ABSTRACT
Breast cancer remains a leading cause of mortality among women, with molecular subtypes significantly influencing prognosis and treatment strategies. Currently, identifying the molecular subtype of cancer requires a biopsy-a specialized, expensive, and time-consuming procedure, often yielding to results that must be supported with additional biopsies due to technique errors or tumor heterogeneity. This study introduces a novel approach for predicting breast cancer molecular subtypes using mammography images and advanced artificial intelligence (AI) methodologies. Using the OPTIMAM imaging database, 1397 images from 660 patients were selected. The pretrained deep learning model ResNet-101 was employed to classify tumors into five subtypes: Luminal A, Luminal B1, Luminal B2, HER2, and Triple Negative. Various classification strategies were studied: binary classifications (one vs. all others, specific combinations) and multi-class classification (evaluating all subtypes simultaneously). To address imbalanced data, strategies like oversampling, undersampling, and data augmentation were explored. Performance was evaluated using accuracy and area under the receiver operating characteristic curve (AUC). Binary classification results showed a maximum average accuracy and AUC of 79.02% and 64.69%, respectively, while multi-class classification achieved an average AUC of 60.62% with oversampling and data augmentation. The most notable binary classification was HER2 vs. non-HER2, with an accuracy of 89.79% and an AUC of 73.31%. Binary classification for specific combinations of subtypes revealed an accuracy of 76.42% for HER2 vs. Luminal A and an AUC of 73.04% for HER2 vs. Luminal B1. These findings highlight the potential of mammography-based AI for non-invasive breast cancer subtype prediction, offering a promising alternative to biopsies and paving the way for personalized treatment plans.
ABSTRACT
From an early age, people are exposed to risk factors that can lead to musculoskeletal disorders like low back pain, neck pain and scoliosis. Medical screenings at an early age might minimize their incidence. The study intends to improve a software that processes images of patients, using specific anatomical sites to obtain risk indicators for possible musculoskeletal problems. This project was divided into four phases. First, markers and body metrics were selected for the postural assessment. Second, the software's capacity to detect the markers and run optimization tests was evaluated. Third, data were acquired from a population to validate the results using clinical software. Fourth, the classifiers' performance with the acquired data was analyzed. Green markers with diameters of 20 mm were used to optimize the software. The postural assessment using different types of cameras was conducted via the blob detection method. In the optimization tests, the angle parameters were the most influenced parameters. The data acquired showed that the postural analysis results were statistically equivalent. For the classifiers, the study population had 16 subjects with no evidence of postural problems, 25 with mild evidence and 16 with moderate-to-severe evidence. In general, using a binary classification with the train/test split validation method provided better results.
ABSTRACT
Ultrasound (US) imaging is used in the diagnosis and monitoring of COVID-19 and breast cancer. The presence of Speckle Noise (SN) is a downside to its usage since it decreases lesion conspicuity. Filters can be used to remove SN, but they involve time-consuming computation and parameter tuning. Several researchers have been developing complex Deep Learning (DL) models (150,000-500,000 parameters) for the removal of simulated added SN, without focusing on the real-world application of removing naturally occurring SN from original US images. Here, a simpler (<30,000 parameters) Convolutional Neural Network Autoencoder (CNN-AE) to remove SN from US images of the breast and lung is proposed. In order to do so, simulated SN was added to such US images, considering four different noise levels (σ = 0.05, 0.1, 0.2, 0.5). The original US images (N = 1227, breast + lung) were given as targets, while the noised US images served as the input. The Structural Similarity Index Measure (SSIM) and Peak Signal-to-Noise Ratio (PSNR) were used to compare the output of the CNN-AE and of the Median and Lee filters with the original US images. The CNN-AE outperformed the use of these classic filters for every noise level. To see how well the model removed naturally occurring SN from the original US images and to test its real-world applicability, a CNN model that differentiates malignant from benign breast lesions was developed. Several inputs were used to train the model (original, CNN-AE denoised, filter denoised, and noised US images). The use of the original US images resulted in the highest Matthews Correlation Coefficient (MCC) and accuracy values, while for sensitivity and negative predicted values, the CNN-AE-denoised US images (for higher σ values) achieved the best results. Our results demonstrate that the application of a simpler DL model for SN removal results in fewer misclassifications of malignant breast lesions in comparison to the use of original US images and the application of the Median filter. This shows that the use of a less-complex model and the focus on clinical practice applicability are relevant and should be considered in future studies.
ABSTRACT
Glioblastoma (GB) is a malignant glioma associated with a mean overall survival of 12 to 18 months, even with optimal treatment, due to its high relapse rate and treatment resistance. The standardized first-line treatment consists of surgery, which allows for diagnosis and cytoreduction, followed by stereotactic fractionated radiotherapy and chemotherapy. Treatment failure can result from the poor passage of drugs through the blood-brain barrier (BBB). The development of novel and more effective therapeutic approaches is paramount to increasing the life expectancy of GB patients. Nanoparticle-based treatments include epitopes that are designed to interact with specialized transport systems, ultimately allowing the crossing of the BBB, increasing therapeutic efficacy, and reducing systemic toxicity and drug degradation. Polymeric nanoparticles have shown promising results in terms of precisely directing drugs to the brain with minimal systemic side effects. Various methods of drug delivery that pass through the BBB, such as the stereotactic injection of nanoparticles, are being actively tested in vitro and in vivo in animal models. A significant variety of pre-clinical studies with polymeric nanoparticles for the treatment of GB are being conducted, with only a few nanoparticle-based drug delivery systems to date having entered clinical trials. Pre-clinical studies are key to testing the safety and efficacy of these novel anticancer therapies and will hopefully facilitate the testing of the clinical validity of this promising treatment method. Here we review the recent literature concerning the most frequently reported types of nanoparticles for the treatment of GB.
ABSTRACT
Currently, breast cancer is the most commonly diagnosed type of cancer worldwide. Digital Breast Tomosynthesis (DBT) has been widely accepted as a stand-alone modality to replace Digital Mammography, particularly in denser breasts. However, the image quality improvement provided by DBT is accompanied by an increase in the radiation dose for the patient. Here, a method based on 2D Total Variation (2D TV) minimization to improve image quality without the need to increase the dose was proposed. Two phantoms were used to acquire data at different dose ranges (0.88-2.19 mGy for Gammex 156 and 0.65-1.71 mGy for our phantom). A 2D TV minimization filter was applied to the data, and the image quality was assessed through contrast-to-noise ratio (CNR) and the detectability index of lesions before and after filtering. The results showed a decrease in 2D TV values after filtering, with variations of up to 31%, increasing image quality. The increase in CNR values after filtering showed that it is possible to use lower doses (-26%, on average) without compromising on image quality. The detectability index had substantial increases (up to 14%), especially in smaller lesions. So, not only did the proposed approach allow for the enhancement of image quality without increasing the dose, but it also improved the chances of detecting small lesions that could be overlooked.
ABSTRACT
Glioblastoma multiforme (GBM) remains a challenging disease, as it is the most common and deadly brain tumour in adults and has no curative solution and an overall short survival time. This incurability and short survival time means that, despite its rarity (average incidence of 3.2 per 100,000 persons), there has been an increased effort to try to treat this disease. Standard of care in newly diagnosed glioblastoma is maximal tumour resection followed by initial concomitant radiotherapy and temozolomide (TMZ) and then further chemotherapy with TMZ. Imaging techniques are key not only to diagnose the extent of the affected tissue but also for surgery planning and even for intraoperative use. Eligible patients may combine TMZ with tumour treating fields (TTF) therapy, which delivers low-intensity and intermediate-frequency electric fields to arrest tumour growth. Nonetheless, the blood-brain barrier (BBB) and systemic side effects are obstacles to successful chemotherapy in GBM; thus, more targeted, custom therapies such as immunotherapy and nanotechnological drug delivery systems have been undergoing research with varying degrees of success. This review proposes an overview of the pathophysiology, possible treatments, and the most (not all) representative examples of the latest advancements.
ABSTRACT
Breast cancer was the most diagnosed cancer around the world in 2020. Screening programs, based on mammography, aim to achieve early diagnosis which is of extreme importance when it comes to cancer. There are several flaws associated with mammography, with one of the most important being tissue overlapping that can result in both lesion masking and fake-lesion appearance. To overcome this, digital breast tomosynthesis takes images (slices) at different angles that are later reconstructed into a 3D image. Having in mind that the slices are planar images where tissue overlapping does not occur, the goal of the work done here was to develop a deep learning model that could, based on the said slices, classify lesions as benign or malignant. The developed model was based on the work done by Muduli et. al, with a slight change in the fully connected layers and in the regularization done. In total, 77 DBT volumes-39 benign and 38 malignant-were available. From each volume, nine slices were taken, one where the lesion was most visible and four above/below. To increase the quantity and the variability of the data, common data augmentation techniques (rotation, translation, mirroring) were applied to the original images three times. Therefore, 2772 images were used for training. Data augmentation techniques were then applied two more times-one set used for validation and one set used for testing. Our model achieved, on the testing set, an accuracy of 93.2% while the values of sensitivity, specificity, precision, F1-score, and Cohen's kappa were 92%, 94%, 94%, 94%, and 0.86, respectively. Given these results, the work done here suggests that the use of single-slice DBT can compare to state-of-the-art studies and gives a hint that with more data, better augmentation techniques and the use of transfer learning might overcome the use of mammograms in this type of studies.
Subject(s)
Breast Neoplasms , Mammography , Humans , Female , Mammography/methods , Breast Neoplasms/diagnostic imaging , Neural Networks, Computer , Imaging, Three-Dimensional/methods , Radiographic Image Interpretation, Computer-Assisted/methodsABSTRACT
Breast cancer was the most diagnosed cancer in 2020. Several thousand women continue to die from this disease. A better and earlier diagnosis may be of great importance to improving prognosis, and that is where Artificial Intelligence (AI) could play a major role. This paper surveys different applications of AI in Breast Imaging. First, traditional Machine Learning and Deep Learning methods that can detect the presence of a lesion and classify it into benign/malignant-which could be important to diminish reading time and improve accuracy-are analyzed. Following that, researches in the field of breast cancer risk prediction using mammograms-which may be able to allow screening programs customization both on periodicity and modality-are reviewed. The subsequent section analyzes different applications of augmentation techniques that allow to surpass the lack of labeled data. Finally, still concerning the absence of big datasets with labeled data, the last section studies Self-Supervised learning, where AI models are able to learn a representation of the input by themselves. This review gives a general view of what AI can give in the field of Breast Imaging, discussing not only its potential but also the challenges that still have to be overcome.
ABSTRACT
Microcalcification clusters (MCs) are among the most important biomarkers for breast cancer, especially in cases of nonpalpable lesions. The vast majority of deep learning studies on digital breast tomosynthesis (DBT) are focused on detecting and classifying lesions, especially soft-tissue lesions, in small regions of interest previously selected. Only about 25% of the studies are specific to MCs, and all of them are based on the classification of small preselected regions. Classifying the whole image according to the presence or absence of MCs is a difficult task due to the size of MCs and all the information present in an entire image. A completely automatic and direct classification, which receives the entire image, without prior identification of any regions, is crucial for the usefulness of these techniques in a real clinical and screening environment. The main purpose of this work is to implement and evaluate the performance of convolutional neural networks (CNNs) regarding an automatic classification of a complete DBT image for the presence or absence of MCs (without any prior identification of regions). In this work, four popular deep CNNs are trained and compared with a new architecture proposed by us. The main task of these trainings was the classification of DBT cases by absence or presence of MCs. A public database of realistic simulated data was used, and the whole DBT image was taken into account as input. DBT data were considered without and with preprocessing (to study the impact of noise reduction and contrast enhancement methods on the evaluation of MCs with CNNs). The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance. Very promising results were achieved with a maximum AUC of 94.19% for the GoogLeNet. The second-best AUC value was obtained with a new implemented network, CNN-a, with 91.17%. This CNN had the particularity of also being the fastest, thus becoming a very interesting model to be considered in other studies. With this work, encouraging outcomes were achieved in this regard, obtaining similar results to other studies for the detection of larger lesions such as masses. Moreover, given the difficulty of visualizing the MCs, which are often spread over several slices, this work may have an important impact on the clinical analysis of DBT images.
ABSTRACT
Breast cancer affects thousands of women across the world, every year. Methods to predict risk of breast cancer, or to stratify women in different risk levels, could help to achieve an early diagnosis, and consequently a reduction of mortality. This paper aims to review articles that extracted texture features from mammograms and used those features along with machine learning algorithms to assess breast cancer risk. Besides that, deep learning methodologies that aimed for the same goal were also reviewed. In this work, first, a brief introduction to breast cancer statistics and screening programs is presented; after that, research done in the field of breast cancer risk assessment are analyzed, in terms of both methodologies used and results obtained. Finally, considerations about the analyzed papers are conducted. The results of this review allow to conclude that both machine and deep learning methodologies provide promising results in the field of risk analysis, either in a stratification in risk groups, or in a prediction of a risk score. Although promising, future endeavors in this field should consider the possibility of the implementation of the methodology in clinical practice.
ABSTRACT
Digital Breast Tomosynthesis (DBT) presents out-of-plane artifacts caused by features of high intensity. Given observed data and knowledge about the point spread function (PSF), deconvolution techniques recover data from a blurred version. However, a correct PSF is difficult to achieve and these methods amplify noise. When no information is available about the PSF, blind deconvolution can be used. Additionally, Total Variation (TV) minimization algorithms have achieved great success due to its virtue of preserving edges while reducing image noise. This work presents a novel approach in DBT through the study of out-of-plane artifacts using blind deconvolution and noise regularization based on TV minimization. Gradient information was also included. The methodology was tested using real phantom data and one clinical data set. The results were investigated using conventional 2D slice-by-slice visualization and 3D volume rendering. For the 2D analysis, the artifact spread function (ASF) and Full Width at Half Maximum (FWHMMASF) of the ASF were considered. The 3D quantitative analysis was based on the FWHM of disks profiles at 90°, noise and signal to noise ratio (SNR) at 0° and 90°. A marked visual decrease of the artifact with reductions of FWHMASF (2D) and FWHM90° (volume rendering) of 23.8% and 23.6%, respectively, was observed. Although there was an expected increase in noise level, SNR values were preserved after deconvolution. Regardless of the methodology and visualization approach, the objective of reducing the out-of-plane artifact was accomplished. Both for the phantom and clinical case, the artifact reduction in the z was markedly visible.
Subject(s)
Algorithms , Breast Neoplasms , Image Processing, Computer-Assisted , Mammography , Artifacts , Breast Neoplasms/diagnostic imaging , Phantoms, Imaging , Signal-To-Noise RatioABSTRACT
Cancer is a major health concern and the prognosis is often poor. Significant advances in nanotechnology are now driving a revolution in cancer detection and treatment. The goal of this study was to develop a novel hybrid nanosystem for melanoma treatment, integrating therapeutic and magnetic targeting modalities. Hence, we designed long circulating and pH-sensitive liposomes loading both dichloro(1,10-phenanthroline) copper (II) (Cuphen), a cytotoxic metallodrug, and iron oxide nanoparticles (IONPs). The synthetized IONPs were characterized by transmission electron microscopy and dynamic light scattering. Lipid-based nanoformulations were prepared by the dehydration rehydration method, followed by an extrusion step for reducing and homogenizing the mean size. Liposomes were characterized in terms of incorporation parameters and mean size. High Cuphen loadings were obtained and the presence of IONPs slightly reduced Cuphen incorporation parameters. Cuphen antiproliferative properties were preserved after association to liposomes and IONPs (at 2 mg/mL) did not interfere on cellular proliferation of murine and human melanoma cell lines. Moreover, the developed nanoformulations displayed magnetic properties. The absence of hemolytic activity for formulations under study demonstrated their safety for parenteral administration. In conclusion, a lipid-based nanosystem loading the cytotoxic metallodrug, Cuphen, and displaying magnetic properties was successfully designed.
ABSTRACT
3D volume rendering may represent a complementary option in the visualization of Digital Breast Tomosynthesis (DBT) examinations by providing an understanding of the underlying data at once. Rendering parameters directly influence the quality of rendered images. The purpose of this work is to study the influence of two of these parameters (voxel dimension in z direction and sampling distance) on DBT rendered data. Both parameters were studied with a real phantom and one clinical DBT data set. The voxel size was changed from 0.085 × 0.085 × 1.0 mm3 to 0.085 × 0.085 × 0.085 mm3 using ten interpolation functions available in the Visualization Toolkit library (VTK) and several sampling distance values were evaluated. The results were investigated at 90º using volume rendering visualization with composite technique. For phantom quantitative analysis, degree of smoothness, contrast-to-noise ratio, and full width at half maximum of a Gaussian curve fitted to the profile of one disk were used. Additionally, the time required for each visualization was also recorded. Hamming interpolation function presented the best compromise in image quality. The sampling distance values that showed a better balance between time and image quality were 0.025 mm and 0.05 mm. With the appropriate rendering parameters, a significant improvement in rendered images was achieved.
ABSTRACT
PURPOSE: Compressed sensing (CS) is a new approach in medical imaging which allows a sparse image to be reconstructed from undersampled data. Total variation (TV) based minimization algorithms are the one CS technique that has achieved great success due to its virtue of preserving edges while reducing image noise. The purpose of this work is to implement and evaluate the performance of a TV minimization filter able to increase the signal difference to noise ratio (SDNR) of digital breast tomosynthesis (DBT) images. METHODS: Assuming a Poisson noise model, the authors present a practical methodology, based on Rudin, Osher, and Fatemi model, which directly applies a TV minimization filter to real phantom and clinical DBT images. Different moments of filter application (before and after image reconstruction) and the suitable Lagrange multiplier (λ) to be used in filter equation are studied. Also, the relationship between background standard deviation (σB) of unfiltered images and optimal λ values is determined, in order to maximize the SDNR. Qualitative and quantitative analyses are conducted between unfiltered and filtered images and between the different moments of filter application. The proposed methodology is also tested with one clinical DBT data set. RESULTS: Using phantom data, when the filter is applied to the projections, the authors observed a decrease of 31.34% in TV and an increase of 5.29% and 5.44% in SDNR and full width at half maximum (FWHM), respectively. When applied after reconstruction, a decrease of 35.48% and 2.59% was achieved for TV and FWHM, respectively, and an increase of 8.32% for SDNR. For each moment of filter application, the optimal λ value found through a comprehensive study was λ = 85 and λ = 60 when the filter is applied before and after reconstruction, respectively. The best fit found for the relationship between σB and the corresponding λ values that allowed the highest filtered SDNR was the logarithmic adjustment. The difference between the λ values obtained by the first approach and the logarithmic adjustment ranges from 0.11% (filter applied before reconstruction) to 2.54% (filter applied after reconstruction). On the other hand, a decrease of 37.63% and 2.42% in TV and FWHM, respectively, and an increase of 24.39% in SDNR were obtained when the filter is applied to clinical data. This great minimization is present through a visual inspection of unfiltered and filtered clinical images, where areas with higher noise level become smoother while preserving edges and details of the structures. CONCLUSIONS: An optimized digital filter for TV minimization in DBT imaging has been presented. The reliability of a logarithmic relation found between σB and λ values was confirmed and can be used in future work. Both quantitative and qualitative analyses performed in a clinical DBT image confirmed the relevance of this approach in improving image quality in DBT imaging. The results obtained are very encouraging about increasing SDNR in a short time and preserving the principal variations in image, the structures' boundary.
Subject(s)
Image Processing, Computer-Assisted/methods , Mammography/methods , Radiographic Image Enhancement/methods , Algorithms , Humans , Phantoms, Imaging , Signal-To-Noise RatioABSTRACT
PURPOSE: The optimization of the collimator design is essential to obtain the best possible sensitivity in single photon emission computed tomography imaging. The aim of this work is to present a methodology for maximizing the sensitivity of convergent collimators, specifically designed to match the pitch of pixelated detectors, for a fixed spatial resolution value and to present some initial results using this approach. METHODS: Given the matched constraint, the optimal collimator design cannot be simply found by allowing the highest level of septal penetration and spatial resolution consistent with the imposed restrictions, as it is done for the optimization of conventional collimators. Therefore, an algorithm that interactively calculates the collimator dimensions, with the maximum sensitivity, which respect the imposed restrictions was developed and used to optimize cone and fan beam collimators with tapered square-shaped holes for low (60-300 keV) and high energy radiation (300-511 keV). The optimal collimator dimensions were locally calculated based on the premise that each hole and septa of the convergent collimator should locally resemble an appropriate optimal matched parallel collimator. RESULTS: The optimal collimator dimensions, calculated for subcentimeter resolutions (3 and 7.5 mm), common pixel sizes (1.6, 2.1, and 2.5 mm), and acceptable septal penetration at 140 keV, were approximately constant throughout the collimator, despite their different hole incidence angles. By using these input parameters and a less strict septal penetration value of 5%, the optimal collimator dimensions and the corresponding mass per detector area were calculated for 511 keV. It is shown that a low value of focal distance leads to improvements in the average sensitivity at a fixed source-collimator distance and resolution. The optimal cone beam performance outperformed that of other optimal collimation geometries (fan and parallel beam) in imaging objects close to the collimator surface. CONCLUSIONS: These results demonstrate the potential of this kind of optimal convergent collimators for the use in small field of view imaging applications.
Subject(s)
Computer-Aided Design , Image Enhancement/instrumentation , Imaging, Three-Dimensional/instrumentation , Signal Processing, Computer-Assisted/instrumentation , Tomography, Emission-Computed, Single-Photon/instrumentation , Equipment Design , Equipment Failure Analysis , Reproducibility of Results , Sensitivity and SpecificityABSTRACT
In a wide range of medical fields, technological advancements have led to an increase in the average collective dose in national populations worldwide. Periodic estimations of the average collective population dose due to medical exposure is, therefore of utmost importance, and is now mandatory in countries within the European Union (article 12 of EURATOM directive 97/43). Presented in this work is a report on the estimation of the collective dose in the Portuguese population due to nuclear medicine diagnostic procedures and the Top 20 diagnostic radiology examinations, which represent the 20 exams that contribute the most to the total collective dose in diagnostic radiology and interventional procedures in Europe. This work involved the collaboration of a multidisciplinary taskforce comprising representatives of all major Portuguese stakeholders (universities, research institutions, public and private healthcare providers, administrative services of the National Healthcare System, scientific and professional associations and private service providers). This allowed us to gather a comprehensive amount of data necessary for a robust estimation of the collective effective dose to the Portuguese population. The methodology used for data collection and dose estimation was based on European Commission recommendations, as this work was performed in the framework of the European wide Dose Datamed II project. This is the first study estimating the collective dose for the population in Portugal, considering such a wide national coverage and range of procedures and consisting of important baseline reference data. The taskforce intends to continue developing periodic collective dose estimations in the future. The estimated annual average effective dose for the Portuguese population was of 0.080±0.017 mSv caput(-1) for nuclear medicine exams and of 0.96±0.68 mSv caput(-1) for the Top 20 diagnostic radiology exams.