A Deep Learning-Based System (Microscan) for the Identification of Pollen Development Stages and Its Application to Obtaining Doubled Haploid Lines in Eggplant

Deep learning-based high-throughput detection of in vitro germination to assess pollen viability from microscopic images

In vitro pollen germination is considered the most efficient method to assess pollen viability. The pollen germination frequency and pollen tube length, which are key indicators of pollen viability, should be accurately measured during in vitro culture. In this study, a Mask R-CNN model trained using microscopic images of tree peony (Paeonia suffruticosa) pollen has been proposed to rapidly detect the pollen germination rate and pollen tube length. To reduce the workload during image acquisition, images of synthesized crossed pollen tubes were added to the training dataset, significantly improving the model accuracy in recognizing crossed pollen tubes. At an Intersection over Union threshold of 50%, a mean average precision of 0.949 was achieved. The performance of the model was verified using 120 testing images. The R2 value of the linear regression model using detected pollen germination frequency against the ground truth was 0.909 and that using average pollen tube length was 0.958. Further, the model was successfully applied to two other plant species, indicating a good generalizability and potential to be applied widely.

Doubled Haploids in Eggplant

Eggplant is a solanaceous crop cultivated worldwide for its edible fruit. Eggplant breeding programs are mainly aimed to the generation of F1 hybrids by crossing two highly homozygous, pure lines, which are traditionally obtained upon several self crossing generations, which is an expensive and time consuming process. Alternatively, fully homozygous, doubled haploid (DH) individuals can be induced from haploid cells of the germ line in a single generation. Several attempts have been made to develop protocols to produce eggplant DHs principally using anther culture and isolated microspore culture. Eggplant could be considered a moderately recalcitrant species in terms of ability for DH production. Anther culture stands nowadays as the most valuable technology to obtain eggplant DHs. However, the theoretical possibility of having plants regenerated from somatic tissues of the anther walls cannot be ruled out. For this reason, the use of isolated microspores is recommended when possible. This approach still has room for improvement, but it is largely genotype-dependent. In this review, we compile the most relevant advances made in DH production in eggplant, their application to breeding programs, and the future perspectives for the development of other, less genotype-dependent, DH technologies.

Deep learning-based high-throughput phenotyping can drive future discoveries in plant reproductive biology

Advances in deep learning are providing a powerful set of image analysis tools that are readily accessible for high-throughput phenotyping applications in plant reproductive biology. High-throughput phenotyping systems are becoming critical for answering biological questions on a large scale. These systems have historically relied on traditional computer vision techniques. However, neural networks and specifically deep learning are rapidly becoming more powerful and easier to implement. Here, we examine how deep learning can drive phenotyping systems and be used to answer fundamental questions in reproductive biology. We describe previous applications of deep learning in the plant sciences, provide general recommendations for applying these methods to the study of plant reproduction, and present a case study in maize ear phenotyping. Finally, we highlight several examples where deep learning has enabled research that was previously out of reach and discuss the future outlook of these methods.