Analysis of neural networks for routine classification of sixteen ultrasound upper abdominal cross sections.

PURPOSE: Abdominal ultrasound screening requires the capture of multiple standardized plane views as per clinical guidelines. Currently, the extent of adherence to such guidelines is dependent entirely on the skills of the sonographer. The use of neural network classification has the potential to better standardize captured plane views and streamline plane capture reducing the time burden on operators by combatting operator variability. METHODS: A dataset consisting of 16 routine upper abdominal ultrasound scans from 64 patients was used to test the classification accuracy of 9 neural networks. These networks were tested on both a small, idealised subset of 800 samples as well as full video sweeps of the region of interest using stratified sampling and transfer learning. RESULTS: The highest validation accuracy attained by both GoogLeNet and InceptionV3 is 83.9% using transfer learning and the large sample set of 26,294 images. A top-2 accuracy of 95.1% was achieved using InceptionV3. Alexnet attained the highest accuracy of 79.5% (top-2 of 91.5%) for the smaller sample set of 800 images. The neural networks evaluated during this study were also successfully able to identify problematic individual cross sections such as between kidneys, with right and left kidney being accurately identified 78.6% and 89.7%, respectively. CONCLUSION: Dataset size proved a more important factor in determining accuracy than network selection with more complex neural networks providing higher accuracy as dataset size increases and simpler linear neural networks providing better results where the dataset is small.

Phantom Study of Arterial Localization using Tactile Sensor Array and a Normal Vs. Shear Pulse Pressure Propagation Method.

OBJECTIVE: Locating the radial artery reliably is a key challenge in reducing patient risks from complications in Trans-Radial Access, which is an important clinical method for catheterization, cardiac monitoring, and neuroendovascular procedures. New tactile sensing technology is being developed to bridge the skill, cost, and performance gap between ultrasonic needle guidance, and manual palpation, for use in developing countries. This paper further develops tactile artery localization with a novel algorithm for arterial localization based on the properties of a curved tactile sensor array. METHODS: Using tactile sensor insensitivity to shear loading, coupled with a radial pulse wave propagation path, the position of the artery can be found at the intersection of a normal and tangential vector from the array corresponding to maximum and minimum pulse pressure measurement locations respectively. This was validated in a simple silicone phantom study Results: The proposed method measured with MAE= 0.58±0.25mm whilst the artery is within range of the tactile array, compared with 0.81±0.57mm for a comparative method of simple pulse localization. This showed improvement in arterial localization and repeatability, and was within 1 arterial radius, expected to reduce the risk of missing the artery, or perforating the side wall.Clinical Relevance- Robust and repeatable arterial localization is important for reducing the failure rate of trans-radial (and other arterial) procedures, and thus reducing the risk of harmful complications.

Prescriptive Method for Optimizing Cost of Data Collection and Annotation in Machine Learning of Clinical Ultrasound.

Machine learning in medical ultrasound faces a major challenge: the prohibitive costs of producing and annotating clinical data. Optimizing the data collection and annotation will improve model training efficiency, reducing project cost and times. This paper prescribes a 2-phase method for cost optimization based on iterative accuracy/sample size predictions, and active learning for annotation optimization. METHODS: Using public breast, fetal, and lung ultrasound datasets we can: Optimize data collection by statistically predicting accuracy for a desired dataset size; and optimize labeling efficiency using Active Learning, where predictions with lowest certainty were labelled manually using feedback. A practical case study on BUSI data was used to demonstrate the method prescribed in this work. RESULTS: With small data subsets, ~10%, dataset size vs. final accuracy relations can be predicted with diminishing results after 50% usage. Manual annotation was reduced by ~10% using active learning to focus the annotation. CONCLUSION: This led to cost reductions of 50%-66%, depending on requirements and initial cost model, on BUSI dataset with a negligible accuracy drop of 3.75% from theoretical maximums.Clinical Relevance- This work provides methodology to optimize dataset size and manual data labelling, this allows generation of cost-effective datasets, of interest to all, but particularly for financially limited trials and feasibility studies, Reducing the time burden on annotating clinicians.

Design and Manufacture of an Optimised Side-Shifted PPM EMAT Array for Use in Mobile Robotic Localisation.

Guided wave Electro Magnetic Acoustic Transducers (EMATs) offer an elegant method for structural inspection and localisation relative to geometric features, such as welds. This paper presents a Lorentz force EMAT construction framework, where a numerical model has been developed for optimising Printed Circuit Board (PCB) coil parameters as well as a methodology for optimising magnet array parameters to a user's needs. This framework was validated experimentally to show its effectiveness through comparison to an industry built EMAT. The framework was then used to design and manufacture a Side-Shifted Unidirectional Periodic Permanent Magnet (PPM) EMAT for use on a mobile robotic system, which uses guided waves for ranging to build internal maps of a given subject, identifying welded sections, defects and other structural elements. The unidirectional transducer setup was shown to operate in simulation and was then manufactured to compare to the bidirectional transmitter and two-receiver configurations on a localisation system. The unidirectional setup was shown to have clear benefits over the bidirectional setup for mapping an unknown environment using guided waves as there were no dead spots of mapping where signal direction could not be interpreted. Additionally, overall package size was significantly reduced, which in turn allows more measurements to be taken within confined spaces and increases robotic crawler mobility.

Non-Invasive Radial Artery Blood Pressure Monitoring Using Error Compensated Tactile Sensors and Patient Specific Oscillometry.

This paper presents a new method of measuring non-invasive blood pressure at the radial artery based on oscillometry and tonometry. A localized capacitive tactile sensor array is used with a novel algorithm based on waveform features for optimizing oscillometry ratios. A novel tonometer is presented with typically 1% base measurement error, with sensor errors compensated using a custom error model, and applied to blood pressure measurement at the radial artery. The tonometer gives a direct arterial waveform, and uses a manual pressure sweep to determine blood pressure. Key points on the oscillogram are correlated with optimal ratios for minimizing mean errors and standard deviation for an individual. This paper details an initial assessment into the dominant sources of error, for the purpose of determining feasibility and directing future research. Over a limited clinical trial of Np = 20, No = 180, the reported BP accuracy is MAE = 0.61/0.38mmHg and 1SD = 7.14/5.91mmHg for systolic and diastolic measurements respectively. The average load on the patient is in the order of 5N, compared with around 1000N for a brachial cuff, which represents a clear improvement in patient comfort. This is a positive result, indicating larger scale performance within AAMI and BHS standards, and stands as a useful benchmark for further development of the system into a clinical product for rapid and comfortable BP measurement. Clinical Relevance This paper demonstrated that direct tonometry can measure blood pressure if sensor error is compensated by the designer. This method uses 200x less load than conventional cuffs suitable for long term and supine use.

Breast Cancer: Model Reconstruction and Image Registration From Segmented Deformed Image Using Visual and Force Based Analysis.

Breast lesion localization using tactile imaging is a new and developing direction in medical science. To achieve the goal, proper image reconstruction and image registration can be a valuable asset. In this paper, a new approach of the segmentation-based image surface reconstruction algorithm is used to reconstruct the surface of a breast phantom. In breast tissue, the sub-dermal vein network is used as a distinguishable pattern for reconstruction. The proposed image capturing device contacts the surface of the phantom, and surface deformation will occur due to applied force at the time of scanning. A novel force based surface rectification system is used to reconstruct a deformed surface image to its original structure. For the construction of the full surface from rectified images, advanced affine scale-invariant feature transform (A-SIFT) is proposed to reduce the affine effect in time when data capturing. Camera position based image stitching approach is applied to construct the final original non-rigid surface. The proposed model is validated in theoretical models and real scenarios, to demonstrate its advantages with respect to competing methods. The result of the proposed method, applied to path reconstruction, ends with a positioning accuracy of 99.7%.