Zeitschrift fur Medizinische Physik最新文献

The use of artificial intelligence systems in clinical routine is still hampered by the necessity of a medical device certification and/or by the difficulty of implementing these systems in a clinic’s quality management system. In this context, the key questions for a user are how to ensure robust model predictions and how to appraise the quality of a model’s results on a regular basis.

In this paper we discuss some conceptual foundation for a clinical implementation of a machine learning system and argue that both vendors and users should take certain responsibilities, as is already common practice for high-risk medical equipment.

We propose the methodology from AAPM Task Group 100 report No. 283 as a conceptual framework for developing risk-driven a quality management program for a clinical process that encompasses a machine learning system. This is illustrated with an example of a clinical workflow. Our analysis shows how the risk evaluation in this framework can accommodate artificial intelligence based systems independently of their robustness evaluation or the user’s in–house expertise. In particular, we highlight how the degree of interpretability of a machine learning system can be systematically accounted for within the risk evaluation and in the development of a quality management system.

For dosimetry of radiopharmaceutical therapies, it is essential to determine the volume of relevant structures exposed to therapeutic radiation. For many radiopharmaceuticals, the kidneys represent an important organ-at-risk. To reduce the time required for kidney segmentation, which is often still performed manually, numerous approaches have been presented in recent years to apply deep learning-based methods for CT-based automated segmentation. While the automatic segmentation methods presented so far have been based solely on CT information, the aim of this work is to examine the added value of incorporating PSMA-PET data in the automatic kidney segmentation.

Methods

A total of 108 PET/CT examinations (53 [⁶⁸Ga]Ga-PSMA-I&T and 55 [¹⁸F]F-PSMA-1007 examinations) were grouped to create a reference data set of manual segmentations of the kidney. These segmentations were performed by a human examiner. For each subject, two segmentations were carried out: one CT-based (detailed) segmentation and one PET-based (coarser) segmentation. Five different u-net based approaches were applied to the data set to perform an automated segmentation of the kidney: CT images only, PET images only (coarse segmentation), a combination of CT and PET images, a combination of CT images and a PET-based coarse mask, and a CT image, which had been pre-segmented using a PET-based coarse mask. A quantitative assessment of these approaches was performed based on a test data set of 20 patients, including Dice score, volume deviation and average Hausdorff distance between automated and manual segmentations. Additionally, a visual evaluation of automated segmentations for 100 additional (i.e., exclusively automatically segmented) patients was performed by a nuclear physician.

Results

Out of all approaches, the best results were achieved by using CT images which had been pre-segmented using a PET-based coarse mask as input. In addition, this method performed significantly better than the segmentation based solely on CT, which was supported by the visual examination of the additional segmentations. In 80% of the cases, the segmentations created by exploiting the PET-based pre-segmentation were preferred by the nuclear physician.

Conclusion

This study shows that deep-learning based kidney segmentation can be significantly improved through the addition of a PET-based pre-segmentation. The presented method was shown to be especially beneficial for kidneys with cysts or kidneys that are closely adjacent to other organs such as the spleen, liver or pancreas. In the future, this could lead to a considerable reduction in the time required for dosimetry calculations as well as an improvement in the results.

Purpose

Whole-body bone scintigraphy (WBS) is one of the most widely used modalities in diagnosing malignant bone diseases during the early stages. However, the procedure is time-consuming and requires vigour and experience. Moreover, interpretation of WBS scans in the early stages of the disorders might be challenging because the patterns often reflect normal appearance that is prone to subjective interpretation. To simplify the gruelling, subjective, and prone-to-error task of interpreting WBS scans, we developed deep learning (DL) models to automate two major analyses, namely (i) classification of scans into normal and abnormal and (ii) discrimination between malignant and non-neoplastic bone diseases, and compared their performance with human observers.

Materials and Methods

After applying our exclusion criteria on 7188 patients from three different centers, 3772 and 2248 patients were enrolled for the first and second analyses, respectively. Data were split into two parts, including training and testing, while a fraction of training data were considered for validation. Ten different CNN models were applied to single- and dual-view input (posterior and anterior views) modes to find the optimal model for each analysis. In addition, three different methods, including squeeze-and-excitation (SE), spatial pyramid pooling (SPP), and attention-augmented (AA), were used to aggregate the features for dual-view input models. Model performance was reported through area under the receiver operating characteristic (ROC) curve (AUC), accuracy, sensitivity, and specificity and was compared with the DeLong test applied to ROC curves. The test dataset was evaluated by three nuclear medicine physicians (NMPs) with different levels of experience to compare the performance of AI and human observers.

Results

DenseNet121_AA (DensNet121, with dual-view input aggregated by AA) and InceptionResNetV2_SPP achieved the highest performance (AUC = 0.72) for the first and second analyses, respectively. Moreover, on average, in the first analysis, Inception V3 and InceptionResNetV2 CNN models and dual-view input with AA aggregating method had superior performance. In addition, in the second analysis, DenseNet121 and InceptionResNetV2 as CNN methods and dual-view input with AA aggregating method achieved the best results. Conversely, the performance of AI models was significantly higher than human observers for the first analysis, whereas their performance was comparable in the second analysis, although the AI model assessed the scans in a drastically lower time.

Conclusion

Using the models designed in this study, a positive step can be taken toward improving and optimizing WBS interpretation. By training DL models with larger and more diverse cohorts, AI could potentially be used to assist physicians in the assessment of WBS images.

The use of synthetic CT (sCT) in the radiotherapy workflow would reduce costs and scan time while removing the uncertainties around working with both MR and CT modalities. The performance of deep learning (DL) solutions for sCT generation is steadily increasing, however most proposed methods were trained and validated on private datasets of a single contrast from a single scanner. Such solutions might not perform equally well on other datasets, limiting their general usability and therefore value. Additionally, functional evaluations of sCTs such as dosimetric comparisons with CT-based dose calculations better show the impact of the methods, but the evaluations are more labor intensive than pixel-wise metrics.

To improve the generalization of an sCT model, we propose to incorporate a pre-trained DL model to pre-process the input MR images by generating artificial proton density, $T1$ and $T2$ maps (i.e. contrast-independent quantitative maps), which are then used for sCT generation. Using a dataset of only $T2w$ MR images, the robustness towards input MR contrasts of this approach is compared to a model that was trained using the MR images directly. We evaluate the generated sCTs using pixel-wise metrics and calculating mean radiological depths, as an approximation of the mean delivered dose.

On $T2w$ images acquired with the same settings as the training dataset, there was no significant difference between the performance of the models. However, when evaluated on $T1w$ images, and a wide range of other contrasts and scanners from both public and private datasets, our approach outperforms the baseline model.

Using a dataset of $T2w$ MR images, our proposed model implements synthetic quantitative maps to generate sCT images, improving the generalization towards other contrasts. Our code and trained models are publicly available.

The impact of including model-averaged excess radiation risks (ER) into a measure of radiation attributed decrease of survival (RADS) for the outcome all solid cancer incidence and the impact on the uncertainties is demonstrated. It is shown that RADS applying weighted model averaged ER based on AIC weights result in smaller risk estimates with narrower 95% CI than RADS using ER based on BIC weights. Further a multi-method-multi-model inference approach is introduced that allows calculating one general RADS estimate providing a weighted average risk estimate for a lunar and a Mars mission. For males the general RADS estimate is found to be 0.42% (95% CI: 0.38%; 0.45%) and for females 0.67% (95% CI: 0.59%; 0.75%) for a lunar mission and 2.45% (95% CI: 2.23%; 2.67%) for males and 3.91% (95% CI: 3.44%; 4.39%) for females for a Mars mission considering an age at exposure of 40 years and an attained age of 65 years. It is recommended to include these types of uncertainties and to include model-averaged excess risks in astronaut risk assessment.