Computer-aided diagnosis tools for Retinopathy of Prematurity (ROP) base their decisions on handcrafted retinal features that highly correlate with expert diagnoses, such as arterial and venous curvature, tortuosity and dilation. Deep learning leads to performance comparable to those of expert physicians, albeit not ensuring that the same clinical factors are learned in the deep representations. In this paper, we investigate the relationship between the handcrafted and the deep learning features in the context of ROP diagnosis. Average statistics on the handcrafted features for each input image were expressed as retinal concept measures. Three disease severity grades, i.e. normal, pre-plus and plus, were classified by a deep convolutional neural network. Regression Concept Vectors (RCV) were computed in the network feature space for each retinal concept measure. Relevant concept measures were identified by bidirectional relevance scores for the normal and plus classes. Results show that the curvature, diameter and tortuosity of the segmented vessels are indeed relevant to the classification. Among the potential applications of this method, the analysis of borderline cases between the classes and of network faults, in particular, can be used to improve the performance.
Mara Graziani, James M. Brown, Vincent Andrearczyk, Veysi Yildiz, J. Peter Campbell, Deniz Erdogmus, Stratis Ioannidis, Michael F. Chiang, Jayashree Kalpathy-Cramer, and Henning Müller, "Improved interpretability for computer-aided severity assessment of retinopathy of prematurity," Proc. SPIE 10950, Medical Imaging 2019: Computer-Aided Diagnosis, 109501R (Presented at SPIE Medical Imaging: February 20, 2019; Published: 13 March 2019); https://doi.org/10.1117/12.2512584.
Conference Presentations are recordings of oral presentations given at SPIE conferences and published as part of the proceedings. They include the speaker's narration with video of the slides and animations. Most include full-text papers. Interactive, searchable transcripts and closed captioning are now available for most presentations.
Search our growing collection of more than 18,000 conference presentations, including many plenaries and keynotes.