Patient-specific modeling of the trochlear morphologic anomalies by means of hyperbolic paraboloids

Abstract

Diagnostic and therapeutic purposes are issuing pressing demands to improve the evaluation of the dysplasia condition of the femoral trochlea. The traditional clinical assessment of the dysplasia, based on Dejour classification, recognized 4 increasing (A, B, C, D) levels of severity. It has been extensively questioned in the literature that this classification methodology can be defective suggesting that quantitative measures can ensure more reliable criteria for the dysplasia severity assessment. This study reports on a novel technique to model the trochlear surface (TS), digitally reconstructed by 3D volumetric imaging, using three hyperbolic paraboloids (HP), one to describe the global trochlear aspect, two to represent the local aspects of the medial and lateral compartments, respectively. Results on a cohort of 43 patients, affected by aspecific anterior knee pain, demonstrate the consistency of the estimated model parameters with the morphologic aspect of the TS. The obtained small fitting error (on average lower than 0.80 mm) demonstrated that the ventral aspect of the trochlear morphology can be modeled with high accuracy by HPs. We also showed that HP modeling provides a continuous representation of morphologic variations in shape parameter space while we found that similar morphologic anomalies of the trochlear aspect are actually attributed to different severity grades in the Dejour classification. This finding is in agreement with recent works in the literature reporting that morphometric parameters can only optimistically be used to discriminate between the Grade A and the remaining three grades. In conclusion, we can assert that the proposed methodology is a further step toward modeling of anatomical surfaces that can be used to quantify deviations to normality on a patient-specific basis.

Stacked sparse autoencoder networks and statistical shape models for automatic staging of distal femur trochlear dysplasia

“…In an attempt to accurately describe the morphologic condition of the trochlear region, quantitative methods were proposed to identify morphometric indexes of the femoral geometry, such as trochlear depth, sulcus angle, lateral trochlear inclination, and trochlear facet asymmetry ratio. These indexes are computed by exploiting anatomical landmarks measured on computed tomography (CT)/magnetic resonance imaging (MRI) scans and/or three‐dimensional (3D) reconstructed shapes of the femur . However, two main aspects make critical the application of such indexes to dysplastic staging.…”

“…First, the reliability of the thresholds to be defined in order to discern the severity of the pathology is contingent to the anthropometric variability (size, age, race, and gender) . Second, as the degeneration progresses, the trochlear morphology detaches from the physiological shape making anatomic landmarks difficult to measure or even meaningless . To overcome this issue, surface‐based techniques were recently explored in place of landmark‐based methods, due to their ability to take the full geometric complexity of trochlear region into account.…”

“…These indexes are computed by exploiting anatomical landmarks measured on computed tomography (CT)/magnetic resonance imaging (MRI) scans [14][15][16][17] and/or three-dimensional (3D) reconstructed shapes of the femur. [18][19][20][21][22] However, two main aspects make critical the application of such indexes to dysplastic staging. First, the reliability of the thresholds to be defined in order to discern the severity of the pathology is contingent to the anthropometric variability (size, age, race, and gender).…”

“…23,24 Second, as the degeneration progresses, the trochlear morphology detaches from the physiological shape making anatomic landmarks difficult to measure or even meaningless. 7,22 To overcome this issue, surface-based techniques were recently explored in place of landmark-based methods, due to their ability to take the full geometric complexity of trochlear region into account. Among different approaches, statistical shape models (SSM) have emerged as promising tools to the task for their ability to represent continuous morphologic variations of anatomical shapes, both globally and locally, including pathologic abnormalities.…”

“…The five considered statistical shape models.Exactly 22,22,22,44, and 66 femur shapes were used to build aSSM, bSSM, cSSM, dSSM, and eSSM, respectively. Out of the overall amount of shapes, 5, 5, and 5 healthy, mild dysplastic, and severe dysplastic shapes, respectively, were not used to build the SSM and considered for SSPA testing input pattern of size a in the network output by means of the intermediate encoding step performed by the hidden neurons, usually featuring logsig activation.…”