Abstract PS2-44: Correlation of thermalytix - an artificial intelligence based thermal breast screening tool in detecting a breast lesion as benign, malignant or normal

Abstract

Abstract Aim: To correlate results of Thermalytix - an artificial intelligence (AI) based breast screening tool with mammography and the final diagnosis as established by histopathology Introduction: Screening and early diagnosis has proved to be useful in improving clinical outcome and survival of breast cancer patients. Mammography, the only breast cancer screening modality approved is not accessible or economically sustainable in low and middle-income countries (LMICs). Thermalytix, an artificial intelligence (AI) based breast cancer screening tool is low cost, portable and radiation free and uses AI-based techniques to analyze and interpret breast thermal images captured using a high-resolution infrared camera. Areas of high thermal activity are identified using relative temperature thresholding while vascular structures are analyzed using a novel image processing technique. These hotspots and vascular patterns are further analyzed to extract a set of features that are input to 3 pre-trained machine learning models to generate quantitative scores. The final scores generated based on these parameters are used to label a breast lesion as malignant, benign or normal at first screening. In this study, we present a comparative analysis of the results of thermalytix and mammography with final diagnosis as established by histopathological diagnosis. Methods:In this retrospective study, 65 patients who had undergone biopsy and histopathological examination of breast for symptoms such as breast lump, pain or discharge were recruited. Each patient had mammography and the non-invasive Thermalytix test done as a preliminary screening modality prior to biopsy. Automated results generated by Thermaytix were then independently compared with the histopathological diagnosis retrospectively. Similarly, the mammography results were also compared with the biopsy findings to compare the coherence of each modalities in detecting breast lesion independently. Results:Out of the 65 symptomatic patients who were followed up by biopsy for any suspicious lesion 48/65 were neoplastic with 37/65 malignant lesions 11/65 benign lesions. Rest 17/65 patients were non-neoplastic comprising of 13/65 inflammatory cases and 4/65 normal cases. 1.Thermalytix detected 31/37 malignant cases while mammography detected 35/37 as malignant. 2.Among benign lesions thermalytix came out positive for 6/11 cases while mammography did so in 10/11 cases. 3.For inflammatory cases thermalytix and mammography raised a suspicion for malignancy in 10/13 and 11/13 cases respectively. 4.All 4 cases which were normal on biopsy were all labelled as suspicious by mammography while 3 were labelled as suspicious by thermalytix. Conclusion:This preliminary study shows that thermalytix fared well with radiological findings in detecting breast lesions as benign, malignant or normal. The findings were somewhat skewed in favour of radiological findings as mammography was done as the primary screening test before the patients underwent biopsy and histopathological examination for any suspicious lesion. Thermalytix can act as a low cost, portable and radiation-free test that can be of great help in detecting neoplastic lesions of breast at first screening in in low- and middle-income countries (LMICs) where mammography is not accessible or economically scalable. With its automated scoring and annotations of potential neoplastic lesion Thermalytix is poised to be a promising modality for breast cancer screening in future. Citation Format: Chandan Kumar, Lakshmi Krishnan, Himanshu Madhu, Geetha Manjunath. Correlation of thermalytix - an artificial intelligence based thermal breast screening tool in detecting a breast lesion as benign, malignant or normal [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS2-44.