Meng Jiang1, Chang-Li Li2, Xiao-Mao Luo3, Zhi-Rui Chuan3, Wen-Zhi Lv4, Xu Li5, Xin-Wu Cui6, Christoph F Dietrich7. 1. Department of Medical Ultrasound, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, 430030, PR China. 2. Department of Geratology, Hubei Provincial Hospital of Integrated Chinese and Western Medicine, 11 Lingjiaohu Avenue, Wuhan, 430015, PR China. 3. Department of Medical Ultrasound, Yunnan Cancer Hospital & the Third Affiliated Hospital of Kunming Medical University, Kunming, 650118, PR China. 4. Department of Artificial Intelligence, Julei Technology, Wuhan, 430030, PR China. 5. School of Biomedical Engineering, South-Central University for Nationalities, 182 Minyuan Road, Wuhan, 430074, PR China. 6. Department of Medical Ultrasound, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, 430030, PR China. Electronic address: cuixinwu@live.cn. 7. Department of Internal Medicine, Hirslanden Clinic, Schänzlihalde 11, Bern, 3013, Switzerland.
Abstract
PURPOSE: The aim of the study was to develop and validate a deep learning radiomic nomogram (DLRN) for preoperatively assessing breast cancer pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) based on the pre- and post-treatment ultrasound. METHODS: Patients with locally advanced breast cancer (LABC) proved by biopsy who proceeded to undergo preoperative NAC were enrolled from hospital #1 (training cohort, 356 cases) and hospital #2 (independent external validation cohort, 236 cases). Deep learning and handcrafted radiomic features reflecting the phenotypes of the pre-treatment (radiomic signature [RS] 1) and post-treatment tumour (RS2) were extracted. The minimum redundancy maximum relevance algorithm and the least absolute shrinkage and selection operator regression were used for feature selection and RS construction. A DLRN was then developed based on the RSs and independent clinicopathological risk factors. The performance of the model was assessed with regard to calibration, discrimination and clinical usefulness. RESULTS: The DLRN predicted the pCR status with accuracy, yielded an area under the receiver operator characteristic curve of 0.94 (95% confidence interval, 0.91-0.97) in the validation cohort, with good calibration. The DLRN outperformed the clinical model and single RS within both cohorts (P < 0.05, as per the DeLong test) and performed better than two experts' prediction of pCR (both P < 0.01 for comparison of total accuracy). Besides, prediction within the hormone receptor-positive/human epidermal growth factor receptor 2 (HER2)-negative, HER2+ and triple-negative subgroups also achieved good discrimination performance, with an AUC of 0.90, 0.95 and 0.93, respectively, in the external validation cohort. Decision curve analysis confirmed that the model was clinically useful. CONCLUSION: A deep learning-based radiomic nomogram had good predictive value for pCR in LABC, which could provide valuable information for individual treatment.
PURPOSE: The aim of the study was to develop and validate a deep learning radiomic nomogram (DLRN) for preoperatively assessing breast cancer pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) based on the pre- and post-treatment ultrasound. METHODS:Patients with locally advanced breast cancer (LABC) proved by biopsy who proceeded to undergo preoperative NAC were enrolled from hospital #1 (training cohort, 356 cases) and hospital #2 (independent external validation cohort, 236 cases). Deep learning and handcrafted radiomic features reflecting the phenotypes of the pre-treatment (radiomic signature [RS] 1) and post-treatment tumour (RS2) were extracted. The minimum redundancy maximum relevance algorithm and the least absolute shrinkage and selection operator regression were used for feature selection and RS construction. A DLRN was then developed based on the RSs and independent clinicopathological risk factors. The performance of the model was assessed with regard to calibration, discrimination and clinical usefulness. RESULTS: The DLRN predicted the pCR status with accuracy, yielded an area under the receiver operator characteristic curve of 0.94 (95% confidence interval, 0.91-0.97) in the validation cohort, with good calibration. The DLRN outperformed the clinical model and single RS within both cohorts (P < 0.05, as per the DeLong test) and performed better than two experts' prediction of pCR (both P < 0.01 for comparison of total accuracy). Besides, prediction within the hormone receptor-positive/human epidermal growth factor receptor 2 (HER2)-negative, HER2+ and triple-negative subgroups also achieved good discrimination performance, with an AUC of 0.90, 0.95 and 0.93, respectively, in the external validation cohort. Decision curve analysis confirmed that the model was clinically useful. CONCLUSION: A deep learning-based radiomic nomogram had good predictive value for pCR in LABC, which could provide valuable information for individual treatment.