Abstract
Ductal carcinoma in situ (DCIS) spans a biologic continuum from atypical ductal hyperplasia (ADH) to high-grade lesions with variable risk of progression to invasive ductal carcinoma (IDC), yet morphologic assessment by hematoxylin and eosin (H&E) remains diagnostically limited, particularly at the benign versus ADH/low-grade DCIS boundary. TRPV4, a mechanosensitive ion channel with pathology-dependent subcellular localization in DCIS, offers a biologically motivated immunohistochemical (IHC) marker that may refine classification beyond routine H&E assessment. We tested whether deep learning models trained on TRPV4 IHC outperform H&E-based models across the DCIS progression spectrum. We assembled a multi-institutional cohort of paired H&E and TRPV4 IHC whole-slide images from 108 patients, comprising an internal development cohort (n=69), an external test cohort (n=39), yielding 24,248 annotated tiles. Histopathological tiles from annotated regions were grouped into four ordered classes: normal/benign, ADH/low-grade DCIS, high-grade DCIS, and IDC. Xception and EfficientNet-B0 convolutional neural networks were trained with patient-level 3-fold cross-validation on the development cohort and evaluated as ensembles on the external test cohort. On external patient-level testing, H&E ensembles achieved macro-F1 values of 0.43 to 0.44 and macro-AUC values of 0.73 to 0.80, whereas TRPV4 IHC ensembles improved performance to macro-F1 values of 0.68 to 0.72 and macro-AUC values of 0.91 to 0.92, corresponding to a 54.5-67.4% relative improvement in patient-level macro-F1. Per-class analyses showed the largest gains with TRPV4 IHC versus H&E for ADH/low-grade DCIS (AUC, 0.83–0.84 vs 0.70–0.81) and IDC (AUC, 0.74–0.79 vs 0.65–0.66). These findings support TRPV4 IHC as a mechanistically grounded complement to H&E that improves patient-level discrimination across the DCIS progression spectrum, with the strongest gains for ADH/low-grade DCIS and IDC, in a pilot multi-institutional setting.
Competing Interest Statement
I. Chung is an inventor on U.S. Patent No. 12,013,398 related to plasma-membrane relocalization based diagnostics. The patent is assigned to The George Washington University, is not licensed, and no commercial products are currently in development. No authors have received or will receive financial compensation related to this patent. The authors declare no other competing interests.
Funding Statement
This work was supported by the Elsa Pardee Foundation Award, the George Washington University Technology Maturation Award, the George Washington Cancer Center, and the Katzen Cancer Research Pilot Award.
Author Declarations
I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.
Yes
The details of the IRB/oversight body that provided approval or exemption for the research described are given below:
Institutional Review Board of The George Washington University gave ethical approval for this work and waived informed consent for the use of de-identified archival breast pathology material (protocol NCR203065, Targeting biomechanical alterations in ductal carcinoma in situ). Only authorized study personnel had access to direct identifiers for linkage purposes. Identifiers were replaced with study codes prior to analysis, and all model development and evaluation were performed on coded/de-identified data. Institutional Review Board of the University of Virginia gave ethical approval for this work and waived informed consent for the use of de-identified archival breast pathology material. Samples were provided by the National Cancer Institute Cooperative Human Tissue Network and were received by the study team in coded/de-identified form.
I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals.
Yes
I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).
Yes
I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable.
Yes
Footnotes
This revised version improves the manuscript's clarity, pathology framing, and clinical focus. The revision clarifies the primary emphasis on the benign versus ADH/low-grade DCIS boundary, strengthens the pathology-facing presentation of representative H&E and TRPV4 IHC examples, adds a summary table of deep learning architecture and training settings, clarifies dataset composition and patient-level cross-validation, adds a comparison table with related DCIS-relevant deep learning studies, and substantially revises the Discussion, limitations, and future-work sections.
Data availability
De-identified data and analysis code will be provided by the corresponding author upon reasonable request, subject to institutional policies and data use agreements.
