ThermoBreast - Non-contact Breast Cancer Imaging Using AI-enhanced Thermography.

Overview

Breast cancer is one of the most worrisome health concerns facing women. Early detection and active patient monitoring are crucial to survival. The chances of a cure are high when detected and treated in the early stages. Standard breast cancer diagnostic methods such as mammography, ultrasound, and MRI have limitations such as ionizing radiation, high false-positive rates, and/or high expenses. Medical Thermography might overcome these limitations: It is a non-invasive , adjunctive physiologic imaging technology that uses a high-resolution infrared camera and computer processing to produce an image (thermogram) of a patient's skin surface temperatures. It is a non-contact screening method, which does not involve radiation exposure or invasive procedures, and is safe for both the patient and the trained personnel performing the screening. While mammography and ultrasound depend primarily on structural and anatomical variation of the tumor from the surrounding breast tissue, thermography detects pathophysiological changes within the breast such as metabolic and vascular changes caused by cancer. The heat transfer in the body is conducted by the circulatory system; hence, pathologies identified by thermography are generally associated with changes in blood perfusion. To date, there has been no completed or ongoing large-scale, prospective, multicenter, international study that evaluates the diagnostic performance of thermal video streams coupled with advanced Artificial Intelligence algorithms for early-stage breast cancer screening and diagnosis. The proposed study will be crucial for the development of a new imaging modality that aims to be both cost-effective and to carry a minimal level of risk, facilitating screening of women of all age groups and breast densities, enabling early detection of abnormalities caused by malignant processes and improving patient monitoring.

This study is designed as a multicenter, prospective, blinded, three cohorts, diagnostic trial. Patients will be recruited at 11 centers in France, Germany, Ireland, Israel, Slovenia, Lithuania, and the US. The primary objective is to compare the diagnostic performance of advanced image processing AI models to automatically predict breast malignancy based on thermograms and individual patient data collected during breast examination (ThermoBreast) with routine breast cancer screening and imaging. The study population consists of women undergoing routine screening for breast cancer or diagnostic evaluation of suspicious breast masses. The study will recruit women into three cohorts: Screening cohort, High-Risk Screening, and Diagnostics cohort. Participants recruited for the study will be assigned to a dedicated cohort based on the reason of the visit. There will be two visits for study participants, partially including study specific and routine procedures, who meet the inclusion criteria in the screening procedure. The first trial visit (V1.1) will take place to provide the patient with detailed information on the study, its aims, the ThermoBreast procedure, and its risks. In- and exclusion criteria will be checked. Informed consent will be obtained. If the patient consents, the study specific ThermoBreast procedure (index test) and a study specific questionnaire regarding user experience will be performed. Moreover, the patient will undergo the routine first breast cancer screening round (screening cohort) or the routine first breast cancer diagnostics round (diagnostic cohort) according to national guidelines which serve as a reference test. In the screening cohort, the patient may undergo routine breast diagnostics following an irregular first screening round (V1.2) according to national guidelines. The second trial visit (V2) serves as a follow-up visit for the reference test - no study specific interventions will be performed here. Patients will undergo routine breast cancer screening or diagnostics according to national guidelines which serve as a reference test. In the screening cohort, the patient may undergo routine breast diagnostics following an irregular second screening round (V2.2) according to national guidelines. No study specific interventions will be performed during this trial visit. There will be no randomization within this study design. The findings of thermal screening will not be disclosed to patients and physicians, to avoid any anxiety or influence on the physicians' decision-making, as the proposed method is under study and development. The ThermoBreast system will only record the thermal imaging but will not automatically display any final assessment with respect to risk of breast cancer. AI-based evaluation of dynamic breast thermography imaging (=ThermoBreast, index test) will serve as the outcome measure. The ThermoBreast evaluation will be performed independently of the routine breast diagnostic evaluation. Some baseline risk variables (e.g. patient age, hormonal status) may be included in the ThermoBreast evaluation. ThermoBreast will provide a risk score for invasive breast cancer on a per-patient basis (not per breast). Routine breast diagnostics will serve as a reference test against which the ThermoBreast outcome will be compared. Depending on national guidelines and clinical scenarios, routine breast cancer diagnostics can consist of several routine procedures, including clinical examination, ultrasound, mammography, tomosynthesis, MRI, follow-up imaging, biopsy, surgical excision, and histopathologic evaluation. All procedures will be performed as indicated and specified by the respective, current national guidelines. The AI-based evaluation will use state-of-the-art image analysis techniques. The algorithm will be trained on data acquired during the training phase of the trial. For the second 6 month validation phase of the trial, a data lock will ensure that outcomes with respect to the reference test (routine breast diagnostics) will not be available anymore for the team responsible for the ThermoBreast algorithm development. A positive reference test is defined as histopathologic diagnosis of invasive breast cancer. A negative reference test is defined as * For the hypothesis without follow-up: a first study visit without biopsy indication (breast imaging BI-RADS 3,2 or 1), or a benign histopathologic evaluation. * For the hypothesis with follow-up: two regular study visits (breast imaging BI-RADS 2 or 1), a regular follow-up imaging after an irregular screening/imaging finding (BI-RADS 3 followed by BI-RADS 2 or 1), or a benign histopathologic evaluation. Analyses will be performed on a per-patient basis. In case of multiple breast masses per patient, the one with the highest risk of breast cancer or histopathological confirmed invasive breast cancer will be considered for the primary efficacy analysis