Lie Cai

📌 Lymph node surgery and CDK4/6 inhibitors in early breast cancer: a pooled analysis from five randomised trials 🔗 thelancet.com/journals/lanon… @OncoAlert #OncoAlertAF @TheLancetOncol @andrepfob

Axillary surgery → CDK4/6 eligibility: what’s the real gain? 🔍5 RCTs, n=19,541 ➡️Patients needed to prevent 1 event: → iDFS: 106–123 → DDFS: 119–129 → OS: 345–807 CDK4/6 benefit exists, but: → Grade ≥3 AEs: ~50–60% → ↑ Lymphedema: +3.9% (SLNB), +10.6% (cALND) 💬Using axillary surgery solely to guide CDK4/6 therapy → high NNT, added morbidity, marginal OS gain thelancet.com/journals/lanon…

At #EBCC15, the first ED session explored how AI is already shaping breast cancer care and what still needs to be addressed. From screening to research, AI shows real impact. But one message stood out: “AI is only as good as the data behind it.” Human oversight remains essential.

@Abdominal_Rad @RadGeek @LetDdiceFlyHigh @ClinMedJournals @SocietyAbdRad Honored to be selected as one of the Pinnacle Papers in 2025 published on @Abdominal_Rad we summarized the application of artificial intelligence in abdominal and pelvic ultrasound. rdcu.be/eZgbJ @andrepfob

2025 Awards are now live, recognizing: • Editorial leadership • Peer review contributions • Pinnacle Papers & top-downloaded articles 🔗 link.springer.com/journal/261/up… Congratulations to all editors, reviewers, and authors @RadGeek @LetDdiceFlyHigh @ClinMedJournals @SocietyAbdRad

📣Special Issue Call for Papers: Aims to promote interdisciplinary dialogue and showcase cutting-edge research in intersection of multi-omics biology, artificial intelligence, and drug resistance biology. We welcome submissions of high-level research/reviews. @cdrjournal

Honored to be selected as a member of the Early-career Editorial Board of Cancer Drug Resistance. 🎉 IF 5.2, JCR Q1. We welcome submissions in the field of cancer drug resistance research.📢 @cdrjournal

🎧Listen to our Dec podcast with @breastDoktor and @andrepfob discussing the Lucerne Toolbox 3 (📑thelancet.com/journals/lanon…), digital health and AI in early breast cancer Spotify: open.spotify.com/show/3JOj0ClP6… Apple: podcasts.apple.com/us/podcast/the…

Check out 3 fantastic examples linked in our editorial: 📌Digital path biomarker identifying men with #prostatecancer who can skip long-term ADT 📌 Foundation model predicting GI cancer prognosis & chemo benefit 📌 Deep learning radiomics to 🔽unnecessary #breast biopsies by 23%

Submit your #AI research to @JCO_ASCO! We are publishing impactful #AI research that shapes #oncology practice. Our new editorial outlines our review standards for AI submissions - from predictive models to #genAI. -> ascopubs.org/doi/10.1200/JC…

AI-guided shear wave elastography model showed similar accuracy to expert readings of B-mode ultrasound for diagnosis of BI-RADS 3/4 masses #BreastCancer 🔸 Significantly reduced false positives (20.4% vs 53.8%; P < .001) From Lie Cai, MD; @DrGolatta ascopost.com/news/september…

Deep Learning Model for Breast Shear Wave Elastography to Improve #BreastCancer Diagnosis (INSPiRED 006) ascopubs.org/doi/abs/10.120… This international, multicenter study evaluated a deep learning model trained on shear wave elastography (SWE) images (AI-SWE) to improve diagnosis of BI-RADS 3 and 4 breast masses. Using data from over 1,200 patients across 12 sites, the model achieved high accuracy (AUROC ~0.93–0.94) in two external validation cohorts, including data from newer SWE software. Compared with B-mode ultrasound alone, AI-SWE maintained equivalent sensitivity (~98%) but reduced false positives by up to 62%, suggesting a potential role in lowering unnecessary biopsies. These findings support AI-SWE as a promising tool to complement breast cancer diagnostics, warranting further integration into clinical workflows. @OncoAlert 🚨 @liecaii @andrepfob @drcgibbons @DrGolatta

Deep Learning Model for Breast Shear Wave Elastography to Improve Breast Cancer Diagnosis (INSPiRED 006) buff.ly/ruQKOjI This study evaluated the use of a deep learning model applied to shear wave elastography (AI-SWE) for improving the diagnosis of BI-RADS 3 and 4 breast masses. Using data from an international multicenter trial across 12 institutions, the model was trained on 4,026 images from 924 patients and externally validated on two independent cohorts. AI-SWE achieved high diagnostic accuracy with AUROCs of 0.94 and 0.93, matching the sensitivity of B-mode ultrasound while significantly reducing false-positive rates by 38–62%, potentially decreasing unnecessary biopsies. These findings suggest that AI-SWE could complement existing imaging methods, and further research should assess its role in multimodal #BreastCancer diagnostics. @liecaii @andrepfob @drcgibbons @DrGolatta #OncoAlertAF @nataliagandur @acampsmalea @BRicciutiMD @yekeduz_emre @HHorinouchi @FadiHaddad_MD @Abdallah81MD @FernandoOnco @ElisaAgostinett @to_be_elizabeth @bavilima @realbowtiedoc @Erman_Akkus @Lucarecco @GaiaGriguolo @JankovicK @MarioBalsaMD @DrMirallas @GIMedOnc @OscarTahuahua @matteolambe @aftimosp @double_whammied @stage4kelly @coffeemommy @itsnot_pink @maryam_lustberg @raalbany @stolaney1 @LoiSher @jamecancerdoc @JavierCortesMD @JaniceTNBCmets @itrisabel

Deep Learning Model for Breast Shear Wave Elastography to Improve #BreastCancer Diagnosis (INSPiRED 006): An International, Multicenter Analysis. Co-authored by @andrepfob, @liecaii, @drcgibbons, et al. Read the full article. brnw.ch/21wVbaO #bcsm

6/ What’s new with AI-SWE? ✔️ Evaluates the whole lesion, not just a cutoff. ✔️ Maintains sensitivity (~98%). ✔️ Cuts false positives (38–62%). ✔️ Works across multiple centers & software versions. From Breast Center Heidelberg to the world 🌍

5/ Trained on 4,026 images from 7 countries using EfficientNetB1 CNN, validated internationally. AI sees patterns, not just peak values.

3/ Our AI doesn’t use a single cutoff. It analyzes the whole SWE image, learning the full distribution of stiffness across a tumor. 4/ Result: AI-SWE = 62% fewer false positives. That means fewer unnecessary biopsies, while keeping 98% cancer detection.

2/ The problem: SWE is promising but limited. Doctors rely on velocity cutoffs (e.g. 3–8 m/s). Cutoffs vary by tissue type → no clear standard, limited reproducibility. Adoption stalled.

1/ 🚨 New in Journal of Clinical Oncology @JCO_ASCO @ASCO We built and validated the largest model for AI-enhanced shear wave elastography (AI-SWE) in breast cancer. 👉 doi.org/10.1200/JCO-24… With @liecaii @DrCGibbons @andrepfob #AIinMedicine #BreastCancer

🚀Honored to share our study with @andrepfob on @JCO_ASCO A deep learning model using shear wave elastography that showed comparable performance in detecting breast malignancy to human experts, and significant reduced unnecessary biopsies ascopubs.org/doi/10.1200/JC…