Landet: an efficient physics-informed deep learning approach for automatic detection of anatomical landmarks and measurement of spinopelvic alignment.

Abstract

An efficient physics-informed deep learning approach for extracting spinopelvic measures from X-ray images is introduced and its performance is evaluated against manual annotations. Two datasets, comprising a total of 1470 images, were collected to evaluate the model's performance. We propose a novel method of detecting landmarks as objects, incorporating their relationships as constraints (LanDet). Using this approach, we trained our deep learning model to extract five spine and pelvis measures: Sacrum Slope (SS), Pelvic Tilt (PT), Pelvic Incidence (PI), Lumbar Lordosis (LL), and Sagittal Vertical Axis (SVA). The results were compared to manually labelled test dataset (GT) as well as measures annotated separately by three surgeons. The LanDet model was evaluated on the two datasets separately and on an extended dataset combining both. The final accuracy for each measure is reported in terms of Mean Absolute Error (MAE), Standard Deviation (SD), and R Pearson correlation coefficient as follows: <math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mo>[</mo> <mi>S</mi> <msup><mi>S</mi> <mo>∘</mo></msup> <mo>:</mo> <mn>3.7</mn> <mrow><mo>(</mo> <mn>2.7</mn> <mo>)</mo></mrow> <mo>,</mo> <mi>R</mi> <mo>=</mo> <mn>0.89</mn> <mo>]</mo></mrow> </math> , <math xmlns="http://www.w3.org/1998/Math/MathML"> <mrow><mrow><mo>[</mo> <mi>P</mi> <msup><mi>T</mi> <mo>∘</mo></msup> <mo>:</mo> <mn>1.3</mn> <mrow><mo>(</mo> <mn>1.1</mn> <mo>)</mo></mrow> <mo>,</mo> <mi>R</mi> <mo>=</mo> <mn>0.98</mn> <mo>]</mo></mrow> <mo>,</mo> <mrow><mo>[</mo> <mi>P</mi> <msup><mi>I</mi> <mo>∘</mo></msup> <mo>:</mo> <mn>4.2</mn> <mrow><mo>(</mo> <mn>3.1</mn> <mo>)</mo></mrow> <mo>,</mo> <mi>R</mi> <mo>=</mo> <mn>0.93</mn> <mo>]</mo></mrow> <mo>,</mo> <mrow><mo>[</mo> <mi>L</mi> <msup><mi>L</mi> <mo>∘</mo></msup> <mo>:</mo> <mn>5.1</mn> <mrow><mo>(</mo> <mn>6.4</mn> <mo>)</mo></mrow> <mo>,</mo> <mi>R</mi> <mo>=</mo> <mn>0.83</mn> <mo>]</mo></mrow> <mo>,</mo> <mrow><mo>[</mo> <mi>S</mi> <mi>V</mi> <mi>A</mi> <mrow><mo>(</mo> <mi>m</mi> <mi>m</mi> <mo>)</mo></mrow> <mo>:</mo> <mn>2.1</mn> <mrow><mo>(</mo> <mn>1.9</mn> <mo>)</mo></mrow> <mo>,</mo> <mi>R</mi> <mo>=</mo> <mn>0.96</mn> <mo>]</mo></mrow> </mrow> </math> . To assess model reliability and compare it against surgeons, the intraclass correlation coefficient (ICC) metric is used. The model demonstrated better consistency with surgeons with all values over 0.88 compared to what was previously reported in the literature. The LanDet model exhibits competitive performance compared to existing literature. The effectiveness of the physics-informed constraint method, utilized in our landmark detection as object algorithm, is highlighted. Furthermore, we addressed the limitations of heatmap-based methods for anatomical landmark detection and tackled issues related to mis-identifying of similar or adjacent landmarks instead of intended landmark using this novel approach.