Deep learning-based quality-controlled spleen assessment from ultrasound images. (July 2022)

• Record Type: Journal Article
• Authors: Yuan, Zhen; Puyol-Antón, Esther; Jogeesvaran, Haran; Smith, Nicola; Inusa, Baba; King, Andrew P.
• Is Part Of: Biomedical signal processing and control. Volume 76(2022)
• Publication Date: 2022-07
• Subjects: Deep learning -- Spleen ultrasound image -- Segmentation -- Splenomegaly -- Sickle cell disease; SCD Sickle cell disease -- QC Quality control -- CT Computed tomography -- MRI Magnetic resonance image -- CNN Convolutional neural network -- ROC Receiver operating characteristic -- CCA Connected components analysis -- LDP Length measurement based on the longest distance between pairs of points on the segmentation boundary -- PCA Length measurement based on principal components analysis -- Post VGG-19 Length measurement using VGG-19 to directly estimate the length from the segmentation -- SB Segmentation-based model: modified U-Net with post-processing on predicted segmentation -- UC Regression-based model: U-Net contracting path direct regression without transferring weights -- UCW Regression-based model: U-Net contracting path direct regression with transferring weights -- VGG Regression-based model: VGG-19 -- QC CV Cross-validation of the quality control system -- HE Human error -- MPLE Mean percentage length error -- R Pearson's correlation coefficient -- IoU Intersection over union -- HD Hausdorff distance -- SEN Sensitivity -- SPE Specificity -- AUC Area under the receiver operating characteristic curve; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.103724
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: First automated pipeline for spleen assessment from the ultrasound images. First deep learning-based segmentation model for spleen ultrasound images. Two novel deep learning-based frameworks were developed to automatically measure spleen length. Our model achieved human expert level on the task of spleen length measurement. Abstract: Objective: Splenomegaly (abnormal splenic enlargement) is a potentially life-threatening condition that occurs in a range of clinical scenarios, including in patients suffering from Sickle cell disease (SCD). Therefore, spleen size assessments from ultrasound imaging are commonly performed in SCD clinics, and typically involve measuring the length of the spleen. However, the current workflow is prone to intra- and inter-observer variability and is dependent on the experience of the sonographer. Our objective was to automate the spleen length measurement process. Methods: Two deep learning-based approaches were investigated to achieve automated spleen length measurement from ultrasound images. One is a segmentation-based approach, where we trained a modified U-Net to obtain a spleen segmentation and then applied post-processing to measure the spleen length from the segmentation. The second approach is based on direct regression of spleen length. We also incorporated a quality control (QC) model to help less experienced sonographers ensure the quality of ultrasound images before measurement. Results: Our best model (segmentation-based approach) reached a mean percentage length error (MPLE) of 4.58% on good quality images, which is within the range of human expert inter-observer variability (5.78%). After including bad quality images, the incorporation of the QC step resulted in a significant reduction in MPLE (from 5.76% to 4.88%). Conclusion: Automated, quality-controlled spleen length measurement from ultrasound has been achieved with human-level accuracy. Significance: This proposed framework has the potential to assist in making robust and accurate assessments of the spleen, especially in settings where there is a lack of experienced sonographers.