Exploring Negated Entites for Named Entity Recognition in Italian Lung Cancer Clinical Reports.

This paper explores the potential of leveraging electronic health records (EHRs) for personalized health research through the application of artificial intelligence (AI) techniques, specifically Named Entity Recognition (NER). By extracting crucial patient information from clinical texts, including diagnoses, medications, symptoms, and lab tests, AI facilitates the rapid identification of relevant data, paving the way for future care paradigms. The study focuses on Non-small cell lung cancer (NSCLC) in Italian clinical notes, introducing a novel set of 29 clinical entities that include both presence or absence (negation) of relevant information associated with NSCLC. Using a state-of-the-art model pretrained on Italian biomedical texts, we achieve promising results (average F1-score of 80.8%), demonstrating the feasibility of employing AI for extracting biomedical information in the Italian language.

Multi input-Multi output 3D CNN for dementia severity assessment with incomplete multimodal data.

Alzheimer's Disease is the most common cause of dementia, whose progression spans in different stages, from very mild cognitive impairment to mild and severe conditions. In clinical trials, Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) are mostly used for the early diagnosis of neurodegenerative disorders since they provide volumetric and metabolic function information of the brain, respectively. In recent years, Deep Learning (DL) has been employed in medical imaging with promising results. Moreover, the use of the deep neural networks, especially Convolutional Neural Networks (CNNs), has also enabled the development of DL-based solutions in domains characterized by the need of leveraging information coming from multiple data sources, raising the Multimodal Deep Learning (MDL). In this paper, we conduct a systematic analysis of MDL approaches for dementia severity assessment exploiting MRI and PET scans. We propose a Multi Input-Multi Output 3D CNN whose training iterations change according to the characteristic of the input as it is able to handle incomplete acquisitions, in which one image modality is missed. Experiments performed on OASIS-3 dataset show the satisfactory results of the implemented network, which outperforms approaches exploiting both single image modality and different MDL fusion techniques.

AI-based methodologies for exoskeleton-assisted rehabilitation of the lower limb: a review.

Over the past few years, there has been a noticeable surge in efforts to design novel tools and approaches that incorporate Artificial Intelligence (AI) into rehabilitation of persons with lower-limb impairments, using robotic exoskeletons. The potential benefits include the ability to implement personalized rehabilitation therapies by leveraging AI for robot control and data analysis, facilitating personalized feedback and guidance. Despite this, there is a current lack of literature review specifically focusing on AI applications in lower-limb rehabilitative robotics. To address this gap, our work aims at performing a review of 37 peer-reviewed papers. This review categorizes selected papers based on robotic application scenarios or AI methodologies. Additionally, it uniquely contributes by providing a detailed summary of input features, AI model performance, enrolled populations, exoskeletal systems used in the validation process, and specific tasks for each paper. The innovative aspect lies in offering a clear understanding of the suitability of different algorithms for specific tasks, intending to guide future developments and support informed decision-making in the realm of lower-limb exoskeleton and AI applications.

Deep pathomics: A new image-based tool for predicting response to treatment in stage III non-small cell lung cancer.

Despite the advantages offered by personalized treatments, there is presently no way to predict response to chemoradiotherapy in patients with non-small cell lung cancer (NSCLC). In this exploratory study, we investigated the application of deep learning techniques to histological tissue slides (deep pathomics), with the aim of predicting the response to therapy in stage III NSCLC. We evaluated 35 digitalized tissue slides (biopsies or surgical specimens) obtained from patients with stage IIIA or IIIB NSCLC. Patients were classified as responders (12/35, 34.7%) or non-responders (23/35, 65.7%) based on the target volume reduction shown on weekly CT scans performed during chemoradiation treatment. Digital tissue slides were tested by five pre-trained convolutional neural networks (CNNs)-AlexNet, VGG, MobileNet, GoogLeNet, and ResNet-using a leave-two patient-out cross validation approach, and we evaluated the networks' performances. GoogLeNet was globally found to be the best CNN, correctly classifying 8/12 responders and 10/11 non-responders. Moreover, Deep-Pathomics was found to be highly specific (TNr: 90.1) and quite sensitive (TPr: 0.75). Our data showed that AI could surpass the capabilities of all presently available diagnostic systems, supplying additional information beyond that currently obtainable in clinical practice. The ability to predict a patient's response to treatment could guide the development of new and more effective therapeutic AI-based approaches and could therefore be considered an effective and innovative step forward in personalised medicine.

Radiomic Machine-Learning Analysis of Multiparametric Magnetic Resonance Imaging in the Diagnosis of Clinically Significant Prostate Cancer: New Combination of Textural and Clinical Features.

BACKGROUND: The aim of our study was to develop a radiomic tool for the prediction of clinically significant prostate cancer. METHODS: From September 2020 to December 2021, 91 patients who underwent magnetic resonance imaging prostate fusion biopsy at our institution were selected. Prostate cancer aggressiveness was assessed by combining the three orthogonal planes-Llocal binary pattern the 3Dgray level co-occurrence matrix, and other first order statistical features with clinical (semantic) features. The 487 features were used to predict whether the Gleason score was clinically significant (≥7) in the final pathology. A feature selection algorithm was used to determine the most predictive features, and at the end of the process, nine features were chosen through a 10-fold cross validation. RESULTS: The feature analysis revealed a detection accuracy of 83.5%, with a clinically significant precision of 84.4% and a clinically significant sensitivity of 91.5%. The resulting area under the curve was 80.4%. CONCLUSIONS: Radiomic analysis allowed us to develop a tool that was able to predict a Gleason score of ≥7. This new tool may improve the detection rate of clinically significant prostate cancer and overcome the limitations of the subjective interpretation of magnetic resonance imaging, reducing the number of useless biopsies.

Multi-objective optimization determines when, which and how to fuse deep networks: An application to predict COVID-19 outcomes.

The COVID-19 pandemic has caused millions of cases and deaths and the AI-related scientific community, after being involved with detecting COVID-19 signs in medical images, has been now directing the efforts towards the development of methods that can predict the progression of the disease. This task is multimodal by its very nature and, recently, baseline results achieved on the publicly available AIforCOVID dataset have shown that chest X-ray scans and clinical information are useful to identify patients at risk of severe outcomes. While deep learning has shown superior performance in several medical fields, in most of the cases it considers unimodal data only. In this respect, when, which and how to fuse the different modalities is an open challenge in multimodal deep learning. To cope with these three questions here we present a novel approach optimizing the setup of a multimodal end-to-end model. It exploits Pareto multi-objective optimization working with a performance metric and the diversity score of multiple candidate unimodal neural networks to be fused. We test our method on the AIforCOVID dataset, attaining state-of-the-art results, not only outperforming the baseline performance but also being robust to external validation. Moreover, exploiting XAI algorithms we figure out a hierarchy among the modalities and we extract the features' intra-modality importance, enriching the trust on the predictions made by the model.

MRI- and Histologic-Molecular-Based Radio-Genomics Nomogram for Preoperative Assessment of Risk Classes in Endometrial Cancer.

High- and low-risk endometrial carcinoma (EC) differ in whether or not a lymphadenectomy is performed. We aimed to develop MRI-based radio-genomic models able to preoperatively assess lymph-vascular space invasion (LVSI) and discriminate between low- and high-risk EC according to the ESMO-ESGO-ESTRO 2020 guidelines, which include molecular risk classification proposed by "ProMisE". This is a retrospective, multicentric study that included 64 women with EC who underwent 3T-MRI before a hysterectomy. Radiomics features were extracted from T2WI images and apparent diffusion coefficient maps (ADC) after manual segmentation of the gross tumor volume. We constructed a multiple logistic regression approach from the most relevant radiomic features to distinguish between low- and high-risk classes under the ESMO-ESGO-ESTRO 2020 guidelines. A similar approach was taken to assess LVSI. Model diagnostic performance was assessed via ROC curves, accuracy, sensitivity and specificity on training and test sets. The LVSI predictive model used a single feature from ADC as a predictor; the risk class model used two features as predictors from both ADC and T2WI. The low-risk predictive model showed an AUC of 0.74 with an accuracy, sensitivity, and specificity of 0.74, 0.76, 0.94; the LVSI model showed an AUC of 0.59 with an accuracy, sensitivity, and specificity of 0.60, 0.50, 0.61. MRI-based radio-genomic models are useful for preoperative EC risk stratification and may facilitate therapeutic management.

Comparison of current methods for anti-dsDNA antibody detection and reshaping diagnostic strategies.

Anti-double-stranded DNA antibodies (anti-dsDNA) are considered a specific marker for systemic lupus erythematosus (SLE). Though the Farr technique was once the reference method for their detection, it has been almost entirely replaced by more recently developed assays. However, there is still no solid evidence of the commutability of these methods in terms of diagnostic accuracy and their correlation with the Crithidia luciliae immunofluorescence test (CLIFT). Anti-dsDNA antibody levels were measured in 80 subjects: 24 patients with SLE, 36 disease controls drawn from different autoimmune rheumatic diseases (14 systemic sclerosis, 10 Sjögren's syndrome, nine autoimmune myositis, three mixed connective tissue disease), 10 inflammatory arthritis and 10 apparently healthy blood donors by eight different methods: fluorescence enzyme immunoassay, microdot array, chemiluminescent immunoassay (two assays), multiplex flow immunoassay, particle multi-analyte technology immunoassay and two CLIFT. At the recommended manufacturer cut-off, the sensitivity varied from 67% to 92%, while the specificity ranged from 84% to 98%. Positive agreement among CLIFT and the other assays was higher than negative agreement. Mean agreement among methods assessed by the Cohen's kappa was 0.715, ranging from moderate (0.588) to almost perfect (0.888). Evaluation of the concordance among quantitative values by regression analysis showed a poor correlation index (mean r2, 0.66). The present study shows that current technologies for anti-dsDNA antibody detection are not fully comparable. In particular, their different correlation with CLIFT influences their positioning in the diagnostic algorithm for SLE (either in association or sequentially). Considering the high intermethod variability, harmonization and commutability of anti-dsDNA antibody testing remains an unachieved goal.

A Multimodal Ensemble Driven by Multiobjective Optimisation to Predict Overall Survival in Non-Small-Cell Lung Cancer.

Lung cancer accounts for more deaths worldwide than any other cancer disease. In order to provide patients with the most effective treatment for these aggressive tumours, multimodal learning is emerging as a new and promising field of research that aims to extract complementary information from the data of different modalities for prognostic and predictive purposes. This knowledge could be used to optimise current treatments and maximise their effectiveness. To predict overall survival, in this work, we investigate the use of multimodal learning on the CLARO dataset, which includes CT images and clinical data collected from a cohort of non-small-cell lung cancer patients. Our method allows the identification of the optimal set of classifiers to be included in the ensemble in a late fusion approach. Specifically, after training unimodal models on each modality, it selects the best ensemble by solving a multiobjective optimisation problem that maximises both the recognition performance and the diversity of the predictions. In the ensemble, the labels of each sample are assigned using the majority voting rule. As further validation, we show that the proposed ensemble outperforms the models learning a single modality, obtaining state-of-the-art results on the task at hand.

CNN-Based Approaches with Different Tumor Bounding Options for Lymph Node Status Prediction in Breast DCE-MRI.

BACKGROUND: The axillary lymph node status (ALNS) is one of the most important prognostic factors in breast cancer (BC) patients, and it is currently evaluated by invasive procedures. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), highlights the physiological and morphological characteristics of primary tumor tissue. Deep learning approaches (DL), such as convolutional neural networks (CNNs), are able to autonomously learn the set of features directly from images for a specific task. MATERIALS AND METHODS: A total of 155 malignant BC lesions evaluated via DCE-MRI were included in the study. For each patient's clinical data, the tumor histological and MRI characteristics and axillary lymph node status (ALNS) were assessed. LNS was considered to be the final label and dichotomized (LN+ (27 patients) vs. LN- (128 patients)). Based on the concept that peritumoral tissue contains valuable information about tumor aggressiveness, in this work, we analyze the contributions of six different tumor bounding options to predict the LNS using a CNN. These bounding boxes include a single fixed-size box (SFB), a single variable-size box (SVB), a single isotropic-size box (SIB), a single lesion variable-size box (SLVB), a single lesion isotropic-size box (SLIB), and a two-dimensional slice (2DS) option. According to the characteristics of the volumes considered as inputs, three different CNNs were investigated: the SFB-NET (for the SFB), the VB-NET (for the SVB, SIB, SLVB, and SLIB), and the 2DS-NET (for the 2DS). All the experiments were run in 10-fold cross-validation. The performance of each CNN was evaluated in terms of accuracy, sensitivity, specificity, the area under the ROC curve (AUC), and Cohen's kappa coefficient (K). RESULTS: The best accuracy and AUC are obtained by the 2DS-NET (78.63% and 77.86%, respectively). The 2DS-NET also showed the highest specificity, whilst the highest sensibility was attained by the VB-NET based on the SVB and SIB as bounding options. CONCLUSION: We have demonstrated that a selective inclusion of the DCE-MRI's peritumoral tissue increases accuracy in the lymph node status prediction in BC patients using CNNs as a DL approach.

Quantib Prostate Compared to an Expert Radiologist for the Diagnosis of Prostate Cancer on mpMRI: A Single-Center Preliminary Study.

Background: To evaluate the clinical utility of an Artificial Intelligence (AI) radiology solution, Quantib Prostate, for prostate cancer (PCa) lesions detection on multiparametric Magnetic Resonance Images (mpMRI). Methods: Prostate mpMRI exams of 108 patients were retrospectively studied. The diagnostic performance of an expert radiologist (>8 years of experience) and of an inexperienced radiologist aided by Quantib software were compared. Three groups of patients were assessed: patients with positive mpMRI, positive target biopsy, and/or at least one positive random biopsy (group A, 73 patients); patients with positive mpMRI and a negative biopsy (group B, 14 patients), and patients with negative mpMRI who did not undergo biopsy (group-C, 21 patients). Results: In group A, the AI-assisted radiologist found new lesions with positive biopsy correlation, increasing the diagnostic PCa performance when compared with the expert radiologist, reaching an SE of 92.3% and a PPV of 90.1% (vs. 71.7% and 84.4%). In group A, the expert radiologist found 96 lesions on 73 mpMRI exams (17.7% PIRADS3, 56.3% PIRADS4, and 26% PIRADS5). The AI-assisted radiologist found 121 lesions (0.8% PIRADS3, 53.7% PIRADS4, and 45.5% PIRADS5). At biopsy, 33.9% of the lesions were ISUP1, 31.4% were ISUP2, 22% were ISUP3, 10.2% were ISUP4, and 2.5% were ISUP5. In group B, where biopsies were negative, the AI-assisted radiologist excluded three lesions but confirmed all the others. In group-C, the AI-assisted radiologist found 37 new lesions, most of them PIRADS 3, with 32.4% localized in the peripherical zone and 67.6% in the transition zone. Conclusions: Quantib software is a very sensitive tool to use specifically in high-risk patients (high PIRADS and high Gleason score).

Artificial Intelligence in Bone Metastases: An MRI and CT Imaging Review.

(1) Background: The purpose of this review is to study the role of radiomics as a supporting tool in predicting bone disease status, differentiating benign from malignant bone lesions, and characterizing malignant bone lesions. (2) Methods: Two reviewers conducted the literature search independently. Thirteen articles on radiomics as a decision support tool for bone lesions were selected. The quality of the methodology was evaluated according to the radiomics quality score (RQS). (3) Results: All studies were published between 2018 and 2021 and were retrospective in design. Eleven (85%) studies were MRI-based, and two (15%) were CT-based. The sample size was <200 patients for all studies. There is significant heterogeneity in the literature, as evidenced by the relatively low RQS value (average score = 22.6%). There is not a homogeneous protocol used for MRI sequences among the different studies, although the highest predictive ability was always obtained in T2W-FS. Six articles (46%) reported on the potential application of the model in a clinical setting with a decision curve analysis (DCA). (4) Conclusions: Despite the variability in the radiomics method application, the similarity of results and conclusions observed is encouraging. Substantial limits were found; prospective and multicentric studies are needed to affirm the role of radiomics as a supporting tool.

Pareto optimization of deep networks for COVID-19 diagnosis from chest X-rays.

The year 2020 was characterized by the COVID-19 pandemic that has caused, by the end of March 2021, more than 2.5 million deaths worldwide. Since the beginning, besides the laboratory test, used as the gold standard, many applications have been applying deep learning algorithms to chest X-ray images to recognize COVID-19 infected patients. In this context, we found out that convolutional neural networks perform well on a single dataset but struggle to generalize to other data sources. To overcome this limitation, we propose a late fusion approach where we combine the outputs of several state-of-the-art CNNs, introducing a novel method that allows us to construct an optimum ensemble determining which and how many base learners should be aggregated. This choice is driven by a two-objective function that maximizes, on a validation set, the accuracy and the diversity of the ensemble itself. A wide set of experiments on several publicly available datasets, accounting for more than 92,000 images, shows that the proposed approach provides average recognition rates up to 93.54% when tested on external datasets.

3T-MRI Artificial Intelligence in Patients with Invasive Breast Cancer to Predict Distant Metastasis Status: A Pilot Study.

BACKGROUND: The incidence of breast cancer metastasis has decreased over the years. However, 20-30% of patients with early breast cancer still die from metastases. The purpose of this study is to evaluate the performance of a Deep Learning Convolutional Neural Networks (CNN) model to predict the risk of distant metastasis using 3T-MRI DCE sequences (Dynamic Contrast-Enhanced). METHODS: A total of 157 breast cancer patients who underwent staging 3T-MRI examinations from January 2011 to July 2022 were retrospectively examined. Patient data, tumor histological and MRI characteristics, and clinical and imaging follow-up examinations of up to 7 years were collected. Of the 157 MRI examinations, 39/157 patients (40 lesions) had distant metastases, while 118/157 patients (120 lesions) were negative for distant metastases (control group). We analyzed the role of the Deep Learning technique using a single variable size bounding box (SVB) option and employed a Voxel Based (VB) NET CNN model. The CNN performance was evaluated in terms of accuracy, sensitivity, specificity, and area under the ROC curve (AUC). RESULTS: The VB-NET model obtained a sensitivity, specificity, accuracy, and AUC of 52.50%, 80.51%, 73.42%, and 68.56%, respectively. A significant correlation was found between the risk of distant metastasis and tumor size, and the expression of PgR and HER2. CONCLUSIONS: We demonstrated a currently insufficient ability of the Deep Learning approach in predicting a distant metastasis status in patients with BC using CNNs.

On the Additional Information Provided by 3T-MRI ADC in Predicting Tumor Cellularity and Microscopic Behavior.

BACKGROUND: to evaluate whether Apparent Diffusion Coefficient (ADC) values of invasive breast cancer, provided by 3T Diffusion Weighted-Images (DWI), may represent a non-invasive predictor of pathophysiologic tumor aggressiveness. METHODS: 100 Patients with histologically proven invasive breast cancers who underwent a 3T-MRI examination were included in the study. All MRI examinations included dynamic contrast-enhanced and DWI/ADC sequences. ADC value were calculated for each lesion. Tumor grade was determined according to the Nottingham Grading System, and immuno-histochemical analysis was performed to assess molecular receptors, cellularity rate, on both biopsy and surgical specimens, and proliferation rate (Ki-67 index). Spearman's Rho test was used to correlate ADC values with histological (grading, Ki-67 index and cellularity) and MRI features. ADC values were compared among the different grading (G1, G2, G3), Ki-67 (<20% and >20%) and cellularity groups (<50%, 50-70% and >70%), using Mann-Whitney and Kruskal-Wallis tests. ROC curves were performed to demonstrate the accuracy of the ADC values in predicting the grading, Ki-67 index and cellularity groups. RESULTS: ADC values correlated significantly with grading, ER receptor status, Ki-67 index and cellularity rates. ADC values were significantly higher for G1 compared with G2 and for G1 compared with G3 and for Ki-67 < 20% than Ki-67 > 20%. The Kruskal-Wallis test showed that ADC values were significantly different among the three grading groups, the three biopsy cellularity groups and the three surgical cellularity groups. The best ROC curves were obtained for the G3 group (AUC of 0.720), for G2 + G3 (AUC of 0.835), for Ki-67 > 20% (AUC of 0.679) and for surgical cellularity rate > 70% (AUC of 0.805). CONCLUSIONS: 3T-DWI ADC is a direct predictor of cellular aggressiveness and proliferation in invasive breast carcinoma, and can be used as a supporting non-invasive factor to characterize macroscopic lesion behavior especially before surgery.

AIforCOVID: Predicting the clinical outcomes in patients with COVID-19 applying AI to chest-X-rays. An Italian multicentre study.

Recent epidemiological data report that worldwide more than 53 million people have been infected by SARS-CoV-2, resulting in 1.3 million deaths. The disease has been spreading very rapidly and few months after the identification of the first infected, shortage of hospital resources quickly became a problem. In this work we investigate whether artificial intelligence working with chest X-ray (CXR) scans and clinical data can be used as a possible tool for the early identification of patients at risk of severe outcome, like intensive care or death. Indeed, further to induce lower radiation dose than computed tomography (CT), CXR is a simpler and faster radiological technique, being also more widespread. In this respect, we present three approaches that use features extracted from CXR images, either handcrafted or automatically learnt by convolutional neuronal networks, which are then integrated with the clinical data. As a further contribution, this work introduces a repository that collects data from 820 patients enrolled in six Italian hospitals in spring 2020 during the first COVID-19 emergency. The dataset includes CXR images, several clinical attributes and clinical outcomes. Exhaustive evaluation shows promising performance both in 10-fold and leave-one-centre-out cross-validation, suggesting that clinical data and images have the potential to provide useful information for the management of patients and hospital resources.

Deep Reinforcement Learning for Fractionated Radiotherapy in Non-Small Cell Lung Carcinoma.

Lung cancer is by far the leading cause of cancer death among both men and women. Radiation therapy is one of the main approaches to lung cancer treatment, and its planning is crucial for the therapy outcome. However, the current practice that uniformly delivers the dose does not take into account the patient-specific tumour features that may affect treatment success. Since radiation therapy is by its very nature a sequential procedure, Deep Reinforcement Learning (DRL) is a well-suited methodology to overcome this limitation. In this respect, in this work we present a DRL controller optimizing the daily dose fraction delivered to the patient on the basis of CT scans collected over time during the therapy, offering a personalized treatment not only for volume adaptation, as currently intended, but also for daily fractionation. Furthermore, this contribution introduces a virtual radiotherapy environment based on a set of ordinary differential equations modelling the tissue radiosensitivity by combining both the effect of the radiotherapy treatment and cell growth. Their parameters are estimated from CT scans routinely collected using the Particle Swarm Optimization algorithm. This permits the DRL to learn the optimal behaviour through an iterative trial and error process with the environment. We performed several experiments considering three rewards functions modelling treatment strategies with different tissue aggressiveness and two exploration strategies for the exploration-exploitation dilemma. The results show that our DRL approach can adapt to radiation therapy treatment, optimizing its behaviour according to the different reward functions and outperforming the current clinical practice.

The Impact of Tumor Edema on T2-Weighted 3T-MRI Invasive Breast Cancer Histological Characterization: A Pilot Radiomics Study.

BACKGROUND: to evaluate the contribution of edema associated with histological features to the prediction of breast cancer (BC) prognosis using T2-weighted MRI radiomics. METHODS: 160 patients who underwent staging 3T-MRI from January 2015 to January 2019, with 164 histologically proven invasive BC lesions, were retrospectively reviewed. Patient data (age, menopausal status, family history, hormone therapy), tumor MRI-features (location, margins, enhancement) and histological features (histological type, grading, ER, PgR, HER2, Ki-67 index) were collected. Of the 160 MRI exams, 120 were considered eligible, corresponding to 127 lesions. T2-MRI were used to identify edema, which was classified in four groups: peritumoral, pre-pectoral, subcutaneous, or diffuse. A semi-automatic segmentation of the edema was performed for each lesion, using 3D Slicer open-source software. Main radiomics features were extracted and selected using a wrapper selection method. A Random Forest type classifier was trained to measure the performance of predicting histological factors using semantic features (patient data and MRI features) alone and semantic features associated with edema radiomics features. RESULTS: edema was absent in 37 lesions and present in 127 (62 peritumoral, 26 pre-pectoral, 16 subcutaneous, 23 diffuse). The AUC-classifier obtained by associating edema radiomics with semantic features was always higher compared to the AUC-classifier obtained from semantic features alone, for all five histological classes prediction (0.645 vs. 0.520 for histological type, 0.789 vs. 0.590 for grading, 0.487 vs. 0.466 for ER, 0.659 vs. 0.546 for PgR, and 0.62 vs. 0.573 for Ki67). CONCLUSIONS: radiomic features extracted from tumor edema contribute significantly to predicting tumor histology, increasing the accuracy obtained from the combination of patient clinical characteristics and breast imaging data.

3T MRI-Radiomic Approach to Predict for Lymph Node Status in Breast Cancer Patients.

BACKGROUND: axillary lymph node (LN) status is one of the main breast cancer prognostic factors and it is currently defined by invasive procedures. The aim of this study is to predict LN metastasis combining MRI radiomics features with primary breast tumor histological features and patients' clinical data. METHODS: 99 lesions on pre-treatment contrasted 3T-MRI (DCE). All patients had a histologically proven invasive breast cancer and defined LN status. Patients' clinical data and tumor histological analysis were previously collected. For each tumor lesion, a semi-automatic segmentation was performed, using the second phase of DCE-MRI. Each segmentation was optimized using a convex-hull algorithm. In addition to the 14 semantics features and a feature ROI volume/convex-hull volume, 242 other quantitative features were extracted. A wrapper selection method selected the 15 most prognostic features (14 quantitative, 1 semantic), used to train the final learning model. The classifier used was the Random Forest. RESULTS: the AUC-classifier was 0.856 (label = positive or negative). The contribution of each feature group was lower performance than the full signature. CONCLUSIONS: the combination of patient clinical, histological and radiomics features of primary breast cancer can accurately predict LN status in a non-invasive way.