Multi-center evaluation of artificial intelligent imaging and clinical models for predicting neoadjuvant chemotherapy response in breast cancer.

BACKGROUND: Neoadjuvant chemotherapy (NAC) plays an important role in the management of locally advanced breast cancer. It allows for downstaging of tumors, potentially allowing for breast conservation. NAC also allows for in-vivo testing of the tumors' response to chemotherapy and provides important prognostic information. There are currently no clearly defined clinical models that incorporate imaging with clinical data to predict response to NAC. Thus, the aim of this work is to develop a predictive AI model based on routine CT imaging and clinical parameters to predict response to NAC. METHODS: The CT scans of 324 patients with NAC from multiple centers in Singapore were used in this study. Four different radiomics models were built for predicting pathological complete response (pCR): first two were based on textural features extracted from peri-tumoral and tumoral regions, the third model based on novel space-resolved radiomics which extract feature maps using voxel-based radiomics and the fourth model based on deep learning (DL). Clinical parameters were included to build a final prognostic model. RESULTS: The best performing models were based on space-resolved and DL approaches. Space-resolved radiomics improves the clinical AUCs of pCR prediction from 0.743 (0.650 to 0.831) to 0.775 (0.685 to 0.860) and our DL model improved it from 0.743 (0.650 to 0.831) to 0.772 (0.685 to 0.853). The tumoral radiomics model performs the worst with no improvement of the AUC from the clinical model. The peri-tumoral combined model gives moderate performance with an AUC of 0.765 (0.671 to 0.855). CONCLUSIONS: Radiomics features extracted from diagnostic CT augment the predictive ability of pCR when combined with clinical features. The novel space-resolved radiomics and DL radiomics approaches outperformed conventional radiomics techniques.

Predictors for late genitourinary toxicity in men receiving radiotherapy for high-risk prostate cancer using planned and accumulated dose.

Background and purpose: Significant deviations between bladder dose planned (DP) and dose accumulated (DA) have been reported in patients receiving radiotherapy for prostate cancer. This study aimed to construct multivariate analysis (MVA) models to predict the risk of late genitourinary (GU) toxicity with clinical and DP or DA as dose-volume (DV) variables. Materials and methods: Bladder DA obtained from 150 patients were compared with DP. MVA models were built from significant clinical and DV variables (p < 0.05) at univariate analysis. Previously developed dose-based-region-of-interest (DB-ROI) metrics using expanded ring structures from the prostate were included. Goodness-of-fit test and calibration plots were generated to determine model performance. Internal validation was accomplished using Bootstrapping. Results: Intermediate-high DA (V30-65 Gy and DB-ROI-20-50 mm) for bladder increased compared to DP. However, at the very high dose region, DA (D0.003 cc, V75 Gy, and DB-ROI-5-10 mm) were significantly lower. In MVA, single variable models were generated with odds ratio (OR) < 1. DB-ROI-50 mm was predictive of Grade ≥ 1 GU toxicity for DA and DP (DA and DP; OR: 0.96, p: 0.04) and achieved an area under the receiver operating curve (AUC) of > 0.6. Prostate volume (OR: 0.87, p: 0.01) was significant in predicting Grade 2 GU toxicity with a high AUC of 0.81. Conclusions: Higher DA (V30-65 Gy) received by the bladder were not translated to higher late GU toxicity. DB-ROIs demonstrated higher predictive power than standard DV metrics in associating Grade ≥ 1 toxicity. Smaller prostate volumes have a minor protective effect on late Grade 2 GU toxicity.

Dependence of LET on material and its impact on current RBE model.

Biological uncertainty remains one of the main sources of uncertainties in proton therapy, and is encapsulated in a scalar quantity known as relative biological effective (RBE). It is currently recognised that a constant RBE of 1.1 is not consistent with radiobiological experiment and may lead to sub-optimal exploitation of the benefits of proton therapy. To overcome this problem, several RBE models have been developed, and in most of these models, there is a dependence of RBE on dose-averaged linear energy transfer (LET), [Formula: see text]. In this work, we show that the [Formula: see text] estimation in these models during the data-fitting (or parameter estimation) phase could be subjected to a huge uncertainty due to not taking into account cellular materials during simulation, and this uncertainty can propagate down to the resulting RBE models. The dosimetric impact of this [Formula: see text] uncertainty is then evaluated on a simple clinical spread out Bragg peak (SOBP) and a prostate example. Our simulation shows that [Formula: see text] uncertainty due to the use of water as cellular material is non-negligible under low [Formula: see text] and low dose (2 Gy), and can be neglected otherwise. Thus, this study indicates that further dose and range margins may be required for low [Formula: see text] target under low dose. This is due to greater uncertainties in RBE model associated with incomplete knowledge of cellular composition for [Formula: see text] computation.