VAVignesh Amal Arasu

Vignesh Amal Arasu

*This is an algorithmically generated author record based on publicly available information.

Vignesh Amal Arasu has a total of 151 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Kaiser Permanente National Cross-Vendor Validation of Mammography Artificial Intelligence Computer-Aided Diagnosis Algorithms in a US-Representative Population

    by Vignesh A. Arasu, Tejomay Gadgil, Mark Westley, Albert Pu, Catherine Lee, Cara Smith Gueye, Jason Balkman, Dorota Wisner, Lili Ivansco, Lawrence Kushi

    Background: Artificial intelligence computer-aided diagnosis (AI CAD) algorithms for screening mammography have shown promise, but independent head-to-head comparisons of commercial algorithms on large, diverse cohorts remain limited. Methods: 786,124 mammography screening exam performed from January 1 2022 to December 31 2023 were identified from electronic medical records. Three FDA-cleared commercial algorithms (A, B, D) and the open-source academic model Mirai (C), were applied to the four standard screening views. Breast cancer within 12 months was ascertained through cancer registry linkage. Algorithms were compared by AUROC; sensitivity, specificity, and PPV at matched AI-positive rates of 5%, 10%, 15%, 20%, 25%; and the proportion of false negatives classified as AI-positive. Results: Of 707,922 examinations with complete data, 4,436 were associated with cancer. Radiologists recalled 6.8% of examinations, with a sensitivity of 67.2%. Algorithm D had the highest AUROC (0.849), followed by algorithm B (0.818), Mirai (0.817), and algorithm A (0.815). Algorithm D also had the highest sensitivity at every AI-positive rate, from 55.3% (95% CI: 54.1, 56.6) at 5% to 77.9% (95% CI: 76.9, 79.0) at 25%, with CIs that did not overlap those of the other algorithms. No algorithm matched radiologist sensitivity at AI-positive rates of 10% or less. Algorithm B flagged a proportion of radiologist false negatives similar to that of algorithm D from 5% through 20% (20.8% vs 20.5% at 5%), despite its lower standalone sensitivity. Conclusions: AI CAD algorithms differed meaningfully in performance on the same large, diverse screening examination cohort, and an open-source academic model performed comparably to two of the three commercial products. Independent local validation at clinically relevant operating points should precede adoption of mammography AI, and algorithm choice and threshold should be matched to the intended workflow.

  • Comparative 10-year performance of mammography artificial intelligence, polygenic, and clinical breast cancer risk models in the Kaiser Permanente Research Bank

    by Vignesh A Arasu, Tejomay Gadgil, Joseph H Rothstein, Stacey E Alexeeff, Ninah S Achacoso, Arjun Bhattacharya, Jason B Cord, Laura J Esserman, Woodward Galbraith, Lawrence D Gerstley, et al.

    Background: We compared performance across 3 breast cancer risk domains—clinical, polygenic, and mammography artificial intelligence—alone and in combination over a 10-year time horizon among women with a negative screening mammogram within a Kaiser Permanente Research Bank (KPRB) prospective cohort. Methods: The study included 82 957 women (61 962 non-Hispanic White, 7256 Asian, 3414 Black, and 5466 Latina) who enrolled in KPRB between 2003 to 2020. Women with a prior history of breast cancer or high/moderate-penetrant gene mutation were excluded. The negative screening mammogram (no clinically visible cancer) closest to enrollment was used to generate the Mirai mammography AI risk score. KPRB survey and electronic health record data were used to generate the Breast Cancer Surveillance Consortium version 3 (BCSCv3) clinical risk score. Genome-wide genotypes were used to compute the 313-SNP polygenic risk score, adjusted for genetic ancestry (PRS313adj). Risks of breast cancer (invasive or ductal carcinoma in situ) at 0 to 10 years after the mammogram were estimated using Cox models, with 5-fold cross-validation used to estimate the C-index. Results: During 10 years of follow-up, 2471 women developed breast cancer. The C-index (95% CI) for the combined model with all 3 risk scores (0.70; 95% CI = 0.69 to 0.71) was significantly higher than for univariate models with only the BCSCv3 (0.62; 95% CI = 0.61 to 0.63), PRS313adj (0.61; 95% CI = 0.60 to 0.62), or Mirai (0.66; 95% CI = 0.65 to 0.67) risk score. Conclusions: Integrating mammographic AI and polygenic risk scores with clinical risk models significantly improved breast cancer risk discrimination, supporting use of combined models for personalized screening and prevention.

  • Open Access

    Cross-Device Adaptation of Mirai for Mammography-Based Breast Cancer Risk Prediction

    by Adriana Sistig, Joseph H. Rothstein, Tejomay Gadgil, Ninah Achacoso, Stacey E. Alexeeff, Lawrence D. Gerstley, Robert J. Klein, Laurie R. Margolies, Albert Pu, Cara L. Smith Gueye, et al.

    Fine-tuning can adapt pretrained medical imaging models to new clinical datasets, but device-specific domain shifts may limit generalizability. We evaluated Mirai, a mammography-based deep learning model for breast cancer risk prediction, in a large screening cohort containing Hologic and General Electric (GE) full-field digital mammography systems, including GE Premium View (GE PV) and Tissue Equalization (GE TE) post-processing software. Native Mirai showed lower performance on TE images than on Hologic or PV images. Fine-tuning on TE images improved TE performance, particularly for short-term risk prediction, but substantially reduced performance on Hologic images, consistent with catastrophic forgetting. To mitigate this effect, we developed a device-invariant model using interleaved multi-device sampling and conditional adversarial training. This approach largely restored Hologic performance while maintaining improved TE performance, providing better robustness across heterogeneous imaging platforms. Comparison of cumulative and annual risk AUCs over a five-year time horizon further showed that performance gains were driven mainly by short- and intermediate-term predictions. These findings highlight both the value and dangers of device-specific fine-tuning and support balanced domain-adaptation strategies for deploying mammography-based risk models across diverse clinical imaging environments.

  • Fatty Breasts and Palpable Lumps: Diagnostic Mammogram Alone May Be Sufficient to Exclude Cancer

    by Dawn C Nwamuo, Grace H Lin, Shivani Gillon, Caitlin M Lydon, Amethyst Leimpeter, Jun Shan, Laurel A Habel, Ninah S Achacoso, Mark Westley, Tony Xu, et al.

    Objective: To assess whether diagnostic mammogram alone is sufficient to identify breast cancer in women with a palpable lump and almost entirely fatty breast density. Methods: All 2D full-field diagnostic mammograms performed in a large health system were retrospectively identified for patients assigned “almost entirely fatty” breast density from January 2014 through December 2018 who presented with a palpable lump and had at least 2 years of follow-up. Women were excluded if they had a history of breast cancer, ipsilateral surgery, or clinical suspicion of abscess. Other exclusions included axillary lumps, lumps located in the retroareolar or posterior one-third of the breast, and lumps in which the lump marker was visible in only 1 view. The investigators performed image-only review of the diagnostic mammogram, with blinding of the image reports and any ensuing outcomes. Subsequently, the associated US was assessed. Cancer outcomes were ascertained via linkage with the regional tumor registry. Results: A total of 1552 eligible women, mean age 58.7 years ± 13.4, with 1791 lumps, were identified. Of these, 1139 women (73%) had no mammographic correlate, and 86 women (6%) had cancers corresponding to their lump, with 1 cancer having no mammographic correlate. The risk of cancer with no mammographic correlate to the palpable lump was 0.1% (1/1139 ; 95% CI, 0.0-0.3). Conclusion: Diagnostic mammography alone has very low risk of missed cancer and is sufficient for evaluation of palpable symptoms in most women with fatty breast density.

  • Comparison of Mammography AI Algorithms with a Clinical Risk Model for 5-year Breast Cancer Risk Prediction: An Observational Study

    by Vignesh A. Arasu, Laurel A. Habel, Ninah S. Achacoso, Diana S. M. Buist, Jason B. Cord, Laura J. Esserman, Nola M. Hylton, M. Maria Glymour, John Kornak, Lawrence H. Kushi, et al.

    Background Although several clinical breast cancer risk models are used to guide screening and prevention, they have only moderate discrimination. Purpose To compare selected existing mammography artificial intelligence (AI) algorithms and the Breast Cancer Surveillance Consortium (BCSC) risk model for prediction of 5-year risk. Materials and Methods This retrospective case-cohort study included data in women with a negative screening mammographic examination (no visible evidence of cancer) in 2016, who were followed until 2021 at Kaiser Permanente Northern California. Women with prior breast cancer or a highly penetrant gene mutation were excluded. Of the 324 009 eligible women, a random subcohort was selected, regardless of cancer status, to which all additional patients with breast cancer were added. The index screening mammographic examination was used as input for five AI algorithms to generate continuous scores that were compared with the BCSC clinical risk score. Risk estimates for incident breast cancer 0 to 5 years after the initial mammographic examination were calculated using a time-dependent area under the receiver operating characteristic curve (AUC). Results The subcohort included 13 628 patients, of whom 193 had incident cancer. Incident cancers in eligible patients (additional 4391 of 324 009) were also included. For incident cancers at 0 to 5 years, the time-dependent AUC for BCSC was 0.61 (95% CI: 0.60, 0.62). AI algorithms had higher time-dependent AUCs than did BCSC, ranging from 0.63 to 0.67 (Bonferroni-adjusted P < .0016). Time-dependent AUCs for combined BCSC and AI models were slightly higher than AI alone (AI with BCSC time-dependent AUC range, 0.66-0.68; Bonferroni-adjusted P < .0016). Conclusion When using a negative screening examination, AI algorithms performed better than the BCSC risk model for predicting breast cancer risk at 0 to 5 years. Combined AI and BCSC models further improved prediction. © RSNA, 2023 Supplemental material is available for this article.


  • Open Access

    Comparison of Mammography Artificial Intelligence Algorithms for 5-year Breast Cancer Risk Prediction

    by Vignesh A. Arasu, Laurel A. Habel, Ninah S. Achacoso, Diana S.M. Buist, Jason B. Cord, Laura J. Esserman, Nola. M. Hylton, M. Maria Glymour, John Kornak, Lawrence H. Kushi, et al.

    PURPOSE: To examine the ability of 5 artificial intelligence (AI)-based computer vision algorithms, most trained to detect visible breast cancer on mammograms, to predict future risk relative to the Breast Cancer Surveillance Consortium clinical risk prediction model (BCSC v2). PATIENTS AND METHODS: In this case-cohort study, women who had a screening mammogram in 2016 at Kaiser Permanente Northern California with no evidence of cancer on final imaging assessment were followed through September 2021. Women with prior breast cancer or a known highly penetrant gene mutation were excluded. From the 329,814 total eligible women, a random subcohort of 13,881 women (4.2%) were selected, of whom 197 had incident cancer. All 4,475 additional incident cancers were also included. Continuous AI-predicted scores were generated from the index 2016 mammogram. Risk estimates were generated with the Kaplan-Meier method and time-varying area under the curve [AUC(t)]. RESULTS: For incident cancers at 0-1 year (interval cancer risk), BCSC demonstrated an AUC(t) of 0.62 (95% CI, 0.58-0.66), and the AI algorithms had AUC(t)s ranging from 0.66-0.71, all significantly higher than BCSC ( P < .05). For incident cancers at 1 to 5 years (5-year future cancer risk), BCSC demonstrated an AUC(t) of 0.61 (95% CI, 0.60-0.62), and the AI algorithms had AUC(t)s ranging from 0.63 to 0.67, all significantly higher than BCSC. Combined BCSC and AI models demonstrated AUC(t)s for interval cancer risk of 0.67-0.73 and for 5-year future cancer risk of 0.66-0.68. CONCLUSION: The AI mammography algorithms we evaluated had significantly higher discrimination than the BCSC clinical risk model for interval and 5-year future cancer risk. Combined AI and BCSC models had slightly higher discrimination than AI alone.


  • Breast MRI during Neoadjuvant Chemotherapy: Lack of Background Parenchymal Enhancement Suppression and Inferior Treatment Response

    by Natsuko Onishi, Wen Li, David C. Newitt, Roy J. Harnish, Fredrik Strand, Alex Anh-Tu Nguyen, Vignesh Amal Arasu, Jessica Gibbs, Ella F. Jones, Lisa J. Wilmes, et al.

    Lack of background parenchymal enhancement suppression at breast MRI during neoadjuvant chemotherapy may indicate inferior treatment response in hormone receptor–positive breast cancer. Background Suppression of background parenchymal enhancement (BPE) is commonly observed after neoadjuvant chemotherapy (NAC) at contrast-enhanced breast MRI. It was hypothesized that nonsuppressed ...


  • Open Access

    Predictive Value of Breast MRI Background Parenchymal Enhancement for Neoadjuvant Treatment Response among HER2− Patients

    by Vignesh A Arasu, Paul Kim, Wen Li, Fredrik Strand, Cody McHargue, Roy Harnish, David C Newitt, Ella F Jones, M Maria Glymour, John Kornak, et al.

    Women with advanced HER2− breast cancer have limited treatment options. Breast MRI functional tumor volume (FTV) is used to predict pathologic complete response (pCR) to improve treatment efficacy. In addition to FTV, background parenchymal enhancement (BPE) may predict response and was explored for HER2− patients in the I-SPY-2 TRIAL. Women with HER2− stage II or III breast cancer underwent prospective serial breast MRIs during four neoadjuvant chemotherapy timepoints. BPE was quantitatively calculated using whole-breast manual segmentation. Logistic regression models were systematically explored using pre-specified and optimized predictor selection based on BPE or combined with FTV. A total of 352 MRI examinations in 88 patients (29 with pCR, 59 non-pCR) were evaluated. Women with hormone receptor (HR)+HER2− cancers who achieved pCR demonstrated a significantly greater decrease in BPE from baseline to pre-surgery compared to non-pCR patients (odds ratio 0.64, 95% confidence interval (CI): 0.39–0.92, P = 0.04). The associated BPE area under the curve (AUC) was 0.77 (95% CI: 0.56–0.98), comparable to the range of FTV AUC estimates. Among multi-predictor models, the highest cross-validated AUC of 0.81 (95% CI: 0.73–0.90) was achieved with combined FTV+HR predictors, while adding BPE to FTV+HR models had an estimated AUC of 0.82 (95% CI: 0.74–0.92). Among women with HER2− cancer, BPE alone demonstrated association with pCR in women with HR+HER2− breast cancer, with similar diagnostic performance to FTV. BPE predictors remained significant in multivariate FTV models, but without added discrimination for pCR prediction. This may be due to small sample size limiting ability to create subtype-specific multivariate models.


  • Open Access

    Comparison of Segmentation Methods in Assessing Background Parenchymal Enhancement as a Biomarker for Response to Neoadjuvant Therapy

    by Alex Anh-Tu Nguyen, Vignesh A. Arasu, Fredrik Strand, Wen Li, Natsuko Onishi, Jessica Gibbs, Ella F. Jones, Bonnie N. Joe, Laura J. Esserman, David C. Newitt, et al.

    Breast parenchymal enhancement (BPE) has shown association with breast cancer risk and response to neoadjuvant treatment. However, BPE quantification is challenging, and there is no standardized segmentation method for measurement. We investigated the use of a fully automated breast fibroglandular tissue segmentation method to calculate BPE from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) for use as a predictor of pathologic complete response (pCR) following neoadjuvant treatment in the I-SPY 2 TRIAL. In this trial, patients had DCE-MRI at baseline (T0), after 3 weeks of treatment (T1), after 12 weeks of treatment and between drug regimens (T2), and after completion of treatment (T3). A retrospective analysis of 2 cohorts was performed: one with 735 patients and another with a final cohort of 340 patients, meeting a high-quality benchmark for segmentation. We evaluated 3 subvolumes of interest segmented from bilateral T1-weighted axial breast DCE-MRI: full stack (all axial slices), half stack (center 50% of slices), and center 5 slices. The differences between methods were assessed, and a univariate logistic regression model was implemented to determine the predictive performance of each segmentation method. The results showed that the half stack method provided the best compromise between sampling error from too little tissue and inclusion of incorrectly segmented tissues from extreme superior and inferior regions. Our results indicate that BPE calculated using the half stack segmentation approach has potential as an early biomarker for response to treatment in the hormone receptor–negative and human epidermal growth factor receptor 2–positive subtype.