Computer Vision Identifies Recurrent and Non-Recurrent Ductal Carcinoma in situ Lesions with Special Emphasis on African American Women

Abstract

ABSTRACTAlthough the existence of non-recurrent and recurrent forms of ductal carcinoma in situ (DCIS) of the breast are observed, no evidence-based test can make this distinction. This retrospective case-control study used archival DCIS samples stained with anti-phospho-Ser226-GLUT1 (glucose transporter type 1) and anti-phosphofructokinase type L (PFKL) antibodies. Immunofluorescence micrographs were used to create machine learning (ML) models of recurrent and non-recurrent biomarker patterns, which were evaluated in cross-validation studies. Clinical performance was assessed by holdout studies using patients’ whose data were not used in training. Micrographs were stratified by the recurrence probability of each image. Recurrent patients were defined by at least one image with a probability of recurrence>98% whereas non-recurrent patients had none. These studies demonstrated no false negatives, identified true positives, and uniquely identified true negatives. Roughly 20% of the microscope fields of recurrent lesions were computationally recurrent. Strong prognostic results were obtained for both Caucasian and African American women. Our machine tool provides the first means to accurately predict recurrent and non-recurrent patient outcomes. We suggest that at least some false positives were true positives that benefitted from surgical intervention. The intracellular locations of phospho-Ser226-GLUT1 and phosphofructokinase type L likely participate in cancer recurrences by accelerating glucose flux, a key feature of the Warburg Effect.