Effect of CT reconstruction settings on the performance of a deep learning based lung nodule CAD system.

PURPOSE: To study the effect of different reconstruction parameter settings on the performance of a commercially available deep learning based pulmonary nodule CAD system. MATERIALS AND METHODS: We performed a retrospective analysis of 24 chest CT scans, reconstructed at 16 different reconstruction settings for two different iterative reconstruction algorithms (SAFIRE and ADMIRE) varying in slice thickness, kernel size and iterative reconstruction level strength using a commercially available deep learning pulmonary nodule CAD system. The DL-CAD software was evaluated at 25 different sensitivity threshold settings and nodules detected by the DL-CAD software were matched against a reference standard based on the consensus reading of three radiologists. RESULTS: A total of 384 CT reconstructions was analysed from 24 patients, resulting in a total of 5786 found nodules. We matched the detected nodules against the reference standard, defined by a team of thoracic radiologists, and showed a gradual drop in recall, and an improvement in precision when the iterative strength levels were increased for a constant kernel size. The optimal DL-CAD threshold setting for use in our clinical workflow was found to be 0.88 with an F2 of 0.73 ±â€¯0.053. CONCLUSIONS: The DL-CAD system behaves differently on IR data than on FBP data, there is a gradual drop in recall, and growth in precision when the iterative strength levels are increased. As a result, caution should be taken when implementing deep learning software in a hospital with multiple CT scanners and different reconstruction protocols. To the best of our knowledge, this is the first study that demonstrates this result from a DL-CAD system on clinical data.

Automated quality control of ultrasound based on in-air reverberation patterns.

Ultrasound image degradation originates primarily from transducer defects and potentially undermines reliable image interpretation. Systematic quantitative quality control is often neglected due to the limited resources available for this task. We propose a quantitative quality control based on in-air reverberation images. These images serve as an initial indication of image degradation. They are easily generated for any (curvi-)linear transducer independent of the level of expertise of the operator. Automated analysis is presented to extract quality parameters based on the in-air reverberation pattern. Static images acquired by the clinical user are transferred to a server where analysis is performed. The results are available to the sonographer prior to clinical use and transducer status can be remotely monitored with trend analysis over time. The method was evaluated for normal functioning and defect transducers. A pilot study was performed over a period of three weeks to assess reproducibility and practical feasibility. All reverberation images were successfully analysed for different transducer types and vendor-specific image presentation. The proposed quality parameters are sensitive to signal loss and allow differentiation of type and severity of image degradation. The pilot study was well received by the sonographers for the simplicity of the method and the measurements were consistent over time. The proposed automated analysis method of ultrasound quality control can monitor (curvi-)linear transducer status in the entire hospital, overcoming previous limitations for periodic quality control. Implementation of the method can reduce the number of defective transducers routinely used in clinical practice.