Biomedical Physics & Engineering Express

BackgroundMammographic (or breast) density is an established risk factor for breast cancer. There are a variety of approaches to measurement including quantitative, semi-automated and automated approaches. We present a new automated measure, AutoCumulus, learnt from applying deep learning to semi-automated measures. MethodsWe used mammograms of 9,057 population-screened women in the BRAIx study for which semi-automated measurements of mammographic density had been made by experienced readers using the CUMULUS software. The dataset was split into training, testing, and validation sets (80%, 10%, 10%, respectively). We applied a deep learning regression model (fine-tuned ConvNeXtSmall) to estimate percentage density and assessed performance by the correlation between estimated and measured percent density and a Bland-Altman plot. The automated measure was tested on an independent CSAW-CC dataset in which density had been measured using the LIBRA software, comparing measures for left and right breasts, sensitivity for high sensitivity, and areas under the receiver operating characteristic curve (AUCs). ResultsBased on the testing dataset, the correlation in percent density between the automated and human measures was 0.95, and the differences were only slightly larger for women with higher density. Based on the CSAW-CC dataset, AltoCumulus outperformed LIBRA in correlation between left and right breast (0.95 versus 0.79; P<0.001), specificity for 95% sensitivity (13% versus 10% (P<0.001)), and AUC (0.638 cf. 0.597; P<0.001). ConclusionWe have created an automated measure of mammographic density that is accurate and gives superior performance on repeatability within a woman, and for prediction of interval cancers, than another well-established automated measure.

BackgroundThe most used instruments for small-field dosimetry have notable limitations, including the need for correction factors, limited scanning speeds, and challenges in alignment for percentage depth dose (PDD) measurements, particularly for extremely small-fields. However, plastic scintillation detectors (PSDs) are an attractive alternative for small-field dosimetry due to their correction-free nature, linear dose response, and fast response time. PurposeThis study evaluates the robustness and accuracy of the dosimetric measurements using a new water-equivalent PSD in small-field dosimetry. The study also aims to report an indirect method for measuring PDD in small-fields, with a scanning time that is 5 to 10 times faster than traditional methods. MethodPDDs, profiles and output factors were measured on a Varian TrueBeam 6XFFF photon beam for the field size of 0.5x0.5 cm2, 1x1 cm2, 2x2 cm2, 3x3 cm2, 4x4 cm2 using a new PSD from Blue Physics (BP-PSD). These measurements were compared with those obtained using a well-established PSD (Standard Imaging W2), micro-diamond (TN60019, PTW-Freiburg, Germany), and micro-silicon detectors (TN60023, PTW-Freiburg, Germany). Owing to its fast response, the BP-PSD enabled the collection of beam profiles at 31 depths, which were used to derive the PDD while avoiding detector misalignment along the beam path. Data was collected in a water tank controlled by the PTW BeamScan software. The pulse-by-pulse raw data from BP-PSD were converted to respective dosimetry data using in-house software. ResultThe BP-PSD demonstrated excellent agreement with other detectors for small-field output factors (FOFs), with a maximum variation of 1.6%. The BP-PSD also showed strong agreements in PDD measurements with an ion chamber (TN31013) for both 3x3 cm2 and 10x10 cm2 field sizes, achieving a 98% gamma passing rate (gamma criteria: 1mm,3%). For the profile measurements, the BP-PSD showed consistency with both the micro-diamond and micro-silicon diode detectors, with less than 1% variation in measured penumbra length. At a 3x3 cm2 field size, the measured penumbra length (4 mm) agreed with previously published data (3.86-4.2 mm). Additionally, for field size less than 3x3 cm2 the indirect PDD measurements derived from profiles showed significant improvement compared to the direct measurements using various detectors, using TPS-calculated PDD as a reference. ConclusionThe BP-PSD has proven to be a robust and reliable detector for small-field dosimetry. It exhibits excellent agreement with other detectors in measuring small FOFs and provides accurate measurements with significantly faster scanning speeds in a water tank. The fast response feature enables the indirect PDD measurement method

This study aims to assess women's perceptions of artificial intelligence (AI) used in breast cancer screening in France by examining their knowledge of AI and the barriers to their participation in organized screening. The results of a survey conducted in June 2025 among a national sample of 2000 women (aged 40-75) reveal limited participation and persistent concerns among women. Nevertheless, despite a low awareness of specific AI applications, a large majority of the women surveyed are very favorable to the use of AI in breast cancer diagnosis, even considering it a lever to increase screening participation.

PurposeBreast density is associated with risk of developing cancer and can be automatically estimated, using deep learning models, from digital mammograms. Our aim is to estimate the capacity and reliability of such models to estimate density from low dose mammograms taken to enable risk estimates for younger women. MethodsWe trained deep learning models on standard and simulated low dose mammograms. The models were then tested on a mammography data-set with paired standard and low-dose image. The effect of different factors (including age, density and dose ratio) on the differences between predictions on standard and low dose are analysed. Methods to improve performance are assessed and factors that reduce model quality are demonstrated. ResultsWe showed that whilst many factors have no significant effect on the quality of low dose density prediction both density and breast area have an impact. For example correlation between density predictions on low and standard dose images of breasts with the largest breast area is 0.985 (0.949-0.995) while with the smallest is 0.882 (0.697-0.961). We also demonstrated that averaging across CC-MLO images and across repeatedly trained models can improve predictive performance. ConclusionLow dose mammography can be used to produce density and risk estimates that are comparable to standard dose images. Averaging across CC-MLO and across model predictions should improve this performance. Model quality is reduced when making predictions on denser and smaller breasts. Code is available at: https://github.com/stevensquires/

ObjectiveTo evaluate the potential diagnostic improvement accrued from using the novel 3D breast imaging technology of Multimodal UltraSonic Tomography (MUST) as adjunct to digital mammography for BI-RADS 4 patients referred to biopsy (clinical trial # EUDAMED/CIV-ID CIV-GR-22-05-039513). MethodsMUST generates 3D tomographic images of pendant breast in water-bath using transmission-mode ultrasound measurements of acoustic refractivity and frequency-dependent attenuation. These measurements are fused via a properly developed algorithm into "advisory diagnostic images" (ADI) depicting the likelihood of malignancy at each voxel of the entire breast volume. In this clinical trial, MUST imaging was performed prior to biopsy on 207 BI-RADS 4 patients presenting micro-calcifications in mammography. The findings of the MUST ADI in the biopsy region were evaluated against the biopsy results ResultsBiopsy histopathology identified malignant lesions in 54 patients (26.2%). MUST ADI detected correctly all these malignancies (down to 2 mm in maximum dimension). In 31 of the 153 participants with negative biopsy (20.3%), MUST ADI found some "likely malignant" lesions within 20-mm radius from the putative point of biopsy. Breast density varied across the cohort, with 65.2% having dense breasts (ACR score 3-4). No dependence of MUST diagnostic performance on breast density was found in this cohort. ConclusionMUST imaging detected correctly all 54 biopsy-confirmed malignant breast lesions (down to 2 mm) among 207 BI-RADS 4 participants (NPV = 100%), while detecting "likely malignancy" in the biopsy region of 31 participants with negative biopsy (PPV = 63.5%), irrespective of breast density. Key messages- MUST detected all 54 biopsy-confirmed malignant lesions in 207 BI-RADS 4 patients (NPV=100%) - MUST detected some malignancy in the region of biopsy in 31 cases with negative biopsy (PPV=63.5%) - The diagnostic performance of MUST was independent of breast density

ObjectivesHigh mammographic density (MD) and excess weight are associated with increased risk of breast cancer. Weight loss interventions could reduce risk, but classically defined percentage density measures may not reflect this due to disproportionate loss of breast fat. We investigate an artificial intelligence-based density method, reporting density changes in 46 women enrolled in a weight-loss study in a family history breast cancer clinic, using a volumetric density method as a comparison. MethodsWe analysed data from women who had weight recorded and mammograms taken at the start and end of the 12-month weight intervention study. MD was assessed at both time points using a deep learning model, pVAS, trained on expert estimates of percent density, and Volpara density software. ResultsThe Spearman rank correlation between reduction in weight and change in density was 0.17 (-0.13 to 0.43) for pVAS and 0.59 (0.36 to 0.75) for Volpara volumetric percent density. ConclusionspVAS percent density measurements were not significantly affected by change in weight. Percent density measured with Volpara increased as weight decreased, driven by changes in fat volume. Advances in knowledgeThe effect of weight change on pVAS mammographic density predictions has not previously been published.

BackgroundBreast cancer is one of the leading causes of death among women worldwide. Therefore, early detection through quality mammography, along with information collected from the mammographic survey, is crucial. The objective of this research is to propose a mammographic survey for patients prior to mammography to gather important information for radiologists specializing in mammography when preparing reports. MethodologyThis research was conducted using a semi-structured survey with both closed and open-ended questions, which was completed by 31 radiologists specializing in mammography. ResultsRelevant patient history and risk factors were collected, which are most useful for radiological reporting concerning patient history and potential risk factors. ConclusionBased on the responses collected from radiologists, we propose a final mammographic survey as a tool to aid radiologists in correlating clinical images with patient history. This survey emphasizes family history, gynecological-obstetric history, surgical history, among other factors.

Rationale and objectivesBreast cancer represents the leading cause of death from cancer in women worldwide. Early detection of breast tumours improves the prognosis and survival rate. Propagation-based phase-contrast computed tomography (PB-CT) is a technique that uses refraction and absorption of the X-ray to produce images for clinical applications. This study compared the performance of photon-counting and flat-panel X-ray detectors in PB-CT breast imaging using synchrotron radiation. Materials and methodsMastectomy specimens underwent PB-CT imaging using the Hamamatsu C10900D Flat Panel and PIXIRAD-8 CdTe single-photon-counting detectors. PB-CT images generated at different imaging conditions were compared to absorption-based CT (AB-CT) reference images acquired with the same detectors to investigate the image quality improvement delivered by PB-CT relative to AB-CT. The image quality of the different image sets was assessed by eleven readers in a visual grading characteristics (VGC) study. ResultsThe intraclass correlation coefficient showed a moderate/good interobserver agreement for the image set analysed (ICC = 0.626, p = <0.001). The area under the curve showed that the image quality improvement in PB-CT images obtained by the PIXIRAD-8 CdTe single-photon-counting detector were consistently higher than the one for flat-panel Hamamatsu detector. The level of improvement in image quality was more substantial at lower radiation doses. ConclusionIn this study, the PIXIRAD-8 photon-counting detector was associated with higher image quality scores at all tested radiation dose levels, which was likely a result of the combined effect of the absence of dark current noise and better spatial resolution, compared to the flat-panel detector.

ObjectiveTo evaluate the performance of Thermalytix, an artificial intelligence-enhanced breast thermal imaging analysis software, against unaided manual interpretation of thermographic images. MethodsIn this multi-reader study, thermal imaging data of 258 symptomatic participants from a previous clinical trial were used. These images were independently manually interpreted by 3 senior trained breast radiologists. The same images were independently evaluated by Thermalytix, which uses sophisticated machine learning analysis of thermal/ vascular radiomic parameters to generate a risk score predictive of cancer . The results of manual interpretation and Thermalytix were compared with reference standard based on standard of care (combination of mammography, ultrasound and histopathology), to determine sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) and area under receiver operating characteristic curves (AUROC). ResultsThermalytix obtained showed a sensitivity and specificity of 95.2% (90% confidence interval (CI), 90.0- 100.0) and 66.7% (CI 60.1-73.3); the NPV and PPV were 97.7% (CI 95.2%-100.3%) and 58.3% (CI 48.5%-68.2%). The (sensitivity, specificity, NPV, PPV) obtained by Reader 1, Reader 2 and Reader 3 were (60.3%, 81.5%, 51.4%, 86.4%), (74.6%,50.8%, 86.1%, 32.9%) and (71.4%, 63.8%, 87.2%, 38.5%), respectively. The AUROC of Thermalytix was 0.85, 13.7% greater than manual interpretation. ConclusionThermalytix demonstrated good clinical performance with 25% higher accuracy than manual interpretation of thermal images. Thermalytix may alleviate the known subjectivity in thennography thereby improving its performance.

The COVID-19 [SARS-COV-2] pandemic has had a devastating global impact, with both the human and socio economic costs being severe. One result of the COVID-19 pandemic is the emergence of an urgent requirement for effective techniques and technologies for screening individuals showing symptoms of infection in a non-invasive and non-contact way. Systems that exploit thermal imaging technology to screen individuals show promise to satisfy the desired criteria, including offering a non-contact, non-invasive method of temperature measurement. Furthermore, the potential for mass and passive screening makes thermal imaging systems an attractive technology where current standard of care methods are not practical. Critically, any fever screening solution must be capable of accurate temperature measurement and subsequent prediction of core temperature. This is essential to ensure a high sensitivity in identifying fever while maintaining a low rate of false positives. This paper discusses the results and analysis of a clinical trial undertaken by Thales UK Ltd and the Queen Elizabeth University Teaching Hospital in Glasgow to assess the accuracy and operation of the High Temperature Detection (HTD) system developed by Thales UK Ltd when used in a clinical setting. Results of this single centre prospective observational cohort study show that the measured laboratory accuracy of the Thales HTD system (RMSE=0.1{degrees}C) is comparable to the accuracy when used in a clinical setting (RMSE = 0.15{degrees}C) when measuring a calibrated blackbody source at typical skin temperature. For measurement of forehead skin temperature, the system produced results commensurate with close contact measurement methods (R = 0.86, Mean error=0.05{degrees}C).. Compared to measured tympanic temperatures, measurement of the forehead skin temperature by the HTD system showed a moderate correlation (R = 0.43),), which is stronger than close contact IR forehead thermometers (R = 0.20,0.35) An improved correlation was observed between the maximum facial temperature measured by the HTD system and measured tympanic temperatures (R = 0.53), which is significantly stronger than the close contact methods. A linear predictive model for tympanic temperature based on the measured maximum facial temperatures resulted in a root mean square error (RMSE = 0.50{degrees}C) that is marginally larger than what is expected as a compound of errors in the measuring devices (RMSE=0.45{degrees}C). The study demonstrates that the HTD could be applied in the clinical and non-clinical setting as a screening mechanism to detect citizens with raised temperature. This approach would enable high volume surveillance and identification of individuals that contribute to further spread of COVID-19. Deployment of the HTD system could be implemented as part of a screening tool to support measures to enhance public safety and confidence in areas of high throughput, such as airports, shopping centres or places of work.

ObjectivesMammographic density is associated with increased risk of developing breast cancer. Automated estimation of density in women below normal screening age would enable earlier risk stratification. We are piloting the use of low dose mammograms combined with models that can make accurate mammographic density estimates. MethodsThree models were trained on a joint set (107,619) of standard dose mammograms with associated density scores and their simulated low dose counterparts such that the models made predictions on standard and low dose mammograms. A second set of models was trained separately on the standard and simulated low dose mammograms. All models were tested on a held-out set from the training data and an independent dataset with 294 pairs of standard and real low dose mammograms. ResultsThe root mean squared errors (RMSE) between the model predictions and density scores on standard and simulated low dose images were 8.26 (8.16-8.36) and 8.27 (8.17-8.38) respectively. The RMSE between predictions on standard and simulated low dose images for the jointly trained models was 1.91 (1.88-1.96). The RMSE of the predictions on the real low dose images compared to the standard dose images is 3.79 (2.75-4.99). ConclusionsDeep learning models make density predictions on low dose images with similar quality as on standard dose images. Such automated analysis of low dose mammograms could contribute to accurate breast cancer risk estimation in younger women enabling stratification for further monitoring and preventative therapy. Advances in knowledgeMammographic density can be estimated in low dose mammograms with similar quality to standard dose mammograms.

PurposeProstate imaging to guide biopsy remains unsatisfactory, with current solutions suffering from high complexity and poor accuracy and reliability. One novel entrant into this field is microultrasound (microUS), which uses a high frequency imaging probe to achieve very high spatial resolution, and achieves prostate cancer detection rates equivalent to multiparametric magnetic resonance imaging (mpMRI). However, the ExactVu transrectal microUS probe has a unique geometry that makes it challenging to acquire controlled, repeatable three-dimensional (3D) transrectal ultrasound (TRUS) volumes. We describe the design, fabrication, and validation of a 3D acquisition system that allows for the accurate use of the ExactVu microUS device for volumetric prostate imaging. MethodsThe design uses a motorized, computer-controlled brachytherapy stepper to rotate the ExactVu transducer about its axis. We carry geometric validation using a phantom with known dimensions and we compare performance with magnetic resonance imaging (MRI) using a commercial quality assurance anthropomorphic prostate phantom. ResultsOur geometric validation shows accuracy of 1 mm or less in all three directions, and images of an anthropomorphic phantom qualitatively match those acquired using MRI and show good agreement quantitatively. ConclusionWe describe the first system to acquire robotically- controlled 3D microUS images using the ExactVu microUS system. The reconstructed 3D microUS images are accurate, which will allow for applications of the ExactVu microUS system in prostate specimen and in vivo imaging.

OverviewA recent study assessing the sporicidal efficacy of UV-C LED high-level disinfection (HLD) system for ultrasound probe reprocessing concluded that UV-C LED disinfection underperforms compared to hydrogen peroxide (H2O2) mist-based devices (Nanosonics: TrophonEPR and Trophon2). However, our independent secondary testing of a UV-C LED disinfection system (Lumicare ONE(R)) under clinically relevant conditions yielded contrasting results, demonstrating equivalent efficacy to H2O2 mist-based devices.1-7 Using the same biological indicators (BIs) (106 Geobacillus stearothermophilus spores, Bionova(R) BT93/6) on stainless steel coupons, we conducted three validation conditionsO_LIBI in glassine packaging/Tyvek(R) (non-clinical scenario) C_LIO_LIBI with packaging removed (bare metal coupon) C_LIO_LIBI with packaging removed and flame-sterilized metal coupon (This validation process has a potential to simulate heat-induced surface changes to the metal coupon) C_LI The Lumicare ONE(R) UV-C LED HLD system achieved a 100% pass rate under both clinically relevant conditions (bare metal and flame-sterilized BIs), while under the non-clinical scenario BI in glassine packaging/Tyvek(R) test type was removed since UV-C is unable to penetrate packaging materials. This paper also identifies methodological concerns in the referenced study, includingO_LIThe use of an inappropriate BI (Bionova(R) BT93), explicitly validated for hydrogen peroxide (H2O2) based sterilization to demonstrate penetration of Tyvek(R) peel packaging but not UV-C applications, as per the manufacturers guidelines. C_LIO_LIA misrepresentation of Therapeutic Goods Administration (TGA) sporicidal efficacy requirements. C_LI The Lumicare ONE(R) system goes beyond TGA mandatory requirements in completing a Carrier Test using Geobacillus stearothermophilus spores. The Lumicare ONE(R) system reduced the numbers of Geobacillus stearothermophilus ATC 7953 by greater than a > 6 log10, thus meeting the pass criteria of the Association of Official Analytical Chemists (AOAC 966.04). The Lumicare ONE(R) system has independently demonstrated a greater than > 6-log10 reduction of TGA-mandated test organisms, including Clostridium sporogenes and Bacillus subtilis, and is registered on the Australian Register of Therapeutic Goods (ARTG) as a hospital-grade disinfectant. Our findings highlight the importance of proper study design, appropriate BI selection, and regulatory alignment when evaluating emerging disinfection technologies. UV-C LED systems represent a validated, effective alternative to hydrogen peroxide (H2O2) mist-based devices in clinical ultrasound probe reprocessing.

IntroductionWhile 2D mammography is the standard for breast cancer screening, its sensitivity is reduced in dense breasts, impacting early detection and extent assessment. This study, for the first time in New Zealand, evaluated the utility of supplementary multi-modality imaging (tomosynthesis, ultrasound, and MRI) in a risk-stratified population. MethodsA retrospective case study (May 2022 - September 2023, New Zealand private clinic) analysed patients by Tyrer-Cuzick (TC) v8 lifetime risk and by Volpara density categories. All patients in the screening pathway (n = 2171) underwent 2D mammography and tomosynthesis. Those patients with high breast density received supplementary ultrasound, and those with TC8 risk scores of [&ge;] 30% were offered abbreviated MRI. Symptomatic patients (n = 230) underwent standard diagnostic workup. Detection rates and extent of disease using multimodality imaging were compared. ResultsOf the 2401 patients, 205 were high-risk criteria ([&ge;]30%) and 19 breast cancers (16 invasive, 3 DCIS only) were diagnosed. Tomosynthesis identified 38% (6/16) more invasive cancers than 2D mammography alone. Ultrasound and MRI detected an additional 27% (4/16) invasive cancers occult on other modalities, predominantly in those women with density D breasts. Ultrasound and particularly MRI demonstrated superior accuracy in assessing disease extent, including identifying multifocal and multicentric disease that was not detected by 2D or Tomosynthesis. ConclusionSupplementary screening modalities, particularly MRI, significantly improve breast cancer detection and assessment of disease extent in high-risk women. These findings support a personalized screening approach integrating risk assessment and breast density to guide imaging modality selection.

BackgroundEarly detection of breast cancer reduces the risk of mortality, with mammography being the gold standard for detecting breast cancer. Currently, there is no consensus on breast screening in transgender patients, with only recommendations from various international health organizations, which differ from one another. The objective of this research was to determine the mammographic projections applicable to transgender patients, in order to propose a mammography protocol and a pre-mammography survey focused on this patient group. MethodologyA semi-structured survey with both closed and open-ended questions was conducted, which was answered by radiologists specializing in mammography. To propose mammographic projections, the survey included questions about the transitional phase, hormone use, surgeries, and breast tissue development. To propose a pre-mammography survey for transgender patients, questions about risk factors and surgical treatments were included. ConclusionsA mammographic protocol is proposed according to the transitional phase, along with a pre-mammography survey for transgender patients, based on information gathered from a survey answered by 20 radiologists specializing in mammography. This contributes significantly to the screening process for this patient group.

BackgroundAccurate radiology disease detection relies on high-resolution, low-dose imaging, yet current systems frequently fail to identify small lesions early enough to alter prognosis while minimizing radiation exposure. This occurs because existing detector architectures cannot maintain high quantum efficiency at small pixel sizes without increasing radiation output. Radiography (70-100 {micro}m), mammography (50-100 {micro}m), CT (250-500 {micro}m), and cone-beam CT (80 {micro}m) detectors are constrained by pixel size, resulting in limited modulation transfer function (MTF), noise power spectrum (NPS), and relative detective quantum efficiency (rDQE). These limitations contribute to 56% of diagnostic errors and require higher milliampere (mA) settings. The objective of this study was to compare a novel CMOS size-2 intraoral X-ray detector prototype with DR, mammography, CT, and CBCT detectors in terms of dose efficiency and spatial resolution (lp/mm), and to evaluate the feasibility of ultra-high-resolution, low-dose X-ray imaging for medical radiology applications. MethodsWe developed and evaluated a complementary metal-oxide semiconductor (CMOS) size-2 intraoral dental and small-field detector prototype comprising a novel back-illuminated (BI) pixel architecture and microlenses (M) to reduce pixel size to 8 {micro}m, enabling low-dose acquisition. MTF, NPS, and rDQE were benchmarked against published data for radiography, mammography, CT, and CBCT in terms of dose efficiency and spatial resolution in line pairs per millimeter (lp/mm). A microfocus X-ray source was used at 70 kVp, 0.3 mA, 20 {micro}m focal spot, 5 cm source-to-detector distance, and 0.25 s exposure time. ResultsThe delivered dose rate during prototype acquisitions was 83.6 mGy/s at a 5 cm source-to-detector distance, corresponding to an air kerma of 20.9 mGy per image for a 0.25 s exposure (0.075 mAs). Inverse-square scaling indicates that incident air kerma would decrease at typical clinical distances once exposure settings are adjusted to maintain signal and image quality. These measurements therefore demonstrated that, at 0.3 mA and a short geometry, the 8 {micro}m BI-M design improved image resolution to 30 lp/mm, exceeding the 5-10 lp/mm limits of existing modalities, while operating at a radiation output several-fold lower in mAs than current systems, providing dose-efficiency headroom for future clinical configurations. Additional system-level analysis demonstrated that microlenses further improved the measured system MTF across clinically relevant spatial frequencies under scintillator-coupled, low-dose imaging conditions. ConclusionAlthough originally developed for intraoral imaging, the novel 8 {micro}m BI prototype detectors pixel architecture, combined with microlenses and a microfocus X-ray source, enabled higher dose efficiency and spatial resolution compared with radiography, mammography, CT, and CBCT detectors. It surpassed MTF, NPS, and rDQE performance at a lower dose, highlighting its potential for early-stage disease detection with reduced radiation burden in microfocus clinical applications. The 8 {micro}m BI-M pixel architecture broke the traditional dose-resolution trade-off by preserving quantum efficiency at small pixel size and enabling microfocus-based clinical imaging at low mA. These technological improvements are particularly relevant for cancer detection, pediatric imaging, and thin-anatomy applications.

PurposeMammographic density is associated with the risk of developing breast cancer and can be predicted using deep learning methods. Model uncertainty estimates are not produced by standard regression approaches but would be valuable for clinical and research purposes. Our objective is to produce deep learning models with in-built uncertainty estimates without degrading predictive performance. ApproachWe analyse data from over 150,000 mammogram images with associated continuous density scores from expert readers in the Predicting Risk Of Cancer At Screening (PROCAS) study. We re-designate the continuous density scores to 100 density classes then train classification and ordinal deep learning models. Distributions and distribution-free methods are applied to extract predictions and uncertainties. A deep learning regression model is trained on the continuous density scores to act as a direct comparison. ResultsThe root mean squared error (RMSE) between expert assigned density labels and predictions of the standard regression model are 8.42 (8.34-8.51) while the RMSE for the classification and ordinal classification are 8.37 (8.28-8.46) and 8.44 (8.35-8.53) respectively. The average uncertainties produced by the models are higher when the density scores from pairs of expert readers density scores differ more, are higher when different mammogram views of the same views are more variable and when two separately trained models show higher variation. ConclusionsUsing either a classification or ordinal approach we can produce model uncertainty estimates without loss of predictive performance.

Breast cancer, a leading cause of cancer-related deaths in women, presents a growing challenge in medical diagnostics. Despite the effectiveness of mammography and ultrasound, the ambiguity in non-invasive scans often necessitates invasive procedures. Our primary goal was to create an AI model that could predict breast cancer with high negative predictive value and reduce unnecessary procedures. This study introduces the Retrieval-Augmented Medical Diagnosis System (RAMDS) for breast cancer, a novel approach combining an AI model with a retrieval-augmented mechanism to enhance diagnostic accuracy and explainability. The RAMDS employs a pretrained ResNet 34 model, fine-tuned on breast ultrasound image datasets from four countries. It integrates a similarity-based weighted adjustment mechanism to compare new cases with historical diagnoses. Its like having an experienced doctor who remembers every case theyve ever seen and uses that knowledge to make better decisions. RAMDS improved sensitivity by 11%, and negative predictive value by 9% when compared to the base model. Notably, the RAMDS improves explainability by linking AI predictions to similar historical cases, aligning with the medical communitys interest in transparent and understandable AI decisions. A unique feature of this system is its adaptability to varied imaging contexts without retraining, addressing the challenge of dataset variability across medical institutions. In conclusion, the RAMDS offers a significant advancement in breast cancer diagnosis, combining enhanced accuracy, explainability, and adaptability. It holds promise for clinical application, though further research is needed to optimize its performance and integrate multi-modal data.

BackgroundWhile mammography is commonly used for breast cancer detection, its widespread implementation in resource-constrained nations is challenging. Artificial intelligence-based Thermalytix is a low-cost, portable, radiation-free, automated test for breast cancer detection in women of all ages. Although used in India, the efficacy of Thermalytix has not been tested in an African population. ObjectivesTo assess the agreement and correlation coefficient of Thermalytix output with radiologist-reported mammography, in a Zambian tertiary care population. MethodologyIn October 2023, 169 women were evaluated with both Thermalytix and standard mammography at Maina Soko Military Hospital at Lusaka. Thermalytix uses advanced machine learning algorithms to interpret breast thermal scans and generates a quantitative score indicating the likelihood of malignancy. All women underwent both tests, with results blinded both ways. Subsequently the Spearman correlation coefficient and level of agreement between Thermalytix output and BIRADs scoring from radiologist-interpreted mammography was calculated. Results144 women with complete data were analysed in this report, with median age of 50 years (53.5% postmenopausal, 65.3% asymptomatic). Six women were assessed as mammography test positive and 138 as mammography negative; in these, the correlation between Thermalytix and mammography using Spearman test of rank correlation was 0.9 [very strong], and using the US FDA recommended test of agreement, positive agreement was obtained in 83.3%. ConclusionDemonstrating a very strong correlation and level of agreement with mammography, along with its good sensitivity, specificity and negative predictive value in previous clinical trials, Thermalytix has the potential to be an additional tool in the early detection of breast cancer in Zambia.

With the emergence of electroencephalography (EEG) as a tool in the cognitive domain, new demands are being placed on the technology to keep up with functional applications, especially in the context of at-home neural monitoring. New use cases have fostered development of wearable EEG (wEEG) devices: portable, low-cost headsets used for EEG monitoring. This evolution of technology and application has not been accompanied by development in technology evaluation, often relying on function-agnostic markers to assess devices for efficacy in this new space. With current methods limited in scope, this study designed, tested and evaluated a novel functionally-focused comparative protocol for wEEG devices. Eight participants undertook a protocol for the evaluation of four established wEEG devices, assessing cognitive resolution and general usability. Compared to a well-established traditional analysis method (eyes open/eyes closed protocol), the novel design proposed here enabled the same analysis of headset resolution, while also providing additional context into user preferences and opening downstream possibilities for specific cognitive insights. Future research could enable the development of this protocol into a standardised method to ensure the performance of wEEG technology can satisfy emerging clinical needs.