n Australia, prostate cancer is, after skin cancer, the most common cancer diagnosed, and it is second only to lung cancer as a cause of cancer-related deaths. [1][2][3][4] In 2005, prostate cancer was the most prevalent cancer for Australian men, accounting for over 29% of all cancer diagnoses. 2 A major clinical treatment for prostate cancer is external-beam radiation therapy (EBRT). Computed tomography (CT) scans are used to provide the required electron density information for radiation therapy dose planning. However, magnetic resonance imaging (MRI) gives superior soft-tissue contrast for visualising the prostate and determining target volume. Here, we describe the development of an efficient, alternative planning method using MRI only for both organ delineation and dose calculation; "pseudo-CT scans" created from the MRI scans are used for dose planning. Standard treatment planning for EBRTIn EBRT for prostate cancer, high-energy x-ray beams from multiple directions deposit energy (dose) within the tumour (prostate) to destroy cancer cells. The standard process for EBRT treatment planning is shown in Box 1(A). The first step is patient imaging (a CT scan sometimes combined with MRI). The treatment targets (the prostate and sometimes the seminal vesicles) and important normal tissues (the rectum, bladder, and femoral heads) are then manually defined from the scans (Box 2). Modern radiotherapy machines offer improved treatment accuracy through better visualisation and the correction of errors in patient setup, making target delineation the most significant uncertainty in radiotherapy planning.In standard treatment planning, if MRI is used for visualising the prostate, then the scans from MRI and CT are aligned to transfer the structure contours defined by MRI to the CT scans for accurate dose calculation. The defined prostate volume is then expanded to become the larger, planning target volume for treatment. This allows for uncertainties in delineation and patient setup, and for prostate movement. The next step is the use of computer planning tools to determine the directions, strengths and shapes of the treatment beams used to deliver the prescribed dose to the defined target, while minimising the dose to normal tissues. Finally, the patient is carefully positioned and the treatment is delivered. The optimal way to align the patient for treatment is to use small implanted fiducial markers in the prostate. These are visible under x-ray imaging and show the precise position of the prostate within the body. Image guidance is used to align the treatment target each day for the entire radiotherapy course.CT v MRI for radiation therapy planning CT scans are acquired using low-energy x-ray beams. An image is created, with each pixel value assigned a CT number. This number can be easily related by a calibration process to the density of electrons within the tissues. The CT scan converted to electron densities can then be used to calculate the dose to be delivered to a patient from a radiotherapy x-ray beam. Energy i...

Our automated deep learning-based approach identifies consolidation/collapse in LUS images to aid in the diagnosisof late stages of COVID-19 induced pneumonia, where consolidation/collapse is one of the possible associated pathologies. A common challenge in training such models is that annotating each frame of an ultrasound video requires high labelling effort. This effort in practice becomes prohibitive for large ultrasound datasets. To understand the impact of various degrees of labelling precision, we compare labelling strategies to train fully supervised models (frame-based method, higher labelling effort) and inaccurately supervised models (video-based methods, lower labelling effort), both of which yield binary predictions for LUS videos ona frame-by-frame level. We moreover introduce a novel sampled quaternary method which randomly samples only 10% of the LUS video frames and subsequently assigns (ordinal) categorical labels to all frames in the video based on the fraction of positively annotated samples. This method outperformed the inaccurately supervised video-based method of our previous work on pleural effusions. More surprisingly, this method outperformed the supervised frame-based approach with respect to metrics such as precision-recall area under curve (PR-AUC) and F1 score that are suitable for the class imbalance scenario of our dataset despite being a form of inaccurate learning. This may be due to the combination of a significantly smaller data set size compared to ourprevious work and the higher complexity of consolidation/collapse compared to pleural effusion, two factors which contribute to label noise and overfitting; specifically, we argue that our video-based method is more robust with respect to label noise and mitigates overfitting in a manner similar to label smoothing. Using clinical expert feedback, separate criteria were developed to exclude data from the training and test sets respectively for our ten-fold cross validation results, which resulted in a PR-AUC score of 73% and an accuracy of 89%. While the efficacy of our classifier using the sampled quaternary method must be verified on a larger consolidation/collapse dataset, when considering the complexity of the pathology, our proposed classifier using the sampled quaternary video-based method is clinically comparable with trained expertsand improves over the video-based method of our previous work on pleural effusions.

No abstract