Deep Learning for Breast Cancer Risk Prediction: Application to a Large Representative UK Screening Cohort.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop an artificial intelligence (AI) deep learning tool capable of predicting future breast cancer risk from a current negative screening mammographic examination and to evaluate the model on data from the UK National Health Service Breast Screening Program. Materials and Methods The OPTIMAM Mammography Imaging Database contains screening data, including mammograms and information on interval cancers, for > 300,000 women who attended screening at three different sites in the UK from 2012 onward. Cancer-free screening examinations from women aged 50-70 years were obtained and classified as risk-positive or risk-negative based on the occurrence of cancer within 3 years of the original examination. Examinations with confirmed cancer and images containing implants were excluded. From the resulting 5264 risk-positive and 191488 risk-negative examinations, training (n = 89285) validation (n = 2106) and test (n = 39351) datasets were produced for model development and evaluation. The AI model was trained to predict future cancer occurrence based on screening mammograms and patient age. Performance was evaluated on the test dataset using the area under the receiver operating characteristic curve (AUC) and compared across subpopulations to assess potential biases. Interpretability of the model was explored, including with saliency maps. Results On the hold-out test set, the AI model achieved an overall AUC of 0.70 (95% CI: 0.69, 0.72). There was no evidence of a difference in performance across the three sites, between patient ethnicities or across age-groups Visualization of saliency maps and sample images provided insights into the mammographic features associated with AI-predicted cancer risk. Conclusion The developed AI tool showed good performance on a multisite, UK-specific dataset. ©RSNA, 2024.

Lessons learned from independent external validation of an AI tool to detect breast cancer using a representative UK data set.

OBJECTIVE: To pilot a process for the independent external validation of an artificial intelligence (AI) tool to detect breast cancer using data from the NHS breast screening programme (NHSBSP). METHODS: A representative data set of mammography images from 26,000 women attending 2 NHS screening centres, and an enriched data set of 2054 positive cases were used from the OPTIMAM image database. The use case of the AI tool was the replacement of the first or second human reader. The performance of the AI tool was compared to that of human readers in the NHSBSP. RESULTS: Recommendations for future external validations of AI tools to detect breast cancer are provided. The tool recalled different breast cancers to the human readers. This study showed the importance of testing AI tools on all types of cases (including non-standard) and the clarity of any warning messages. The acceptable difference in sensitivity and specificity between the AI tool and human readers should be determined. Any information vital for the clinical application should be a required output for the AI tool. It is recommended that the interaction of radiologists with the AI tool, and the effect of the AI tool on arbitration be investigated prior to clinical use. CONCLUSION: This pilot demonstrated several lessons for future independent external validation of AI tools for breast cancer detection. ADVANCES IN KNOWLEDGE: Knowledge has been gained towards best practice procedures for performing independent external validations of AI tools for the detection of breast cancer using data from the NHS Breast Screening Programme.

Virtual Versus In-person Grand Rounds in Orthopaedics: A Framework for Implementation and Participant-reported Outcomes.

INTRODUCTION: Grand rounds have been weekly gatherings at academic orthopaedic surgery programs across the country for decades. During the 50th year of grand rounds at our institution, the COVID-19 pandemic prompted the transformation of this in-person forum into a virtual setting. The purpose of this study was to detail this initiative and to report survey data providing participant-reported perceptions and satisfaction of virtual versus in-person grand rounds. MATERIALS AND METHODS: Once in-person meetings were discontinued, virtual grand rounds commenced using the Zoom video application. At the conclusion of the 2020 to 2021 academic year, a 30-item online survey was sent to all residents, faculty, and visiting faculty to assess their perspective and satisfaction. A five-point Likert scale ranging from 1 to 5, with 5 being extremely effective, was used. A 21-item follow-up survey was sent to all speakers as well. RESULTS: Thirty-six virtual grand rounds were successfully hosted. The response rate for the survey was excellent-80 of 86 (93.0%) surveys returned completed. Respondents found that virtual grand rounds were more convenient to attend, were more convenient to obtain Continuing Medical Education, and were more satisfied with virtual grand rounds. Respondents reported that in-person grand rounds were more effective for stimulating social collegiality and networking. Speakers found that virtual grand rounds were more effective for uploading the presentation and overall convenience, whereas they were less effective at retaining audience attention and receiving audience feedback. Improved faculty attendance after the switch to virtual grand rounds was also noted. CONCLUSION: This study found that respondents across all groups appreciated the convenience of attending and obtaining Continuing Medical Educations at virtual grand rounds while also noting the merits of in-person grand rounds for promoting peer interaction, camaraderie, and departmental identity. All respondents strongly recommended continuation of this program in a hybrid format. Virtual orthopaedic grand rounds are viable, readily implemented and demonstrate improved participant satisfaction.

OPTIMAM Mammography Image Database: A Large-Scale Resource of Mammography Images and Clinical Data.

Supplemental material is available for this article.

Effect of intervertebral disc degeneration on mechanical and electric signals at the interface between disc and vertebra.

Intervertebral disc (IVD) degeneration is significantly correlated with the changes in structure and material properties of adjacent vertebral bone, possibly through mechanical and electrical interactions. However, the mechanisms underlying the alteration of the mechanical and electrical environment at the disc-vertebra interface related with disc degeneration have not been well studied. The objective of this study was to numerically investigate the long-term distributions of mechanical and electrical signals on the disc-vertebra interface with disc degeneration. A three-dimensional finite element model of a human lumbar IVD was used to study the mechanical and electric signals at the interface between disc and vertebral body. The disc degeneration was simulated by reducing the nutrition levels on the nucleus pulposus (NP)-vertebra interface and on the annulus fibrosus (AF) periphery to 30% and 60% of its normal values, respectively. In the simulation, the total external mechanical load applied to the disc-vertebra segment was assumed unchanged during disc degeneration. The simulation results showed that the compressive stress of solid matrix changed by up to ~37 kPa on the NP-vertebra interface, while it increased by up to ~32 kPa on the AF-vertebra interface. The shear stress increased by up to ~37 kPa with disc degeneration. The absolute value of the electric potential on the disc-vertebra interface of the disc slightly decreased with the disc degeneration (~0.5 mV). The knowledge of these spatial and temporal variations of the mechanical stresses and electric potential on the disc-vertebra interface is important for understanding the vertebrae adaptation and remodeling during disc degeneration.

Transport of Vancomycin and Cefepime Into Human Intervertebral Discs: Quantitative Analyses.

STUDY DESIGN: Simulation of antibiotics transport into human intervertebral disc with intravenous infusion. OBJECTIVE: The objective of this study was to quantitatively investigate antibiotic concentrations in the disc. SUMMARY OF BACKGROUND DATA: Intravenous infusion of antibiotics is typically used to treat intervertebral disc infection in clinics. However, it is difficult to evaluate the drug concentrations within discs in vivo. METHODS: A computational model was used in this study. The variation of drug charge with pH was considered in the model. Thirty-minute infusions of two commonly used antibiotics in clinic-vancomycin and cefepime-were numerically investigated. Spatial and temporal concentration distributions of these drugs in both nondegenerated and moderately degenerated discs were calculated. RESULTS: For intravenous infusion of 1 g vancomycin and 2 g cefepime in 30 minutes repeated every 12 hours, it was predicted that vancomycin concentration in the disc fluctuated between 17.0 and 31.0 times of its minimum inhibitory concentration (1 ug/mL) and cefepime concentration fluctuated between 1.1 and 4.2 times of its minimum inhibitory concentration (i.e., 8 ug/mL) in about 2 days. It was also found that vancomycin concentration in moderately degenerated disc was lower than that in the nondegenerated disc. CONCLUSION: This study provides quantitative guidance on selecting proper dosage for treating disc infection. The method used in this study could be used to provide quantitative information on transport of other antibiotics and drugs in discs as well. LEVEL OF EVIDENCE: N/A.

Effects of diurnal loading on the transport of charged antibiotics into intervertebral discs.

The objective of this study was to quantitatively analyze the effect of diurnal loading on the transport of various charged antibiotics into negatively charged human intervertebral disc (IVD). Transport of charged antibiotics into a human lumbar disc was analyzed using a 3D finite element model. The valence (z) of the electrical charge of antibiotics varied from z = +2 (positively charged) to z = -2 (negatively charged). An uncharged antibiotic (z = 0) was used as a control. Cases with transient antibiotic concentration at disc boundaries [to mimic intravenous (IV) infusion] were simulated. Our results showed that diurnal compression increased the concentrations in the nucleus pulposus (NP) region, but degreased the concentrations in the annulus fibrosus (AF) region for all charged or non-charged drugs. The overall concentration (averaged over disc) increased with diurnal compression. The diurnal compression had more effects on negatively charged antibiotics than positively charged ones. For example, at day 5 with diurnal compression, the diurnal compression increased the concentration of negatively charged drug (z = -1) in NP by 18.3%, but only by 6.6% for positively charged one (z = +1). In AF, diurnal compression decreased the concentration by 13.2% for negatively charged drug (z = -1) versus 1.2% for positively charged one (z = +1). Note these percentages are the averaged values over day 5. This study provides quantitative information on understanding the mechanisms of charged drug transport in human IVDs.

Diffusion of antibiotics in intervertebral disc.

Delivering charged antibiotics to the intervertebral disc is challenging because of the avascular, negatively charged extracellular matrix (ECM) of the tissue. The purpose of this study was to measure the apparent diffusion coefficient of two clinically relevant, charged antibiotics, vancomycin (positively charged) and oxacillin (negatively charged) in IVD. A one-dimensional steady state diffusion experiment was employed to measure the apparent diffusion coefficient of the two antibiotics in bovine coccygeal annulus fibrosus (AF) tissue. The averaged apparent diffusion coefficient for vancomycin under 20% compressive strain was 7.94 ±â€¯2.00 × 10-12 m2/s (n = 10), while that of oxacillin was 2.26 ±â€¯0.68 × 10-10 m2/s (n = 10). A student's t-test showed that the diffusivity of vancomycin was significantly lower than that of oxacillin. This finding may be attributed to two factors: solute size and possible binding effects. Vancomycin is approximately 3 times larger in molecular weight than oxacillin, meaning that steric hindrance likely plays a role in the slower transport. Reversible binding between positive vancomycin and the negative ECM could also slow down the rate of diffusion. Therefore, more investigation is necessary to determine the specific relationship between net charge on antibiotic and diffusion coefficients in IVD. This study provides essential quantitative information regarding the transport rates of antibiotics in the IVD, which is critical in using computational modeling to design effective strategies to treat disc infection.

Simulation of water content distributions in degenerated human intervertebral discs.

The objective of this study was to investigate the spatial and temporal variations of water content in intervertebral discs during degeneration and repair processes. We hypothesized that the patterns of water content distribution in the discs are related to the intensity patterns observed in T2-weighted MRI images. Water content distributions in the mildly (e.g., 80% viable cells in the disc, 2.3% decrease in disc height) and moderately (e.g., 40% viable cells in the disc, 9.3% decrease in disc height) degenerated discs were predicted using a finite element model. The variation of water content in the degenerated discs treated with three biological therapies (i.e., increasing the cell density in the nucleus pulposus [Case I], increasing glycosaminoglycan synthesis rate in the nucleus pulposus [Case II], and decreasing glycosaminoglycan degradation rate in the nucleus pulposus [Case III]) were also predicted. It was found that two patterns of water content distributions, a horizontal region with lower water content at the mid-axial plane of nucleus pulposus and a spot with higher water content at the posterior region, were shown during the degeneration progress for the disc simulated in this study. These two patterns disappeared after treatment in Case I, but in Case II and Case III. The implication of these patterns for the horizontal gray band and high intensity zone in T2-weighted MRI images was discussed. This study provided new guidance to develop a novel method for diagnosing disc degeneration and assessing outcomes of biological therapies with MRI techniques. © 2016 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 35:147-153, 2017.

Kinetics of charged antibiotic penetration into human intervertebral discs: A numerical study.

Little quantitative information exists on the kinetics of charged antibiotic penetration into human intervertebral discs (IVD). This information is crucial for determining the dosage to use, timing of administration, and duration of treatment for infected IVDs. The objective of this study was to quantitatively analyze the transport of various charged antibiotics into human lumbar IVDs. Penetration of charged and uncharged antibiotics into a human lumbar disc was analyzed using a 3D finite element model. The valence (z) of the electrical charge of antibiotics varied from z=+2 (positively charged) to z=-2 (negatively charged). An uncharged antibiotic (z=0) was used as a control. Cases with intravenous (IV) administrations of different charged antibiotics were simulated. Our results showed that the electrical charge had great effects on kinetics of an antibiotic penetration into the IVD; with higher concentrations and uptakes for positively charged antibiotics than those for negatively charged ones. This study provides quantitative information on selecting antibiotics for treating intervertebral disc infections.

The relationship between cancer detection in mammography and image quality measurements.

PURPOSE: To investigate the relationship between image quality measurements and the clinical performance of digital mammographic systems. METHODS: Mammograms containing subtle malignant non-calcification lesions and simulated malignant calcification clusters were adapted to appear as if acquired by four types of detector. Observers searched for suspicious lesions and gave these a malignancy score. Analysis was undertaken using jackknife alternative free-response receiver operating characteristics weighted figure of merit (FoM). Images of a CDMAM contrast-detail phantom were adapted to appear as if acquired using the same four detectors as the clinical images. The resultant threshold gold thicknesses were compared to the FoMs using a linear regression model and an F-test was used to find if the gradient of the relationship was significantly non-zero. RESULTS: The detectors with the best image quality measurement also had the highest FoM values. The gradient of the inverse relationship between FoMs and threshold gold thickness for the 0.25mm diameter disk was significantly different from zero for calcification clusters (p=0.027), but not for non-calcification lesions (p=0.11). Systems performing just above the minimum image quality level set in the European Guidelines for Quality Assurance in Breast Cancer Screening and Diagnosis resulted in reduced cancer detection rates compared to systems performing at the achievable level. CONCLUSIONS: The clinical effectiveness of mammography for the task of detecting calcification clusters was found to be linked to image quality assessment using the CDMAM phantom. The European Guidelines should be reviewed as the current minimum image quality standards may be too low.

Simulation of biological therapies for degenerated intervertebral discs.

The efficacy of biological therapies on intervertebral disc repair was quantitatively studied using a three-dimensional finite element model based on a cell-activity coupled multiphasic mixture theory. In this model, cell metabolism and matrix synthesis and degradation were considered. Three types of biological therapies-increasing the cell density (Case I), increasing the glycosaminoglycan (GAG) synthesis rate (Case II), and decreasing the GAG degradation rate (Case III)-to the nucleus pulposus (NP) of each of two degenerated discs [one mildly degenerated (e.g., 80% viable cells in the NP) and one severely degenerated (e.g., 30% viable cells in the NP)] were simulated. Degenerated discs without treatment were also simulated as a control. The cell number needed, nutrition level demanded, time required for the repair, and the long-term outcomes of these therapies were analyzed. For Case I, the repair process was predicted to be dependent on the cell density implanted and the nutrition level at disc boundaries. With sufficient nutrition supply, this method was predicted to be effective for treating both mildly and severely degenerated discs. For Case II, the therapy was predicted to be effective for repairing the mildly degenerated disc, but not for the severely degenerated disc. Similar results were predicted for Case III. No change in cell density for Cases II and III were predicted under normal nutrition level. This study provides a quantitative guide for choosing proper strategies of biological therapies for different degenerated discs.

Influences of Nutrition Supply and Pathways on the Degenerative Patterns in Human Intervertebral Disc.

STUDY DESIGN: Investigation of the effects of the impairment of different nutritional pathways on the intervertebral disc degeneration patterns in terms of spatial distributions of cell density, glycosaminoglycan content, and water content. OBJECTIVE: The aim of this study was to test the hypothesis that impairment of different nutritional pathways would result in different degenerative patterns in human discs. SUMMARY OF BACKGROUND DATA: Impairment of nutritional pathways has been found to affect cell viability in the disc. However, details on how impairment of different nutritional pathways affects the disc degeneration patterns are unknown. METHODS: A 3D finite element model was used for this study. This finite element method was based on the cell-activity coupled mechano-electrochemical theory for cartilaginous tissues. Impairment of the nutritional pathways was simulated by lowering the nutrition level at the disc boundaries. Effects of the impartment of cartilaginous endplate-nucleus pulposus (CEP-NP) pathway only (Case 1), annulus fibrosus (AF) pathway only (Case 2), and both pathways (Case 3) on disc degeneration patterns were studied. RESULTS: The predicted critical levels of nutrition for Case 1, Case 2, and Case 3 were around 30%, 20%, and 50% of the reference values, respectively. Below this critical level, the disc degeneration would occur. Disc degeneration appeared mainly in the NP for Case 1, in the outer AF for Case 2, and in both the NP and inner to middle AF for Case 3. For Cases 1 and 3, the loss of water content was primarily located in the mid-axial plane, which is consistent with the horizontal gray band seen in some T2-weighted magnetic resonance imaging (MRI). For the disc geometry used in this study, it was predicted that there existed a high-intensity zone (for Case 3), as seen in some T2-weighted MRI images. CONCLUSION: Impairment of different nutrition pathways results in different degenerative patterns. LEVEL OF EVIDENCE: N/A.

Breast cancer detection rates using four different types of mammography detectors.

OBJECTIVE: To compare the performance of different types of detectors in breast cancer detection. METHODS: A mammography image set containing subtle malignant non-calcification lesions, biopsy-proven benign lesions, simulated malignant calcification clusters and normals was acquired using amorphous-selenium (a-Se) detectors. The images were adapted to simulate four types of detectors at the same radiation dose: digital radiography (DR) detectors with a-Se and caesium iodide (CsI) convertors, and computed radiography (CR) detectors with a powder phosphor (PIP) and a needle phosphor (NIP). Seven observers marked suspicious and benign lesions. Analysis was undertaken using jackknife alternative free-response receiver operating characteristics weighted figure of merit (FoM). The cancer detection fraction (CDF) was estimated for a representative image set from screening. RESULTS: No significant differences in the FoMs between the DR detectors were measured. For calcification clusters and non-calcification lesions, both CR detectors' FoMs were significantly lower than for DR detectors. The calcification cluster's FoM for CR NIP was significantly better than for CR PIP. The estimated CDFs with CR PIP and CR NIP detectors were up to 15% and 22% lower, respectively, than for DR detectors. CONCLUSION: Cancer detection is affected by detector type, and the use of CR in mammography should be reconsidered. KEY POINTS: The type of mammography detector can affect the cancer detection rates. CR detectors performed worse than DR detectors in mammography. Needle phosphor CR performed better than powder phosphor CR. Calcification clusters detection is more sensitive to detector type than other cancers.

Simulation of the progression of intervertebral disc degeneration due to decreased nutritional supply.

STUDY DESIGN: Simulate the progression of human disc degeneration. OBJECTIVE: The objective of this study was to quantitatively analyze and simulate the changes in cell density, nutritional level, proteoglycan (PG) content, water content, and volume during human disc degeneration using a numerical method. SUMMARY OF BACKGROUND DATA: Understanding the cause and progression of intervertebral disc degeneration is crucial for developing effective treatment strategies for intervertebral disc degeneration-related diseases. During tissue degeneration, the disc undergoes losses of cell viability and activities, changes in extracellular matrix composition and structure, and compromise of the tissue-level integrity and function, which is significantly influenced by the intercoupled biological, chemical, electrical, and mechanical signals in the disc. Characterizing these signals in human discs in vivo is difficult. METHODS: A realistic 3-dimensional finite element model of the human intervertebral disc was developed on the basis of biomechanoelectrochemical continuum mixture theory. The theoretical framework and the constitutive relationships were all biophysics based. All the material properties were obtained from experimental results. The cell-mediated disc degeneration process caused by lowered nutritional levels at disc boundaries was simulated and validated by comparing with experimental results. RESULTS: Cell density reached equilibrium state in 30 days after reduced nutritional supply at the disc boundary, whereas the PG and water contents reached a new equilibrium state in 55 years. The simulated results for the distributions of PG and water contents within the disc were consistent with the results measured in the literature, except for the distribution of PG content in the sagittal direction. CONCLUSION: Poor nutritional supply has a long-term effect on disc degeneration.

The effect of image processing on the detection of cancers in digital mammography.

OBJECTIVE. The objective of our study was to investigate the effect of image processing on the detection of cancers in digital mammography images. MATERIALS AND METHODS. Two hundred seventy pairs of breast images (both breasts, one view) were collected from eight systems using Hologic amorphous selenium detectors: 80 image pairs showed breasts containing subtle malignant masses; 30 image pairs, biopsy-proven benign lesions; 80 image pairs, simulated calcification clusters; and 80 image pairs, no cancer (normal). The 270 image pairs were processed with three types of image processing: standard (full enhancement), low contrast (intermediate enhancement), and pseudo-film-screen (no enhancement). Seven experienced observers inspected the images, locating and rating regions they suspected to be cancer for likelihood of malignancy. The results were analyzed using a jackknife-alternative free-response receiver operating characteristic (JAFROC) analysis. RESULTS. The detection of calcification clusters was significantly affected by the type of image processing: The JAFROC figure of merit (FOM) decreased from 0.65 with standard image processing to 0.63 with low-contrast image processing (p = 0.04) and from 0.65 with standard image processing to 0.61 with film-screen image processing (p = 0.0005). The detection of noncalcification cancers was not significantly different among the image-processing types investigated (p > 0.40). CONCLUSION. These results suggest that image processing has a significant impact on the detection of calcification clusters in digital mammography. For the three image-processing versions and the system investigated, standard image processing was optimal for the detection of calcification clusters. The effect on cancer detection should be considered when selecting the type of image processing in the future.

Objective assessment of cancer genes for drug discovery.

Selecting the best targets is a key challenge for drug discovery, and achieving this effectively, efficiently and systematically is particularly important for prioritizing candidates from the sizeable lists of potential therapeutic targets that are now emerging from large-scale multi-omics initiatives, such as those in oncology. Here, we describe an objective, systematic, multifaceted computational assessment of biological and chemical space that can be applied to any human gene set to prioritize targets for therapeutic exploration. We use this approach to evaluate an exemplar set of 479 cancer-associated genes, reveal the tension between biological relevance and chemical tractability, and describe major gaps in available knowledge that could be addressed to aid objective decision-making. We also propose drug repurposing opportunities and identify potentially druggable cancer-associated proteins that have been poorly explored with regard to the discovery of small-molecule modulators, despite their biological relevance.

Nutrient transport in human annulus fibrosus is affected by compressive strain and anisotropy.

The avascular intervertebral disc (IVD) receives nutrition via transport from surrounding vasculature; poor nutrition is believed to be a main cause of disc degeneration. In this study, we investigated the effects of mechanical deformation and anisotropy on the transport of two important nutrients--oxygen and glucose--in human annulus fibrosus (AF). The diffusivities of oxygen and glucose were measured under three levels of uniaxial confined compression--0, 10, and 20%--and in three directions--axial, circumferential, and radial. The glucose partition coefficient was also measured at three compression levels. Results for glucose and oxygen diffusivity in AF ranged from 4.46 × 10(-7) to 9.77 × 10(-6) cm(2)/s and were comparable to previous studies; the glucose partition coefficient ranged from 0.71 to 0.82 and was also similar to previous results. Transport properties were found to decrease with increasing deformation, likely caused by fluid exudation during tissue compression and reduction in pore size. Furthermore, diffusivity in the radial direction was lower than in the axial or circumferential directions, indicating that nutrient transport in human AF is anisotropic. This behavior is likely a consequence of the layered structure and unique collagen architecture of AF tissue. These findings are important for better understanding nutritional supply in IVD and related disc degeneration.