Conventional age-based breast cancer screening ignores substantial inter-individual risk variation, contributing to overdiagnosis, false positives, and missed opportunities for earlier detection in high-risk women. Mammography-based artificial intelligence (AI) may enable risk-stratified screening and more efficient workflows. To systematically review evidence on mammography-based AI for personalized breast cancer screening, covering risk prediction, detection/triage, decision support, and associated ethical, economic, and equity implications. We searched MEDLINE/PubMed, Embase, Scopus, Web of Science, and the Cochrane Library (January 2015–November 2025) for studies evaluating AI-enabled personalization in breast cancer screening. Two reviewers independently screened 612 records, assessed 77 full texts, and included 30 studies; data were synthesized narratively. Image-based deep-learning risk models consistently outperformed traditional clinical risk tools and enriched future cancers within small high-risk strata, including cancers presenting as interval cancers in recent validations. Prospective trials and real-world implementations indicate that AI-supported reading can maintain or modestly improve cancer detection while reducing radiologist workload by roughly 40–50

Background: This study introduces a new qualitative enhancement descriptor for contrast-enhanced mammography (CEM), termed Ground-Glass Enhancement (GGE). The objective was to categorize breast lesions using this descriptor and evaluate its association with malignancy and markers of tumor aggressiveness. Methods: In this single-center retrospective study, 249 patients with a single enhancing lesion on CEM were included. Lesions were classified into pure Ground-Glass Enhancement (PGGE), Heterogeneous Ground-Glass Enhancement (HGGE), or Opaque Enhancement (OE) based on the degree of obscuration of the underlying parenchyma. Clinical, imaging, and pathological features were compared across groups. Multivariable logistic regression was used to identify independent predictors of malignancy. Results: Significant differences across enhancement patterns were found in lesion conspicuity, enhancement type, size, background enhancement, and patient age. OE lesions more frequently showed high conspicuity (83% vs. 62% in HGGE and 20% in PGGE) and a mass-like appearance (94% vs. 73% in HGGE and 81% in PGGE). HGGE lesions had the largest median size (25 mm, vs. 17 mm in OE and 13 mm in PGGE), and OE lesions most often exhibited minimal background enhancement (77%, vs. 50% in HGGE). In multivariable analysis, mass-like enhancement (OR = 4.59), larger size (OR = 1.27 per +5 mm), and high conspicuity (OR = 3.43) were independently associated with malignancy. Although GGE categories correlated with malignancy in univariable analysis, this was not confirmed in the adjusted model. OE lesions were significantly associated with higher Ki-67 expression (73% with Ki-67 >20%), indicating increased proliferative activity compared with PGGE (43%) and HGGE (57%). Conclusions: The GGE descriptor captures clinically relevant imaging features and may support visual stratification of breast lesions on CEM. While not an independent predictor of malignancy, it appears more closely related to markers of tumor aggressiveness.

This study evaluates the capabilities of large language models, specifically GPT-4, in interpreting mammographic images. The analysis involved 120 mammographic images equally divided between cases with and without mammography's findings. Without additional context, the LLM was tasked to generate reports based solely on these images. GPT-4 correctly identified mammographic projections in 53.3% of cases and showed varying degrees of accuracy in identifying microcalcifications and masses. The study highlighted GPT-4's embryonic interpretative abilities with a sensitivity of 50.0% and specificity of 37.5%. However, a significant rate of false positives and false negatives, along with hallucinations, underscored the model's limitations. This exploratory test offers insights into the potential and risks of using LLMs in mammography interpretation, also underscoring the need for dedicated training, validation, and regulation of AI tools in healthcare to ensure their reliability and safety in clinical practice.

BACKGROUND:Therapeutic approach used for pancreatic ductal adenocarcinoma is usually translated also for the rarer acinar counterpart, which shows a different mutational landscape nevertheless. While dMMR/MSI-H status is rare in the ductal histotype, it appears to be more prevalent in pancreatic acinar cell carcinoma (PACC). CASE PRESENTATION:We report the case of a patient with locally advanced MSI-H PACC in whom the treatment with the anti-PD-1 pembrolizumab, administered as third line, made possible surgical resection, achieving even an exceptional pathological complete response. CONCLUSIONS:Treatment of PACC should be tailored based on the peculiar molecular features that distinguish PACC from ductal adenocarcinoma. Evaluation of potentially therapeutically targetable alterations should be mandatory in case of PACC diagnosis.

Objectives: This meta-analysis compares the efficacy, limitations, and clinical implications of abbreviated breast MRI (AB-MRI) and full protocol MRI (FP-MRI), focusing on diagnostic accuracy across diverse populations. It extends previous analyses by including studies conducted after 2019 in both screening and diagnostic contexts. Methods: We conducted a systematic review (November 2019 to December 2022), using a bivariate model to calculate summary estimates of sensitivity and specificity. Random effect models were applied for summary area under the curve (AUC), and probability distributions for negative and positive predictive values were obtained. Subgroup analyses explored differences in sensitivity, specificity, and AUC between AB-MRI and FP-MRI. Results: From 11 eligible studies (1 prospective, 10 retrospective), statistical analysis revealed a significant difference in sensitivity between FP-MRI (95%) and AB-MRI (86%, P = .005), with no significant difference in specificity (P = .50). AB-MRI's shorter acquisition time suggests potential for higher patient throughput, but challenges remain in detecting small lesions and nonmass enhancements. Some studies recommend additional sequences, like diffusion-weighted imaging, to improve diagnostic performance. Conclusions: While FP-MRI remains the gold standard in breast cancer detection, AB-MRI offers a quicker alternative, especially in high-risk screening. However, its lower sensitivity limits its use as a standalone diagnostic tool. Future research should optimize AB-MRI protocols and consider patient-specific factors to enhance breast cancer screening and diagnostic strategies. Advances in knowledge: This meta-analysis expands understanding of AB-MRI's role in breast cancer detection, highlighting its benefits and limitations compared to FP-MRI, particularly in terms of sensitivity and screening efficiency.

4056 Background: REGO is an oral multi-targeted TKI that showed promising activity in advanced HER2-neg gastric cancer pts. The a-MANTRA study aimed to evaluate the efficacy and safety of REGO as maintenance after first-line (1L) therapy in advanced GC/GEJ tumors. Methods: This is a randomized, double-blind, placebo-controlled, multicenter Phase-II study in which HER2 neg advanced GC/GEJ pts with disease control after 1L platinum and fluoropyrimidines-based therapy, were randomized (1:1 ratio) to receive maintenance placebo (ARM A) or REGO (ARM B) starting at 80 mg with an escalation up to 160 mg (once daily on d1-21 q28 days), until intolerance or PD. The primary endpoint was Progression Free Survival 1 (PFS1). Two-sided 80% CIs were calculated to detect HR of 0.57, using a one-sided α=0.10 to have 90% power, translating to 3 months of improvement in mPFS. 118 subjects were required to observe 88 events. The interim analysis (IA) was planned after 44 events. Results: 67 pts were randomized in 18 Italian Cancer Centers; 64 pts (33 in ARM A and 31 in ARM B) received treatment. Most of the pts were male (64.1%), Caucasian (96.8%), ECOG PS 0 (81.2%), and median age 66 (40-80) yrs; the main primary tumor side was proximal (64.1%) with intestinal subtype (54.7%). Peritoneum metastases were present in 42.2%. IA showed a 98% probability of achieving a positive and statistically significant result for mPFS1. The study was stopped early due to the introduction of ICIs as 1L in PD-L1-positive pts. The objective responses to 1L chemotherapy were RP (50.0%), SD (42.2%), and CR (7.8%). At a median follow-up of 31 months (IQR 19.1-33-8), 28 (84.8%) and 26 (83.8%) events for each arm were reported. The main reason for discontinuation was PD (66.7% and 35.5%). The mPFS was 3.91 (80% CI, 2.27-5.98) and 5.19 months (80% CI, 4.0-7.26) with HR= 0.736 (80%CI, 0.51-1.04; p=0.1318). The mOS was 11.25 and 16.97 months (80%CI, HR=0.596 [0.318-1.103], p=0.1003). The most common G3-4 toxicities in ARM A and B were fatigue (3.0 vs 6.4%), thrombocytopenia (3.0 vs 3.2%), hand-foot syndrome (0 vs 12.9%), diarrhea and skin rash (0 vs 3,2%, respectively). Conclusions: Despite the promising trend, considering the sample was not sized for the statistical study hypothesis, REGO maintenance therapy after 1L chemotherapy did not reach statistically significant results in mPFS1. No significant safety concerns were raised. New study designs following the REGO with ICIs are currently underway. Clinical trial information: NCT03627728 .

The aim of this study was to evaluate the diagnostic performance of contrast-enhanced spectral mammography (CESM) in predicting breast lesion malignancy due to microcalcifications compared to lesions that present with other radiological findings. Three hundred and twenty-one patients with 377 breast lesions that underwent CESM and histological assessment were included. All the lesions were scored using a 4-point qualitative scale according to the degree of contrast enhancement at the CESM examination. The histological results were considered the gold standard. In the first analysis, enhancement degree scores of 2 and 3 were considered predictive of malignity. The sensitivity (SE) and positive predictive value (PPV) were significative lower for patients with lesions with microcalcifications without other radiological findings (SE = 53.3% vs. 82.2%, p-value < 0.001 and PPV = 84.2% vs. 95.2%, p-value = 0.049, respectively). On the contrary, the specificity (SP) and negative predictive value (NPV) were significative higher among lesions with microcalcifications without other radiological findings (SP = 95.8% vs. 84.2%, p-value = 0.026 and NPV = 82.9% vs. 55.2%, p-value < 0.001, respectively). In a second analysis, degree scores of 1, 2, and 3 were considered predictive of malignity. The SE (80.0% vs. 96.8%, p-value < 0.001) and PPV (70.6% vs. 88.3%, p-value: 0.005) were significantly lower among lesions with microcalcifications without other radiological findings, while the SP (85.9% vs. 50.9%, p-value < 0.001) was higher. The enhancement of microcalcifications has low sensitivity in predicting malignancy. However, in certain controversial cases, the absence of CESM enhancement due to its high negative predictive value can help to reduce the number of biopsies for benign lesions.

Recent technological advances in the field of artificial intelligence hold promise in addressing medical challenges in breast cancer care, such as early diagnosis, cancer subtype determination and molecular profiling, prediction of lymph node metastases, and prognostication of treatment response and probability of recurrence. Radiomics is a quantitative approach to medical imaging, which aims to enhance the existing data available to clinicians by means of advanced mathematical analysis using artificial intelligence. Various published studies from different fields in imaging have highlighted the potential of radiomics to enhance clinical decision making. In this review, we describe the evolution of AI in breast imaging and its frontiers, focusing on handcrafted and deep learning radiomics. We present a typical workflow of a radiomics analysis and a practical "how-to" guide. Finally, we summarize the methodology and implementation of radiomics in breast cancer, based on the most recent scientific literature to help researchers and clinicians gain fundamental knowledge of this emerging technology. Alongside this, we discuss the current limitations of radiomics and challenges of integration into clinical practice with conceptual consistency, data curation, technical reproducibility, adequate accuracy, and clinical translation. The incorporation of radiomics with clinical, histopathological, and genomic information will enable physicians to move forward to a higher level of personalized management of patients with breast cancer.

Abstract Background Breast cancer screening through mammography is crucial for early detection, yet the demand for mammography services surpasses the capacity of radiologists. Artificial intelligence (AI) can assist in evaluating microcalcifications on mammography. We developed and tested an AI model for localizing and characterizing microcalcifications. Methods Three expert radiologists annotated a dataset of mammograms using histology-based ground truth. The dataset was partitioned for training, validation, and testing. Three neural networks (AlexNet, ResNet18, and ResNet34) were trained and evaluated using specific metrics including receiver operating characteristics area under the curve (AUC), sensitivity, and specificity. The reported metrics were computed on the test set (10% of the whole dataset). Results The dataset included 1,000 patients aged 21–73 years and 1,986 mammograms (180 density A, 220 density B, 380 density C, and 220 density D), with 389 malignant and 611 benign groups of microcalcifications. AlexNet achieved the best performance with 0.98 sensitivity, 0.89 specificity of, and 0.98 AUC for microcalcifications detection and 0.85 sensitivity, 0.89 specificity, and 0.94 AUC of for microcalcifications classification. For microcalcifications detection, ResNet18 and ResNet34 achieved 0.96 and 0.97 sensitivity, 0.91 and 0.90 specificity and 0.98 and 0.98 AUC, retrospectively. For microcalcifications classification, ResNet18 and ResNet34 exhibited 0.75 and 0.84 sensitivity, 0.85 and 0.84 specificity, and 0.88 and 0.92 AUC, respectively. Conclusions The developed AI models accurately detect and characterize microcalcifications on mammography. Relevance statement AI-based systems have the potential to assist radiologists in interpreting microcalcifications on mammograms. The study highlights the importance of developing reliable deep learning models possibly applied to breast cancer screening. Key points • A novel AI tool was developed and tested to aid radiologists in the interpretation of mammography by accurately detecting and characterizing microcalcifications. • Three neural networks (AlexNet, ResNet18, and ResNet34) were trained, validated, and tested using an annotated dataset of 1,000 patients and 1,986 mammograms. • The AI tool demonstrated high accuracy in detecting/localizing and characterizing microcalcifications on mammography, highlighting the potential of AI-based systems to assist radiologists in the interpretation of mammograms. Graphical Abstract

Abstract Background Image-guided vacuum-assisted breast biopsy (VABB) of the tumour bed, performed after neoadjuvant therapy, is increasingly being used to assess residual cancer and to potentially identify to identify pathological complete response (pCR). In this study, the accuracy of preoperative VABB specimens was assessed and compared with surgical specimens in patients with triple-negative or human epidermal growth factor receptor 2 (HER2)-positive invasive ductal breast cancer after neoadjuvant therapy. As a secondary endpoint, the performance of contrast-enhanced MRI of the breast and PET–CT for response prediction was assessed. Methods This single-institution prospective pilot study enrolled patients from April 2018 to April 2021 with a complete response on imaging (iCR) who subsequently underwent VABB before surgery. Those with a pCR at VABB were included in the primary analysis of the accuracy of VABB. The performance of imaging (MRI and PET–CT) was analysed for prediction of a pCR considering both patients with an iCR and those with residual disease at postneoadjuvant therapy imaging. Results Twenty patients were included in the primary analysis. The median age was 44 (range 35–51) years. At surgery, 18 of 20 patients showed a complete response (accuracy 90 (95 per cent exact c.i. 68 to 99) per cent). Only two patients showed residual ductal intraepithelial neoplasia of grade 2 and 3 respectively. In the secondary analysis, accuracy was similar for MRI and PET–CT (77 versus 78 per cent; P = 0.76). Conclusion VABB in patients with an iCR might be a promising method to select patients for de-escalation of surgical treatment in triple-negative or HER2-positive breast cancer. The present results support such an approach and should inform the design of future trials on de-escalation of surgery.

The mainstay treatment for patients with immediate resectable pancreatic cancer remains upfront surgery, which represents the only potentially curative strategy. Nevertheless, the majority of patients surgically resected for pancreatic cancer experiences disease relapse, even when a combination adjuvant therapy is offered. Therefore, aiming at improving disease free survival and overall survival of these patients, there is an increasing interest in evaluating the activity and efficacy of neoadjuvant and perioperative treatments. In this view, it is of utmost importance to find biomarkers able to select patients who may benefit from a preoperative therapy rather than upfront surgical resection. Defined genomic alterations and a dynamic inflammatory microenvironment are the major culprits for disease recurrence and resistance to chemotherapeutic treatments in pancreatic cancer patients. Signal transduction pathways or tumor immune microenvironment could predict early recurrence and response to chemotherapy. In the last decade, distinct molecular subtypes of pancreatic cancer have been described, laying the bases to a tailored therapeutic approach, started firstly in the treatment of advanced disease. Patients with homologous repair deficiency, in particular with mutant germline BRCA genes, represent the first subgroup demonstrating to benefit from specific therapies. A fraction of patients with pancreatic cancer could take advantage of genome sequencing with the aim of identifying possible targetable mutations. These genomic driven strategies could be even more relevant in a potentially curative setting. In this review, we outline putative predictive markers that could help in the next future in tailoring the best therapeutic strategy for pancreatic cancer patients with a potentially curable disease.

Although gastroenteropancreatic neuroendocrine neoplasms (GEP-NENs) have always been considered rare tumors, their incidence has risen over the past few decades. They represent a highly heterogeneous group of neoplasms with several prognostic factors, including disease stage, proliferative index (Ki67), and tumor differentiation. Most of these neoplasms express somatostatin receptors on the cell surface, a feature that has important implications in terms of prognosis, diagnosis, and therapy. Although International Guidelines propose algorithms aimed at guiding therapeutic strategies, GEP-NEN patients are still very different from one another, and the need for personalized treatment continues to increase. Radical surgery is always the best option when feasible; however, up to 80% of cases are metastatic upon diagnosis. Regarding medical treatments, as GEP-NENs are characterized by relatively long overall survival, multiple therapy lines are adopted during the lifetime of these patients, but the optimum sequence to be followed has never been clearly defined. Furthermore, although new molecular markers aimed at predicting the response to therapy, as well as prognostic scores, are currently being studied, their application is still far from being part of daily clinical practice. As they represent a complex disease, with therapeutic protocols that are not completely standardized, GEP-NENs require a multidisciplinary approach. This review will provide an overview of the available therapeutic options for GEP-NENs and attempts to clarify the possible approaches for the management of these patients and to discuss future perspectives in this field.

Objective: During the last decades, advances in computing power, structured data, and algorithm development, developed a technology based on Artificial intelligence (AI) which is currently applied in medicine. Nowadays, the main use of AI in breast imaging is in decision support in mammography, where it facilitates human decision-making as opposed to replacing radiologists. In this paper, we analyze how AI is currently involved in radiological decision-making and how will change both interpretation efficacy and workflow efficiency in breast imaging. Mechanism: We performed a non-systematic review on Pubmed and Scopus and Web of Science electronic databases from January 2001 to January 2022, using the following keywords: artificial intelligence, machine and deep learning, breast imaging and mammography. Findings in Brief: Many retrospective studies showed that AI can match or even enhance performances of radiologists in mammography interpretation. However, to assess the real role of AI in clinical practice compelling evidence from accurate perspective studies in large cohorts is needed. Breast imaging must face with the exponential growth in imaging requests (and consequently higher costs) and a predicted reduced number of trained radiologists to read imaging and provide reports. To mitigate these urges, solution is being sought with increasing investments in the application of AI to improve the radiology workflow efficiency as well as patient outcomes. Conslusions: This paper show the background on the evolution and the application of AI in breast imaging in 2022, in addition to exploring advantages and limitations of this innovative technology, as well as ethical and legal issues that have been identified so far.

Background: We aimed to create a model of radiological and pathological criteria able to predict the upgrade rate of low-grade ductal carcinoma in situ (DCIS) to invasive carcinoma, in patients undergoing vacuum-assisted breast biopsy (VABB) and subsequent surgical excision. Methods: A total of 3100 VABBs were retrospectively reviewed, among which we reported 295 low-grade DCIS who subsequently underwent surgery. The association between patients’ features and the upgrade rate to invasive breast cancer (IBC) was evaluated by univariate and multivariate analysis. Finally, we developed a nomogram for predicting the upstage at surgery, according to the multivariate logistic regression model. Results: The overall upgrade rate to invasive carcinoma was 10.8%. At univariate analysis, the risk of upgrade was significantly lower in patients with greater age (p = 0.018), without post-biopsy residual lesion (p < 0.001), with a smaller post-biopsy residual lesion size (p < 0.001), and in the presence of low-grade DCIS only in specimens with microcalcifications (p = 0.002). According to the final multivariable model, the predicted probability of upstage at surgery was lower than 2% in 58 patients; among these 58 patients, only one (1.7%) upstage was observed, showing a good calibration of the model. Conclusions: An easy-to-use nomogram for predicting the upstage at surgery based on radiological and pathological criteria is able to identify patients with low-grade carcinoma in situ with low risk of upstaging to infiltrating carcinomas.

: The impact of COVID-19 on the world of breast cancer care has been unprecedented, with worrisome short- and long-term consequences, and there remains a long road ahead to recover and unbury the breast imaging departments from their current backlog. Radiologists have to consider what the new normal will be going forward. At present time, because of widescale COVID-19 vaccination, benign vaccine-related reactive lymphadenopathy is likely to be encountered in oncologic patients and we need data-driven guidelines to manage unilateral lymphadenopathy and avoid unnecessary biopsies. In the next years, some procedures like wearing masks and maintaining social distancing will probably remain in use, as radiologists show patients that they are concerned about patient safety. Accordingly, odds are it will incorporate novel protocols for patient safety, innovative technologies (such as telemedicine and Artificial Intelligence algorithms), and changes in radiology workflow to create an environment that feels safe to both patients and radiologists, preventing backlogs (preventive service must not to be declined anymore) and burnouts (we need to take medical staff's mental health seriously). However, there is hope on the horizon with new lessons learned from this pandemic that can help clear the backlog and improve the working in breast imaging departments to achieve what is most important: saving lives in the fight against breast cancer.

Purpose: In order to evaluate the use of un-enhanced magnetic resonance imaging (MRI) for detecting breast cancer, we evaluated the accuracy and the agreement of diffusion-weighted imaging (DWI) through the inter-reader reproducibility between expert and non-expert readers. Material and Methods: Consecutive breast MRI performed in a single centre were retrospectively evaluated by four radiologists with different levels of experience. The per-breast standard of reference was the histological diagnosis from needle biopsy or surgical excision, or at least one-year negative follow-up on imaging. The agreement across readers (by inter-reader reproducibility) was examined for each breast examined using Cohen’s and Fleiss’ kappa (κ) statistics. The Wald test was used to test the difference in inter-reader agreement between expert and non-expert readers. Results: Of 1131 examinations, according to our inclusion and exclusion criteria, 382 women were included (49.5 ± 12 years old), 40 of them with unilateral mastectomy, totaling 724 breasts. Overall inter-reader reproducibility was substantial (κ = 0.74) for expert readers and poor (κ = 0.37) for non- expert readers. Pairwise agreement between expert readers and non-expert readers was moderate (κ = 0.60) and showed a statistically superior agreement of the expert readers over the non-expert readers (p = 0.003). Conclusions: DWI showed substantial inter-reader reproducibility among expert-level readers. Pairwise comparison showed superior agreement of the expert readers over the non-expert readers, with the expert readers having higher inter-reader reproducibility than the non-expert readers. These findings open new perspectives for prospective studies investigating the actual role of DWI as a stand-alone method for un-enhanced breast MRI.

Italy has one of the highest COVID-19 clinical burdens in the world and Lombardy region accounts for more than half of the deaths of the country. Since COVID-19 is a novel disease, early impactful decisions are often based on experience of referral centres.We report the re-organisation which our institute (IEO, European Institute of Oncology), a cancer referral centre in Lombardy, went through to make our breast-imaging division pandemic-proof. Using personal-protective-equipment and innovative protocols, we provided essential breast-imaging procedures during COVID-19 pandemic without compromising cancer outcomes.The emergency management and infection-control-measures implemented in our division protected both the patients and the staff, making this experience useful for other radiology departments dealing with the pandemic.

Multiple synchronous (multifocal or multicentric) ipsilateral breast cancers with heterogeneous histopathology are a rare clinical occurrence, however, their incidence is increasing due to the use of MRI for breast cancer screening and staging. Some studies have demonstrated poorer clinical outcomes for this pattern of breast cancer, but there is no evidence to guide clinical practice. In this multidisciplinary review, we reflect on pathology and molecular characteristics, imaging findings, surgical management including conservation and reconstructive options and approach to the axilla, and the role of chemotherapy and radiotherapy. Multidisciplinary discussions appear decisive in planning an appropriate surgical choice and defining the correct systemic treatment tailored to each clinical condition.

The three-dimensional automated breast ultrasound system (3D ABUS) is a new device which represents a huge innovation in the breast ultrasound field, with several application scenarios of great interest. ABUS's aim is to solve some of the main defects of traditional ultrasound, such as lack of standardization, high level of skill non-reproducibility, small field of view and high commitment of physician time. ABUS has proven to be an excellent non-ionising alternative to other supplemental screening options for women with dense breast tissue; also, it has appeared to be very promising in daily clinical practice. The purpose of this paper is to present a summary of current applications of ABUS, focusing on clinical applications and future perspectives as ABUS is particularly promising for studies involving artificial intelligence, radiomics and evaluation of breast molecular subtypes.