patient. T‐scores were recalculated from the shifted aBMD values, followed by reassigned diagnoses from the World Health Organization's T‐score‐based scheme. Results The unshifted original subcohort of women aged 65 and older had 599 normal diagnoses (21.1% of the cohort), 1784 osteopenia diagnoses (62.8%), and 455 osteoporosis diagnoses (16.1%). Osteoporosis diagnosis rates were highly sensitive to aBMD shifts. At the extrema, a −10% aBMD shift led to +161% osteoporosis cases, and a +10% aBMD shift led to −64.5% osteoporosis cases. Within the more plausible ±4% aBMD error range, the osteoporosis diagnosis rate changed −10.5% per +1% aBMD shift as indicated by linear regression ( R 2 = 0.98). Except for the men aged 49 years and younger subcohort, the total cohort and five subcohorts had fit line slopes ranging between −9.7% and −12.1% with R 2 ≥ 0.98. Cross‐calibration bias had greater influence for diagnosis count rates compared to measurement precision, that is, LSC. Conclusions These results quantify the degree of misdiagnosis that can occur in a clinically relevant cohort due to cross‐calibration bias. In medical practices where patients may be scanned on more than one DXA unit, ensuring cross‐calibration quality is a critical and high‐value quality control task with direct impact on patient diagnosis and treatment course. The clinical impact and incidence of poor DXA quality control practices, and cross‐calibration in particular, should be studied further.

AbstractPurposeThe low exposures, unique x‐ray beam geometry, and scanning design in dual‐energy x‐ray absorptiometry (DXA) make measurement and quality‐control strategies different from traditional x‐ray equipment. This study examines the dependence of measured entrance‐air‐kerma (EAK) on both dose sensor type and scan length. The feasibility of using EAK to compare scanner output between different scan modes, individual scanners, and scanner platforms was also established. Finally, the congruence between measured and vendor‐reported EAK was analyzed.MethodsFour Hologic DXA scanners at two institutions and all four available scan modes were tested. EAK was measured directly by three types of Radcal dose sensors: 60‐cc pancake ion‐chamber (IC), 180‐cc pancake IC, and solid‐state detector. The coefficient of variation (COV) was used to assess the dependence of EAK on scan length. Variations in EAK between the types of dose sensors as well as measured versus vendor‐reported values were evaluated using Bland–Altman analysis: mean ±95% prediction interval (PI): 1.96σ.ResultsDose sensor variations in EAK were minimal, with a −3.5 ± 3.5% (mean ±95% PI) percent difference between the two sizes of IC's. The solid‐state detector produced highly similar measurements to the 180‐cc IC. These small differences were consistent across all scanners and all scan modes tested. Neither measured nor vendor‐reported EAK values were found to show relevant dependence on scan length, with all COV values ≤4%. Differences between measured and reported EAK were higher at −6 ± 48%. Likely errors in vendor‐reported EAK calculations were also identified.ConclusionIt is feasible to quantify DXA scanner stability using EAK as a quality‐control metric with a variety of solid‐state and IC dose sensors, and the scan length used is not critical. Although vendor‐reported EAK was consistent among scanners of the same platform, measured EAK varied significantly from scanner to scanner. As a result, measured and reported EAK may not always be comparable.