Breast cancer is a global health concern, ranking as the second leading cause of death among women. Current screening methods, such as mammography, face limitations, particularly for women under 50 due to radiation concerns and frequency of examination restrictions. MammoWave, utilizing microwave signals (1 to 9 GHz), emerges as an innovative and safe technology for breast cancer detection. This paper focuses on the numerical data extracted from MammoWave, presenting a hierarchical approach to address challenges posed by a diverse dataset of over 1000 samples from two European hospitals. The proposed approach involves unsupervised clustering to classify data into two main groups, followed by binary classification within each group to distinguish healthy and non-healthy cases. Careful consideration is given to feature extraction methods and classifiers at each step. The unique influence of sub-bands within the 1 to 9 GHz range on the diagnosis model is observed, leading to the selection of suitable sub-bands, feature extraction methods, and classification models. An optimization algorithm and a defined cost function are employed to achieve high and balanced sensitivity, specificity, and accuracy values. Experimental results showcase a promising overall balanced performance of around 70 %, representing a significant milestone in breast cancer detection using microwave imaging. MammoWave, with its novel approach, provides a solution that overcomes age and frequency of examination related limitations associated with existing screening methods, contributing to enhanced breast health monitoring for a broader population.

In this paper, we present preliminary outcomes derived from a prospective multicentric clinical trial focusing on microwave breast imaging involving 336 women. This investigation was conducted as part of the RadioSpin project. Our primary objective was to evaluate the prospective performance of MammoWave, our microwave imaging device, in differentiating between breasts with and without radiological findings, utilizing specific microwave images features' thresholds. Beyond this primary assessment, we explored the use of individual frequency sub-bands to categorize breasts into two groups: healthy (without findings or with benign findings) and non-healthy (malignant findings), drawing on features identified in a previous clinical trial. Our findings reveal a sensitivity of 72% in detecting radiological findings and a noteworthy 78% in identifying non-healthy (cancerous) breasts.

This paper presents an innovative paradigm for breast cancer detection by leveraging a Support Vector Machine (SVM) based model fueled with numerical data obtained from the cutting-edge MammoWave device. Operating in the microwave spectrum between 1 to 9 GHz and boasting a 5 MHz sampling rate, MammoWave emerges as a groundbreaking solution, specifically addressing the limitations posed by conventional methods, particularly for women under 50. This technological advancement opens a promising avenue for more frequent and precise breast health monitoring. To enhance the efficacy of the SVM model, our research introduces a metaheuristic-based methodology, strategically navigating the selection of frequencies crucial for breast cancer detection within the MammoWave dataset. Overcoming the challenge of judicious frequency selection, our approach employs wrapper methods in metaheuristic algorithms. These algorithms iterate through subsets of frequencies, guided by the SVM model's performance, culminating in the identification of the optimal frequency subset that significantly refines precision in breast cancer detection. Moreover, a novel cost function is proposed to strike a balanced trade-off between sensitivity and specificity, ensuring an acceptable accuracy rate. The results exhibit a noteworthy 10% increase in specificity, a milestone achievement for the MammoWave device, yielding an overall detection rate of approximately 62%. This research underscores the potential of seamlessly integrating metaheuristic algorithms into frequency selection, thereby contributing significantly to the ongoing refinement of MammoWave's capabilities in breast cancer detection.

In this paper, our primary objective is to present an AI-based model designed for the detection of breast cancer through the analysis of microwave data. The data utilized in this study were collected by MammoWave scanner, with the magnitude of the S21 complex number serving as the raw data input. We employed an Auto Encoder-Decoder for feature extraction, followed by the utilization of a Probabilistic Neural Network for the detection of malignant lesions. The results obtained indicate that although there is an imbalance in the distribution between healthy and non-healthy classes, our proposed method demonstrates the ability to address this challenge effectively through fine-tuning. Our approach achieves an approximate 60% rate for accuracy, sensitivity, and specificity, showcasing its promising potential in breast cancer detection.

MammoWave, a novel microwave breast imaging device, employs a frequency range spanning from 1 to 9 GHz to acquire dielectric properties information from the breast. This study conducts a comprehensive analysis of various sub-bands to identify those or their combinations that enhance the efficacy of cancer detection. The findings reveal that leveraging the combined information from the 5-6, 7-8, and 8-9 GHz sub-bands yields notable improvements. These results underscore the potential of strategically utilizing specific frequency sub-bands for enhanced performance in breast cancer detection using MammoWave.

In this work, we present our preliminary findings from a prospective multicentric microwave breast imaging clinical trial involving 218 women, within the framework of the RadioSpin project. In addition to the prospective performance evaluation of our microwave imaging device, MammoWave, in classifying breasts with and without radiological findings (using features with appropriate thresholds), we conducted further investigations and analyses to assess the impact of individual frequency sub-bands and the classification of breasts into healthy (without any findings or with benign findings) and non-healthy (malignant findings), based on selected features obtained from a prior clinical trial. Our results reveal the potential for achieving a sensitivity of up to 78% in the detection of radiological findings and up to 89% in the detection of non-healthy (with cancer) breasts.

Microwave breast imaging is being investigated by research groups worldwide for its promising applications in early cancer detection, overcoming key limitations of conventional imaging systems. In this framework, artificial intelligence may play an important role to enhance the performances of new systems, based on this novel technology, for breast cancer detection. Research is being carried out to demonstrate the potential of implementing machine learning tools that have already been investigated for conventional mammography and MRI. This work presents the retrospective implementation of several supervised machine learning approaches on the microwave data obtained by MammoWave device in the framework of a clinical trial. Two different approaches are explored and explained in detail: the application of artificial intelligence directly on the MammoWave raw data and on dedicated features extracted from microwave images. Both approaches lead to promising results with high (>80%) and quite balanced specificity and sensitivity.

In recent years, new technologies focused on dielectric principles have been developed for medical applications. Conductivity and permittivity of biological tissues have been described to vary among benign and malignant tissues, so many efforts are being made to implement new systems based on safe low-power microwaves able to capture these inhomogeneities for medical imaging. However, such conductivity and permittivity parameters are being investigated for several different applications. The dielectric characterization of tissues in vivo during surgeries or via excised tissue may offer clinicians new tools for optimizing hospital routines in the diagnostic pathway. This work presents the application of several Machine Learning (ML) approaches to dielectric data gathered from excised breast tissues using a novel open-ended coaxial probe.

Novel techniques, such as microwave imaging, have been implemented in different prototypes and are under clinical validation, especially for breast cancer detection, due to their harmless technology and possible clinical advantages over conventional imaging techniques. In the prospective study presented in this work, we aim to investigate through a multicentric European clinical trial (ClinicalTrials.gov Identifier NCT05300464) the effectiveness of the MammoWave microwave imaging device, which uses a Huygens-principle-based radar algorithm for image reconstruction and comprises dedicated image analysis software. A detailed clinical protocol has been prepared outlining all aspects of this study, which will involve adult females having a radiologist study output obtained using conventional exams (mammography and/or ultrasound and/or magnetic resonance imaging) within the previous month. A maximum number of 600 volunteers will be recruited at three centres in Italy and Spain, where they will be asked to sign an informed consent form prior to the MammoWave scan. Conductivity weighted microwave images, representing the homogeneity of the tissues' dielectric properties, will be created for each breast, using a conductivity = 0.3 S/m. Subsequently, several microwave image parameters (features) will be used to quantify the images' non-homogenous behaviour. A selection of these features is expected to allow for distinction between breasts with lesions (either benign or malignant) and those without radiological findings. For all the selected features, we will use Welch's t-test to verify the statistical significance, using the gold standard output of the radiological study review.

Microwave imaging is a safe new technology for breast imaging, avoiding ionizing radiation and the patient discomfort due to breast compression. In this paper we present results from the first prospective microwave breast imaging study where both symptomatic and asymptomatic subjects were recruited. For this purpose, a novel microwave imaging device (MammoWave) was examined on 353 women enrolled in the study to allow distinction between breasts with and without radiological findings. We investigated MammoWave’s performance using both features from the reconstructed images (prospective investigation) and through the use of artificial intelligence (AI) models (retrospective investigation). Our results indicate the importance of AI for specificity enhancement, and show a sensitivity greater than 80% for all the investigations. These high sensitivity values are maintained when considering breast cancers only.

This paper presents for the first time, preliminary measurement results obtained using an innovative microwave brain imaging device, named StrokeWave. Brain strokes are the world's second largest cause of death and the largest cause of adult disability, and their correct timely detection is vital for increasing patient's chance of survival. The StrokeWave prototype introduced here employs only two antennas which operate without matching medium, and uses a radar backpropagation algorithm based on the Huygens' principle. We present results obtained through both a cylindrical phantom and the head of a patient with hemorrhagic stroke who participated in our first round of clinical trials. Our measurement results indicate the capability of the algorithm in detecting and localizing the dielectric contrast, paving the potential for brain stroke detection and classification.

This work presents the first multicentric, single arm, prospective study to evaluate the ability of MammoWave, a microwave imaging prototype, in breast lesions detection. This study was the first breast microwave imaging study during which both symptomatic and asymptomatic subjects were recruited. MammoWave output consists of a selection of microwave images' features, determined prior to the beginning of the trials, to quantify images' non-homogenous behavior. Our results on 382 breasts show a sensitivity of 82% using a statistical significance of p<0.05, indicating MammoWave's ability in distinguishing breasts with and without radiological findings. This prospective clinical trial may pave the way for introducing microwave imaging into clinical practice, for assisting in identification of breast lesions in asymptomatic women of all ages, without safety limitations.

Microwave imaging for breast cancer detection has attracted growing global attention with a small number of prototypes advancing to the clinical trial stage. This investigation aims to provide an overview of MammoWave, a novel microwave-based imaging system for breast lesion detection and to assess its introduction into the clinical routine and its potential role in future breast screening programs. As a key focus of this work, we will describe in detail the various aspects of the clinical protocol procedure that has enabled us to perform a successful clinical trial. Obtained preliminary results indicate the ability of our device to distinguish breasts with no radiological finding and those with radiological findings, with a sensitivity of 89.6%.