ABSTRACT
We present a structured approach to combine explainability of artificial intelligence (AI) with the scientific method for scientific discovery. We demonstrate the utility of this approach in a proof-of-concept study where we uncover biomarkers from a convolutional neural network (CNN) model trained to classify patient sex in retinal images. This is a trait that is not currently recognized by diagnosticians in retinal images, yet, one successfully classified by CNNs. Our methodology consists of four phases: In Phase 1, CNN development, we train a visual geometry group (VGG) model to recognize patient sex in retinal images. In Phase 2, Inspiration, we review visualizations obtained from post hoc interpretability tools to make observations, and articulate exploratory hypotheses. Here, we listed 14 hypotheses retinal sex differences. In Phase 3, Exploration, we test all exploratory hypotheses on an independent dataset. Out of 14 exploratory hypotheses, nine revealed significant differences. In Phase 4, Verification, we re-tested the nine flagged hypotheses on a new dataset. Five were verified, revealing (i) significantly greater length, (ii) more nodes, and (iii) more branches of retinal vasculature, (iv) greater retinal area covered by the vessels in the superior temporal quadrant, and (v) darker peripapillary region in male eyes. Finally, we trained a group of ophthalmologists (N=26) to recognize the novel retinal features for sex classification. While their pretraining performance was not different from chance level or the performance of a nonexpert group (N=31), after training, their performance increased significantly (p<0.001, d=2.63). These findings showcase the potential for retinal biomarker discovery through CNN applications, with the added utility of empowering medical practitioners with new diagnostic capabilities to enhance their clinical toolkit.
ABSTRACT
Deep learning (DL) techniques have seen tremendous interest in medical imaging, particularly in the use of convolutional neural networks (CNNs) for the development of automated diagnostic tools. The facility of its non-invasive acquisition makes retinal fundus imaging particularly amenable to such automated approaches. Recent work in the analysis of fundus images using CNNs relies on access to massive datasets for training and validation, composed of hundreds of thousands of images. However, data residency and data privacy restrictions stymie the applicability of this approach in medical settings where patient confidentiality is a mandate. Here, we showcase results for the performance of DL on small datasets to classify patient sex from fundus images-a trait thought not to be present or quantifiable in fundus images until recently. Specifically, we fine-tune a Resnet-152 model whose last layer has been modified to a fully-connected layer for binary classification. We carried out several experiments to assess performance in the small dataset context using one private (DOVS) and one public (ODIR) data source. Our models, developed using approximately 2500 fundus images, achieved test AUC scores of up to 0.72 (95% CI: [0.67, 0.77]). This corresponds to a mere 25% decrease in performance despite a nearly 1000-fold decrease in the dataset size compared to prior results in the literature. Our results show that binary classification, even with a hard task such as sex categorization from retinal fundus images, is possible with very small datasets. Our domain adaptation results show that models trained with one distribution of images may generalize well to an independent external source, as in the case of models trained on DOVS and tested on ODIR. Our results also show that eliminating poor quality images may hamper training of the CNN due to reducing the already small dataset size even further. Nevertheless, using high quality images may be an important factor as evidenced by superior generalizability of results in the domain adaptation experiments. Finally, our work shows that ensembling is an important tool in maximizing performance of deep CNNs in the context of small development datasets.
