RPS 402 - Updates in the evaluation of response to neoadjuvant therapy

Author Block: L. Guo, S. Du, L. Zhang; Shenyang/CN
Purpose or Learning Objective: To investigate the value of radiomics-based tumour heterogeneity changes after one cycle of neoadjuvant chemotherapy (NAC) using a high spatiotemporal resolution DCE-MRI for early prediction of pathological complete response (pCR) in patients with breast cancer.
Methods or Background: A total of 140 patients (training: test=7:3) with breast cancer underwent Differential Subsampling with Cartesian Ordering (DISCO) DCE-MRI before (t0) and after one cycle (t1) of NAC. Radiomic features were extracted from postcontrast early (CEe), peak (CEp) and delay (CEd) CE-MRI phases, respectively. Feature changes (t∆) were computed for each phase. For training, the strongest features that associated with pCR were selected. Logistic regression classifiers based on DCE-MRI at t0, t1, t∆ were constructed for differentiating pCR patients. The performance under different contrast phases was evaluated and compared. Clinicopathological information and the optimal imaging classifier were fused to enhance the predictive performance.
Results or Findings: Imaging classifiers using radiomic feature changes (t∆) achieved superior performance in both training and test cohort for CEe (AUC=0.840/0.825), CEp (AUC=0.772/0.744) and CEd (AUC=0.757/0.728) compared with DCE-MRI at t0 (CEe: AUC=0.725/0.689; CEp: AUC=0.609/0.572; CEd: AUC=0.678/0.611), with a significant difference of NRI and IDI (all p<0.05 in the training cohort). For DCE-MRI at t0 and t∆, the CEe was the optimal contrast phase for pCR classification. After adding receptor status to CEe at t∆, the classifier showed the highest AUC of 0.828 in the testing.
Conclusion: Using DISCO DCE-MRI, changes in radiomic features after one cycle of NAC, that reflect tumour heterogeneity changes could provide a non-invasive approach for early prediction of breast cancer response regardless of knowledge of the receptor status.
Limitations: Single-centre research; no voxel-level analysis; no biological explanation.
Ethics committee approval: Approved by the Ethics Committee of First Affiliated Hospital of China Medical University
Funding for this study: Funding was received from the National Scientific Foundation of China (81971695).

Author Block: R. J. Weinfurtner, N. Raghunand, O. Stringfield, M. Abdalah, B. Niell, M. C. Lee, H. Han, B. Czerniecki, H. Soliman; Tampa, FL/US
Purpose or Learning Objective: To quantitatively evaluate perfusion-based intratumoural habitats on dynamic contrast-enhanced (DCE) breast MRI to predict triple-negative breast cancer (TNBC) NAI/NAC response.
Methods or Background: TVEC is a modified oncolytic herpes simplex 1 virus. Subjects with TNBC in this phase II trial underwent ultrasound-guided intratumoral TVEC injections followed by neoadjuvant chemotherapy prior to surgery. Baseline and post-treatment breast MRIs were evaluated for partial vs complete response (mCR). MRI quantitative analysis was performed on dynamic contrast-enhanced T1-weighted images with voxels assigned 8 habitats based on two criteria: high (H) or low (L) maximum contrast enhancement per Otsu algorithm, and the sequentially numbered dynamic sequence of maximum enhancement (H1-4, L1-4). Then, the % habitat makeup (%HM) of tumour and whole breasts (%HM of habitat X=habitat X voxels/total voxels in the segmented volume) were evaluated and correlated with pathologic response (PR). Statistical analyses were performed using paired and unpaired t-tests, with p<0.05 considered statistically significant.
Results or Findings: Twenty patients were included in the study, and 11 achieved pathologic complete response (pCR). Prediction of pCR with mCR yielded accuracy of 65% (sensitivity 66.7%, specificity of 63.6%, PPV 60% and NPV 70%). Pre-NAI/NAC tumour %HM for each habitat differed significantly from whole breasts (p=0.001 or less). The %HM of habitat H1 (early phase, high enhancement) decreased the most after treatment (-18%, p=0.0004) followed by H2 and H3 (-14% and -4%, respectively). Conversely, %HM H4 (late phase, high enhancement) increased (12%, p=0.031). The H1-3 combination decreased 34%, and this was more pronounced in patients with pCR (-44% vs -22%, p=0.036).
Conclusion: In patients undergoing TVEC NAI/NAC treatment for TNBC, a decrease in %HM of early and mid-phase high enhancement habitats correlates with pathologic response.
Limitations: Sample size.
Ethics committee approval: IRB-approved.
Funding for this study: Funding was received from the Moffitt Cancer Center.

Author Block: P. Kapetas, P. Clauser, T. H. Helbich, R-I. Milos, P. A. Baltzer; Vienna/AT
Purpose or Learning Objective: To evaluate whether machine-learning (ML) models using selected clinical and radiomic features from pre-therapeutic, B-mode breast ultrasound (US) can predict response to neoadjuvant chemotherapy (NAC).
Methods or Background: 253 patients with invasive breast cancer undergoing NAC were included. One B-mode US image of each tumour from the baseline examination was selected. Tumours were manually segmented (without-ROI1 and with inclusion of surrounding tissue-ROI2) and overall 851 radiomic features were extracted using dedicated software. Two analyses were performed: first, principal component analysis was used for feature reduction and a multilayer perceptron neural network was trained using remaining radiomic and selected clinical features. 1/3 of the cases were used as an external validation set. Second, a decision tree using the exhaustive chi-squared automatic interaction detection method with 10-fold cross-validation was constructed. Postoperative histology was the reference standard. Diagnostic performance was evaluated using the area under the ROC curve (AUC).
Results or Findings: 104 patients (41.1%) achieved pathological complete response. In the first model, the combination of radiomic features, age and molecular subtype showed the highest AUC, both for ROI1 (0.715; 95%CI: 0.606-0.809) and for ROI2 (0.709; 95%CI: 0.599-0.803; p>0.05) in the validation set. The model resulted in 66.3% correct classifications considering ROI1 and 55.4% using ROI2 in the validation set. On the other hand, decision trees based on molecular subtype and 5-6 radiomic features showed AUCs of 0.852 (95%CI: 0.802-0.893) for ROI1 and 0.896 (95%CI: 0.852-0.931) for ROI2 (p>0.05). The corresponding correct classifications were 83.4% and 82.6%.
Conclusion: ML models based on radiomic features from baseline B-mode breast US and simple clinical features have the potential to predict response to NAC.
Limitations: This is a retrospective, monocentric study with a limited patient number.
Ethics committee approval: This was an IRB-approved retrospective study.
Funding for this study: Not applicable.

Author Block: J. R. D. Reis, O. Thomas, M. Lyngra, H. Schandiz, J. Boavida, K-I. Gjesdal, T. Sauer, J. Geisler, J. T. Geitung; Lørenskog/NO
Purpose or Learning Objective: To correlate MRI morphological-response patterns with histopathological tumour regression grading system based on tumour cellularity in locally advanced breast cancer (LABC) treated neoadjuvant with third-generation aromatase inhibitors.
Methods or Background: Fifty postmenopausal patients with ER-positive/HER-2 negative LABC treated with neoadjuvant letrozole and exemestane were given sequentially in an intra-patient cross-over regimen for at least 4 months with MRI response monitoring at baseline as well as after at least 2 and 4 months on treatment. The MRI morphological response pattern was classified into 6 categories: 0/complete imaging response, I/concentric shrinkage, II/fragmentation, III/diffuse, IV/stable and V/progressive. Histopathological tumour regression was assessed based on the recommendations from The Royal College of Pathologists regarding tumour cellularity.
Results or Findings: Following 2 and 4 months with therapy, the most common MRI pattern was pattern II (24/50 and 21/50, respectively). After 4 months of therapy, the most common histopathological tumour regression grade was grade 3 (21/50). After 4 months an increasing correlation is observed between MRI patterns and histopathology. The overall correlation, between the largest tumour diameter obtained from MRI and histopathology, was moderate and positive (r=0.50, P-value=2e-04). Among them, the correlation was highest in type IV (r=0.53).
Conclusion: The type II MRI pattern "fragmentation" was more frequent in the histopathological responder group; and types I and IV in the non-responder group. Type II pattern showed the best endocrine responsiveness and a relatively moderate correlation between sizes obtained from MRI and histology, whereas type IV pattern indicated endocrine resistance but the strongest correlation between MRI and histology.
Limitations: A single-centre study.
Ethics committee approval: The NEOLETEXE trial was registered on March 23rd, 2015 and approved by the regional ethical committee of the South-Eastern Health Region in Norway (registration number: REK-SØ-84-2015).
Funding for this study: Funding was received from the Bodil and Magne`s Cancer Research Fund.

Author Block: G. Cimino, M. Conti, A. Franco, A. De Filippis, C. Esposito, E. Bufi, D. Terribile, P. Belli, R. Manfredi; Rome/IT
Purpose or Learning Objective: To investigate the influence of radiological features measured on magnetic resonance imaging (MRI), digital mammography (DM), in addition to clinical factors on surgeons' choice between breast conservative surgery (BCS), oncoplastic surgery (OPS) or conservative mastectomy (CM), in patients treated with neoadjuvant chemotherapy (NACT).
Methods or Background: Preoperative MRI and DM of 255 women who underwent BCS (118 patients), OPS (50 patients) or CM (87 patients) after NACT in 2016-2021 were retrospectively reviewed. Preoperative radiological features were analysed including the extent of microcalcifications, tumour volume/breast volume ratio (TVBVR) on DM and MRI, multifocality/multicentricity, in addition to histotype, grading and clinical features (menopause, BRCA mutations, ptosis, previous breast surgery). All parameters were correlated with final surgeons’ choice.
Results or Findings: On univariate analysis BCS has proved to be the surgeons’ choice in patients with a unifocal tumour, the extent of microcalcifications <54,9 mm, TVBVR pre-NACT <0,42 % on DM and <3,53 % on MRI. OPS showed a significant correlation with multifocal tumour, the extent of microcalcifications 55-79,9 mm, TVBVR ratio pre-NACT 0,4-1,38 % on DM and 3,54-10,27 % on MRI. CM was correlated with multicentric tumours, the extent of microcalcifications >80 mm, TVBVR pre-NACT >1,38 % on DM and >10,27 % on MRI. On multivariate analysis these results were confirmed and were used to create a specific score to define which type of surgery to suggest to patients. ROC curves were produced for each type of surgery (AUC BCS=0,840, OPS=0,751 and MC=0.871).
Conclusion: Radiological features measured on DM and MRI play a fundamental role in the surgeons’ choice between BCS, OPS or CM in patients treated with NACT.
Limitations: No limitations were identified.
Ethics committee approval: Not applicable.
Funding for this study: Not applicable.

Author Block: M. Nadrljanski, I. Krusac, V. Urban, M. Mihajlovic; Belgrade/RS
Purpose or Learning Objective: Manually defined ROIs for computation of ADC-values may be technically challenging, prone to sampling errors and operator-dependent. Different ADC parameters (ADCdiff, ADCmax, ADCmin, ADCmean) may provide a more objective approach. ADCdiff corresponding to the level of intratumoral heterogeneity points more objectively to invasive components and may have prognostic significance.
Methods or Background: Retrospective analysis of (N=34) consecutive patients undergoing neoadjuvant chemotherapy (NACT) on 1.5T breast DWI-MRI (b50, b850 [s/mm2]), for: a.) Pre-treatment and b.) Early response assessment (after 2nd cycle of NACT) for morpho-dynamic and DWI parameters: ADCdiff, ADCmax, ADCmin, ADCmean, including the analysis of subgroups of patients with pCR (n1=11) and non-pCR (n2=23) following NACT.
Results or Findings: ADCdiff and ADCmean are significantly different between n1 and n2 on pre-treatment exam [10-6 x mm2/s]: (403.64+/-122.05 vs. 285.56+/-69.47; p=.008); (808.36+/-60.50 vs. 1015.69+/-78.97; p<.00001) and after the 2nd cycle: (407.36+/-92.00 vs. 261.78+/-60.69; p=.001); (963.27+/-67.51 vs. 1059.65+/-73.63; p=.002). In n1, ADCdiff is not different before and after 2nd cycle of NACT: (403.64+/-122.05 vs. 407.36+/-92.00; p=.096). In n2, ADCdiff is significantly different after 2nd cycle of NACT: (285.56+/-69.47 vs. 261.78+/-60.69; p=.048).
Conclusion: There are complex factors determining the response to NACT. There is no standardised ADC cut-off value determining the response. ADCdiff with indirect interpretation of microvascular changes and the rate of growth may also provide information regarding the response to NACT, as higher ADCdiff values in pCR group on pre-treatment MRI and early response assessment exam may contribute to better response and more adequate assessment.
Limitations: A relatively small number of patients in a single-centre retrospective study.
Ethics committee approval: Referent board approval was obtained for retrospective analysis.
Funding for this study: No funding was provided for this study.