Systems and methods for assessing the severity of neutrophilic skin diseases having visible cutaneous manifestations - Patents.com

Abstract

The present invention relates to a computer-implemented system (100) and method for detecting and assessing the severity of a neutrophilic skin disease (ND) condition associated with visible skin manifestations in a patient (10) before or after treatment with an anti-interleukin-36 receptor (anti-IL-36R) antibody.

Description

[Technical field]

[0001] The present invention relates to a computer system and a computer-implemented method for assessing (predicting) the severity of a neutrophilic dermatopathy (ND) condition in a patient before or after treatment with an anti-interleukin-36 receptor (anti-IL-36R) antibody, more particularly to a computer-implemented system and a method for predicting the severity of generalized pustular psoriasis (GPP) or palmoplantar pustulosis (PPP) in a patient before or after treatment with spesolimab (BI655130). [Background technology]

[0002] Neutrophilic dermatoses form a heterogeneous group of inflammatory skin disorders that show distinctive clinical features but are united by the presence of a sterile, predominantly neutrophilic infiltrate on histopathology. The heterogeneity of the morphology of the skin lesions associated with these disorders makes diagnosis difficult. Furthermore, a thorough evaluation is required to exclude diseases that may mimic these disorders.

[0003] Currently, the diagnosis and severity assessment of neutrophilic skin diseases is performed by licensed dermatologists, who also prescribe treatment regimens based on their subjective diagnosis and assessment.

[0004] Unfortunately, manual diagnosis of neutrophilic skin diseases is often subjective and therefore inaccurate. More frequently, different dermatologists may classify ND conditions at different levels of severity, leading to inconsistent treatment of patients suffering from ND. For example, incorrect classification of the severity of an ND condition may result in undertreatment, which may result in a longer recovery time for the patient. Similarly, incorrect classification of the severity of an ND condition may result in overtreatment, which may result in the patient being treated with less medication than is necessary. Summary of the Invention

[0005] Therefore, there is a need for a rapid, automatic and non-subjective method for assessing the severity of a patient's ND condition before or after treatment that does not rely on a physician's manual diagnosis or evaluation and that assists the physician in diagnosis. As used herein, the term "patient" refers to a human or non-human animal diagnosed with a neutrophilic skin disease condition that develops one or more symptoms of a neutrophilic skin disease, such as, for example, neutrophil infiltration in affected skin tissue, pustules, and / or has visible skin symptoms including erythema, pustules, and / or scales.

[0006] The present invention addresses the above needs by providing a computer system and a computer-implemented method (each as set forth in the independent claims) for automatically assessing the severity of a patient's ND condition related to GPP or PPP before or after treatment with a therapeutically effective amount of an anti-IL-36R antibody. In other words, the methods disclosed herein can automatically assess the severity of skin lesions in patients suffering from ND (especially ND with visible signs and symptoms such as skin erythema, pustules and / or scales), and can help treating physicians quickly and easily form an objective assessment of the patient's disease severity, which is required for effective treatment.

[0007] definition A phrase such as "aspect" does not imply that such aspect is essential to the invention or that such aspect applies to all configurations of the subject technology. Disclosure of an aspect may apply to one or more configurations. An aspect may provide one or more examples of the present disclosure. A phrase such as "an embodiment" does not imply that such an embodiment is essential to the subject technology or that such an embodiment applies to all configurations of the subject technology. An embodiment may also include any feature. Disclosure of an embodiment may also apply to higher-level embodiments. An embodiment may provide one or more examples of the present disclosure.

[0008] The term "about" is generally intended to mean an acceptable degree of error or variation for the quantity measured, given the nature or precision of the measurement. Typical and exemplary degrees of error or variation are within 5%, within 3%, or within 1% of a given value or range of values. For example, the expression "about 100" includes 105 and 95, 103 and 97, or 101 and 99, and all values ​​therebetween (e.g., for the range of 95-105, 95.1, 95.2, etc.; for the range of 97-103, 97.1, 97.2, etc.; or for the range of 99-101, 99.1, 99.2, etc.). Numerical quantities given herein are approximate unless otherwise stated, meaning that the term "about" can be inferred when not expressly stated.

[0009] A "pharmaceutical composition" in this context refers to a liquid or powder preparation that is in a form that allows the biological activity of the active ingredient(s) to be demonstrably effective and does not contain additional ingredients that are significantly toxic to the patient to whom the composition is administered. Such compositions are sterile.

[0010] The term "patient" as used in reference to treatment refers to any animal classified as a mammal, including humans, domestic and farm animals, and zoo, sports, or pet animals, such as dogs, horses, cats, cows, etc., that suffers from one or more symptoms of a neutrophilic skin disease. Preferably, the mammal is a human.

[0011] As used herein, the terms "treat", "treating" and the like refer to relieving symptoms, eliminating the cause of symptoms, either temporarily or permanently, or preventing or delaying the onset of symptoms of a specified disorder or condition. These terms are meant to include therapeutic and prophylactic or suppressive measures for a disease or disorder, which result in any clinically desirable or beneficial effect, including, but not limited to, alleviating or reducing one or more symptoms, regressing, slowing, or halting the progression of a disease or disorder. Thus, for example, the term "treatment" includes administering an agent before or after the onset of symptoms of a disease or disorder, thereby preventing or eliminating one or more signs of a disease or disorder. As another example, the term includes administering an agent after clinical manifestation of a disease to combat symptoms of the disease. Additionally, administration of an agent after onset and after clinical symptoms have occurred includes "treatment" or "therapy" as used herein, if the administration affects clinical parameters of the disease or disorder, such as the extent of tissue damage, regardless of whether the treatment results in an improvement of the disease. Furthermore, so long as the composition of the present invention, alone or in combination with another therapeutic agent, alleviates or ameliorates at least one symptom of the disorder being treated compared to the symptom in the absence of the humanized anti-IL-36R antibody composition, the result should be considered an effective treatment of the underlying disorder, regardless of whether all symptoms of the disorder are alleviated.

[0012] The term "therapeutically effective amount" is used to refer to an amount of an active agent that eliminates or ameliorates one or more of the symptoms of the disorder being treated. In another aspect, a "therapeutically effective amount" refers to a target serum concentration that has been shown to be effective, for example, in slowing the progression of a disease.

[0013] As used herein, the term "skin lesion" refers to an area of ​​the skin of an ND patient that exhibits visible skin manifestations, or signs and symptoms, including erythema, pustules, and / or scaling.

[0014] Although any methods and materials similar or equivalent to those described herein can be used in the practice of the present invention, the preferred methods and materials are described herein. All publications mentioned herein are incorporated by reference as if set forth in their entirety.

[0015] Neutrophilic dermatoses (ND) are a diverse group of conditions with common features and overlapping pathophysiology. They include generalized pustular psoriasis (GPP), palmoplantar pustulosis (PPP), hidradenitis suppurativa (HS), acute generalized exanthematous pustulosis, acute febrile neutrophilic dermatosis (Sweet's syndrome), nonmicrobial pustulosis of the skin folds (APF), Behçet's disease, intestinal bypass syndrome (gut-associated dermatitis-arthritis syndrome), gut-associated dermatitis-arthritis syndrome (BADAS), CARD14-mediated pustular psoriasis (CAMPS), cryopyrin-associated periodic syndrome (CAPS), interleukin-36 receptor antagonist deficiency (DIRTA), interleukin-I receptor antagonist deficiency (DIRA), persistent elevated erythema. plaques; histiocytic neutrophilic dermatitis; acropustulosis of childhood; neutrophilic skin disease of the dorsum of the hands; neutrophilic eccrine hidradenitis; neutrophilic urticarial skin disease; palisade neutrophilic granulomatous dermatitis; psoriasis vulgaris; pyoderma gangrenosum, acne, hidradenitis suppurativa (PASH) syndrome; pyoderma gangrenosum (PG); pyoderma gangrenosum and acne (PAPA); septic arthritis; the skin lesions of Behçet's disease; Still's disease; subcorneal pustulosis (Sneddon-Wilkerson disease); synovitis, acne, pustulosis-osteophytosis; osteitis (SAPHO) syndrome; rheumatic neutrophilic dermatitis (RND), and ichthyosis (and its subtypes including Netherton syndrome (NS)).

[0016] For example, hidradenitis suppurativa (HS) is an inflammatory disease characterized by recurrent painful abscesses and fistulas. Patients with HS objectively have one of the lowest quality of life measures of any skin disease. Lesions characteristically occur in the axilla, groin, submammary, and / or anogenital regions of the body. HS lesions can progress to form sinus tracts and expanding abscesses.

[0017] IL36R is a novel member of the IL1R family that forms a heterodimeric complex with IL1R accessory protein (IL1RAcp) and IL1Rrp2, which are associated with epithelial-mediated inflammation and barrier dysfunction. The heterodimeric IL36R system with stimulatory (IL36α, IL36β, IL36γ) and inhibitory (IL36Rα and IL38) ligands shares some structural and functional similarities with other members of the IL1 / ILR family, such as IL1, IL18, and IL33. All IL1 family members (IL1α, IL1β, IL18, IL36α, IL36β, IL36γ, and IL38) signal through unique cognate receptor proteins that, upon ligand binding, recruit a common IL1RAcP subunit and activate NFγB and MAP kinase pathways in receptor-positive cell types (Dinarello, 2011; Towne et al., 2004; Towne et al., 2011). Genetic human studies have established a strong link between IL36R signaling and skin inflammation, as demonstrated by the development of generalized pustular psoriasis in patients with loss-of-function mutations in IL36Rα that result in uncontrolled IL36R signaling ( Marrakchi et al., 2011 ).

[0018] Given the strong association between the IL36 pathway and neutrophilic skin diseases, and without wishing to be bound by this theory, it is believed that IL36R biology contributes to the pathophysiology of these conditions and therefore blocking IL36R activation would be of benefit in patients suffering from neutrophilic skin diseases. However, objective assessment of disease severity in ND patients is important for timely and effective treatment.

[0019] Antibodies of the Invention The anti-IL 36R antibodies of the present invention are disclosed in U.S. Pat. No. 9,023,995 or WO 2013 / 074569, the contents of each of which are incorporated herein by reference in their entirety.

[0020] As used herein, the term "antibody" includes immunoglobulin molecules and multimers thereof (e.g., IgM) that contain four polypeptide chains, two heavy (H) chains and two light (L) chains, interconnected by disulfide bonds. In a typical antibody, each heavy chain contains a heavy chain variable region (referred to herein as HCVR or V H The heavy chain constant region is made up of three domains: H 1. C H 2, and C H Each light chain comprises a light chain variable region (referred to herein as LCVR or V L The light chain constant region comprises one domain (C L 1) is included. H Area and V L The regions can be further subdivided into regions of hypervariability, called complementarity determining regions (CDRs), interspersed with more conserved regions, called framework regions (FRs). H and V L is composed of three CDRs and four FRs, which are arranged from amino terminus to carboxy terminus in the following order: FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4. In different embodiments of the present invention, the FRs of an anti-IL-36R antibody (or an antigen-binding portion thereof) may be identical to the human germline sequence or may be naturally or artificially modified. An amino acid consensus sequence may be defined based on the parallel analysis of two or more CDRs.

[0021] The term "antibody" as used herein also includes antigen-binding fragments of complete antibody molecules. The terms "antigen-binding portion" of an antibody, "antigen-binding fragment" of an antibody, and the like, as used herein, include any naturally occurring, enzymatically obtainable, synthetic, or genetically engineered polypeptide or glycoprotein that specifically binds to an antigen to form a complex. Antigen-binding fragments of antibodies can be obtained from complete antibody molecules using any suitable standard technique, such as, for example, proteolytic digestion, or recombinant genetic engineering techniques involving the manipulation and expression of DNA encoding antibody variable domains and optionally constant domains. Such DNA is known and / or readily available, for example, from commercial sources, DNA libraries (including, for example, phage-antibody libraries), or can be synthesized. The DNA can be sequenced and manipulated chemically or by using molecular biology techniques, for example, to place one or more variable and / or constant domains in the appropriate configuration, or to introduce codons, create cysteine ​​residues, modify, add, or delete amino acids, etc.

[0022] Non-limiting examples of antigen-binding fragments include: (i) Fab fragments; (ii) F(ab')2 fragments; (iii) Fd fragments; (iv) Fv fragments; (v) single chain Fv (scFv) molecules; (vi) dAb fragments; and (vii) minimal recognition units consisting of amino acid residues mimicking the hypervariable regions of an antibody (e.g., isolated complementarity determining regions (CDRs), such as CDR3 peptides), or constrained FR3-CDR3-FR4 peptides. Other engineered molecules, such as domain-specific antibodies, single domain antibodies, domain deleted antibodies, chimeric antibodies, CDR-grafted antibodies, diabodies, triabodies, tetrabodies, minibodies, nanobodies (e.g., monovalent nanobodies, bivalent nanobodies, etc.), small modular immunopharmaceuticals (SMIPs), and shark variable IgNAR domains, are also encompassed by the term "antigen-binding fragment" as used herein.

[0023] Antigen-binding fragments of antibodies typically contain at least one variable domain. A variable domain may be of any size or amino acid composition and generally contains at least one CDR adjacent to or in frame with one or more framework sequences. L Domain-associated V H In an antigen-binding fragment having a domain, H Domain and V L The domains may be positioned relative to each other in any suitable configuration. For example, the variable region may be a dimer, with the V H -V H , V H -V L , or V L -V L Alternatively, the antigen-binding fragment of the antibody may contain a monomeric V H Or V L It may contain domains.

[0024] The antibody used in the method of the present invention may be a human antibody. The term "human antibody", as used herein, is intended to include antibodies having variable and constant regions derived from human germline immunoglobulin sequences. Nevertheless, the human antibody of the present invention may include amino acid residues (e.g., mutations introduced by random or in vitro site-directed mutagenesis or by in vivo somatic mutation), e.g., in the CDRs, particularly CDR3, that are not encoded by human germline immunoglobulin sequences. However, the term "human antibody", as used herein, is not intended to include antibodies in which CDR sequences derived from the germline of another mammalian species (e.g., mouse) have been grafted onto human framework sequences.

[0025] The antibody used in the method of the present invention may be a recombinant human antibody. The term "recombinant human antibody" as used herein is intended to include all human antibodies prepared, expressed, generated, or isolated by recombinant means, such as, for example, antibodies expressed using a recombinant expression vector transfected into a host cell (described further below), antibodies isolated from a recombinant combinatorial human antibody library (described further below), antibodies isolated from an animal (e.g., a mouse) that is transgenic for human immunoglobulin genes (e.g., Taylor et al. (1992) Nucl. Acids Res. 20:6287-6295), or human antibodies prepared, expressed, generated, or isolated by any other means involving splicing of human immunoglobulin gene sequences to other DNA sequences. Such recombinant human antibodies have variable and constant regions derived from human germline immunoglobulin sequences. However, in certain embodiments, such recombinant human antibodies are subjected to in vitro mutagenesis (or in vivo somatic mutagenesis, when animals transgenic for human Ig sequences are used) to thereby improve the V H Area and V L The amino acid sequence of the region is H Sequence and V L These are sequences that are derived from and related to sequences, but may not naturally occur within the human antibody germline repertoire in vivo.

[0026] According to certain embodiments, the antibody used in the method of the present invention specifically binds to IL-36R. The term "specifically binds" and the like means that the antibody or its antigen-binding fragment forms a complex with the antigen that is relatively stable under physiological conditions. Methods for determining whether an antibody specifically binds to an antigen are well known in the art, and include, for example, equilibrium dialysis, surface plasmon resonance, and the like. For example, an antibody that "specifically binds" to IL-36R, as used in the context of the present invention, has a Kd of less than about 1000 nM, less than about 500 nM, less than about 300 nM, less than about 200 nM, less than about 100 nM, less than about 90 nM, less than about 80 nM, less than about 70 nM, less than about 60 nM, less than about 50 nM, less than about 40 nM, less than about 30 nM, less than about 20 nM, less than about 10 nM, less than about 5 nM, less than about 4 nM, less than about 3 nM, less than about 2 nM, less than about 1 nM, or less than about 0.5 nM, as measured in a surface plasmon resonance assay. Kd and includes antibodies that bind to IL-36R or a portion thereof. However, an isolated antibody that specifically binds human IL-36R may have cross-reactivity to other antigens, such as IL-36R molecules from other (non-human) species.

[0027] In certain exemplary embodiments related to any aspect of the invention, an anti-IL-36R antibody or antigen-binding fragment thereof that can be used in the context of the methods of the invention comprises: a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 35, 102, 103, 104, 105, 106, or 140 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 or 141 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, 111, or 142 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3).

[0028] Embodiment In a first aspect, a computer-implemented method is provided for predicting the severity of a palmoplantar pustulosis condition of a patient using a deep neural network model. Predicting such a severity level (severity score) using a deep neural network means that the neural network implements a classifier suitable for classifying the patient's condition according to a classification scheme learned by the neural network in a training phase based on suitable training data. In general, the term "predict" as used herein is used in the commonly known sense in the context of classification tasks performed by neural networks. In this context, the individually trained neural network classification model returns a predicted class label in response to a test input provided to the neural network. That is, the term "predict" should be understood as classifying the current test input, and does not mean that the patient's future medical condition is predicted. In the context of this specification, the test input is classified into a severity score that reflects the severity level of the patient's current ND condition. In other words, the prediction / classification corresponds to an evaluation of the patient's ND condition according to the test input.

[0029] A computer system executing the computer-implemented method receives a test input, which is a digital image showing a skin area of ​​a patient. The computer system receives the digital image via a suitable interface. A patient suffering from PPP typically exhibits erythema, pustules, and / or scaling on the skin of characteristic body parts typically affected by PPP. In the case of PPP, such characteristic body parts include the patient's left and right palms and the left and right soles of the feet. Those skilled in the art will appreciate that in the case of an animal body, the palms and soles may correspond to the respective feet of the animal body. The test input digital image includes a grid having at least a pair of tiles. Advantageously, the tiles are of equal size, a first tile showing a skin area on the front of the characteristic body part of the patient, and a second tile showing a skin area on the back of the characteristic body part of the patient. For a patient suffering from PPP, at least one of the skin areas shows at least one of erythema, pustules, and scaling on the skin. For example, a pair of tiles may include two images showing the front and back of the patient's right palm. A second pair of tiles may be included within a grid showing the front and back of the patient's left palm. Further, the rows of the grid may include further pairs of tiles associated with the left and / or right soles of the patient. The sub-images within each tile may be captured by using a digital camera. The image grid may be composed of single images showing either the back or the front of the respective body part. Such image synthesis into a final image grid may be performed by the digital camera or by a pre-processing step that combines the individual images into a corresponding image grid. That is, the tiles of the input image grid show sub-images of characteristic body parts of the same patient captured within a relatively short time interval such that all tiles reflect the same medical condition of that patient.

[0030] In a further step, the computer system predicts the Palmoplantar Pustulosis Global Assessment (PPPG A) score of the subject patient using a predictor module with a deep neural network model. The received image grid thereby serves as an input to a deep neural network model (DNN). The DNN is pre-trained with a training dataset including a plurality of training images having the same structure as the test input. That is, the training images include the same number of tiles showing the same characteristic body parts of the test patient in the same order as the test input images. The images were captured from a plurality of test patients selected according to predefined inclusion / exclusion criteria that ensure that the training patients have a history of generalized pustular psoriasis without any conflicting disease. The inclusion / exclusion criteria are described in detail in the "Detailed Description" section. The training dataset includes a plurality of training images of each test patient captured at different times during a predefined minimum time interval. Each training image is annotated with one or more severity scores associated with erythema, pustules, and scale as ground truths reflecting the severity of the palmoplantar pustulosis condition of the respective training subject patient at the time the training image was captured. In other words, one training image may have a single annotation for the overall severity score of the training patient as the ground truth, which already reflects the total severity score taking into account all of the PPP symptoms. Alternatively, the training image may be annotated with individual severity scores for each of the erythema, pustules, and scale symptoms. Based on such individual severity scores, the total severity score for the patient is finally determined by the predictor. To assess the severity of ND condition, the training images do not need to include images of healthy patients. However, some of the selected training patients may not show visible signs of erythema, pustules, or scale. However, there is no restriction on using healthy patients as training patients.

[0031] In a second aspect, a computer-implemented method is provided for predicting the severity of a patient's generalized pustular psoriasis condition using a deep neural network model. The architecture of the deep neural network model is the same as that of the first aspect. However, there are differences between the received test input and the corresponding training data used in the second aspect. In the second aspect, the characteristic body parts shown in the test input are selected from the trunk, the left and right lower limbs, and the left and right upper limbs. To process the test input of the second aspect, the deep neural network model is previously trained with a training data set, which has the same structure as the test input and includes training images captured from test patients according to a predetermined inclusion / exclusion criteria that determine that the training patients have a history of generalized pustular psoriasis without any reciprocal disease. Again, the training data set includes multiple training images of each test patient captured at different times during a predetermined minimum time interval, and each training image is annotated with one or more severity scores associated with erythema, pustules, and scale as ground truths that reflect the severity of the generalized pustular psoriasis condition of the respective training patient at the time the training image is captured. Once the DNN is trained accordingly, it predicts a Global Generalized Pustular Psoriasis Physician Assessment (GPPGA) score for a patient based on the test inputs received.

[0032] The following optional features may be used in the computer-implemented methods according to the first and second aspects.

[0033] In one implementation, the training images can be annotated with one of at least three severity score values ​​covering a severity range from none to severe. A single annotated severity score value can reflect the average of the individual erythema severity score, pustule severity score, and scaling severity score of each training image. For this training scenario, the trained deep neural network model provides a single severity score value as output for the test input of that patient.

[0034] In an alternative embodiment, each training image is annotated with an individual erythema severity score value, a pustule severity score value, and a scale severity score value. Each individual severity score value is one of at least three severity score values ​​covering a severity range from none to severe for each erythema severity, pustule severity, and scale severity on the training image. In this implementation, the trained deep neural network model provides as output an individual severity score value for each of the erythema severity, pustule severity, and scale severity. Thus, the predictor module is configured to determine a single severity score for the test input of the patient based on averaging the determined individual severity score values.

[0035] The severity ranges of the above two implementations can also include more than three severity score values. For example, the following five score values ​​for severity levels (none, almost none, mild, moderate, and severe, with severity increasing from none to severe) can be used. The severity levels can be implemented using a corresponding five-point scale, including 0 if the predicted severity score is equal to zero, 1 if the predicted severity score is greater than 0 but less than 1.5, 2 if the predicted severity score is 1.5 or greater but less than 2.5, 3 if the predicted severity score is 2.5 or greater but less than 3.5, and 4 if the predicted severity score is 3.5 or greater. A predicted severity score of 2 or greater can be an indication for administering a recommended dose to the patient based on the pharmacologic effective amount of anti-IL-36R antibody. The five-point scale with a threshold value of "2" is a commonly used scale for severity assessment. However, other scales can be used as well.

[0036] In one embodiment, the invention relates to a method for treating a patient suffering from a neutrophilic skin disease condition, comprising: (a) identifying skin lesions exhibiting erythema, pustules and / or scaling in said patient, scoring the severity of each of the erythema, pustules and scaling and calculating a total Neutrophilic Skin Disease Global Assessment (NDGA) score or Neutrophilic Skin Disease Global Pustule (NDGP) score for the patient, wherein the identifying, scoring and calculating is a computer-implemented method comprising: (i) providing a digital image of a skin area of ​​the patient; (ii) scoring the severity of erythema, pustules and / or scaling of the skin lesions using a 5-point severity scale of 0 to 4, the 5-point severity scale including 0 for absent, 1 for almost none, 2 for mild, 3 for moderate and 4 for severe; and (iii) calculating the total NDGA score or NDGP score for the patient on the 5-point scale of 0 to 4. The total NDGA score is the sum of the erythema severity score, pustule severity score, and scale severity score obtained in step (ii) divided by 3, and the NDGP score is the pustule severity score obtained in step (ii), and the 5-point scale includes 0, 1, 2, 3, and 4, where 0 is when the total NDGA score or NDGP score is equal to 0, 1 is when the total NDGA score or NDGP score is greater than 0 but less than 1.5, 2 is when the total NDGA score or NDGP score is 1.5 or more but less than 2.5, 3 is when the total NDGA score or NDGP score is 2.5 or more but less than 3.5, and 4 is when the total NDGA score or NDGP score is 3.5 or more. The above method also includes (b) determining whether the patient exhibits a score of 2 or more for the total NDGA score or NDGP score, and administering to the patient a recommended dose based on a pharma- ceutical effective amount (depending on the first or second aspect) of an anti-IL-36R antibody for treating PPP or GPP, respectively.

[0037] For example, a pharmacologic effective amount of an anti-IL-36R antibody is in the range of about 0.001 to about 1200 mg. In one embodiment, a pharmacologic effective amount of spesolimab (BI655130) is in the range of about 0.001 to about 1200 mg. Thus, an anti-IL-36R antibody (e.g., spesolimab (BI655130)) may be administered within the range of 0.001 to 1200 mg.

[0038] In one embodiment, the severity prediction based on the test input image grid can be further improved by combining a set of features of the patient's tabulated clinical data with features extracted from the patient's individual digital image for severity prediction of the patient, whereby a particular set of features of the tabulated clinical data and the respective corresponding digital image are associated with the same severity of the patient's condition. In other words, the test input images and the respective tabulated clinical data are captured within a given time interval (e.g., on the same day) to ensure that the digital images and the respective corresponding tabulated clinical data reflect the same severity score for either PPP or GPP (depending on the test input and the corresponding predictor implementation) of the patient.

[0039] For this embodiment, two alternative implementations can also be used. In one implementation, the features of the tabulated clinical data are normalized and concatenated with the respective features extracted from the convolutional layer of the deep neural network, and the concatenated feature set is used as the input layer of the classification (fully connected) layer of the deep neural network model. The deep neural network in this implementation is trained with respective augmented training data including multiple training images and associated feature sets of the tabulated clinical data. The deep neural network model may implement any of the following algorithms: convolutional neural networks (e.g., ResNet, EfficientNet, or ConvNeXT architectures), vision transformers, and combination networks with convolutional and attention layers.

[0040] In an alternative implementation, the predictor further comprises a clinical data classifier, which is trained on the tabulated clinical data features associated with each training image of the deep neural network model using the same ground truth. That is, if a training image is annotated with a particular severity score, the corresponding set of tabulated clinical data features is also annotated with the same particular severity score as the ground truth. The deep neural network and the clinical data classifier are then trained together by using ensemble learning. The clinical data classifier may be based on an architecture different from DNN. For example, it may be implemented using logistic regression, gradient boosting, random forest, or support vector machine. In this alternative implementation, the predictor finally combines the output of the deep neural network model and the output of the clinical data classifier into a single severity score.

[0041] In one embodiment there is provided a computer program product for predicting severity of a palmoplantar pustulosis condition according to the first aspect or a generalized pustular psoriasis condition according to the second aspect, which when loaded into a memory of a computing device and executed by at least one processor of the computing device causes the at least one processor to perform the steps of the computer-implemented methods disclosed herein.

[0042] In one embodiment, a method for treating a patient suffering from palmoplantar pustulosis is provided. In a first step, the computer-implemented method according to the first aspect is performed to predict the severity score of the palmoplantar pustulosis condition of the patient. Then, the computer compares the predicted severity score with a predetermined drug administration threshold. If the predicted severity score is equal to or greater than the predetermined drug administration threshold, the computer determines a recommended dose based on a pharmacologic effective amount of an anti-interleukin-36 receptor antibody. Finally, the recommended dose of anti-interleukin-36 receptor for the treatment of palmoplantar pustulosis is administered.

[0043] In a further embodiment, a method for treating a patient suffering from generalized pustular psoriasis is provided. In a first step, the computer-implemented method according to the first aspect is performed to predict the severity score of the patient's generalized pustular psoriasis condition. Then, the computer compares the predicted severity score with a predetermined drug administration threshold. If the predicted severity score is equal to or greater than the predetermined drug administration threshold, the computer determines a recommended dose based on a pharmacologic effective amount of anti-interleukin-36 receptor antibody. Finally, the patient is administered a recommended dose of anti-interleukin-36 receptor antibody for the treatment of generalized pustular psoriasis.

[0044] In one embodiment, a computer system is provided for predicting the severity of a palmoplantar pustular psoriasis condition (GPP) and / or the severity of a generalized pustular psoriasis condition (GPP) of a patient using one or more respectively trained deep neural network models. The computer system may load and execute the above computer program product to perform the computer-implemented steps of the methods disclosed herein.

[0045] The computer system comprises an interface configured to receive a test input digital image having a skin area of ​​the patient. For example, the interface may receive the test input via a wireless communication interface or a wired interface (e.g., an interface for data exchange over a local area network). The test input digital image comprises a plurality of images arranged as tiles in a grid. The grid comprises at least a pair of tiles, a first tile showing a skin area on the front of the characteristic body part of the patient and a second tile showing a skin area on the back of the characteristic body part of the patient. Advantageously, the tiles of the test input are of the same size. In a patient suffering from PPP or GPP, at least one of the skin areas shows at least one of erythema, pustules, and scales on the skin. If the patient suffers from PPP, the characteristic body parts are selected from the left and right palms, the left and right soles of the feet. If the patient suffers from GPP, the characteristic body parts are selected from the trunk, the left and right lower legs, the left and right upper legs. The skilled person is aware that for animal patients, the corresponding characteristic body parts may be called differently. The grid image may be received directly from a separate digital camera, or from a pre-processing unit that constructs the grid image from a number of single images captured by a standard digital camera.

[0046] The system further includes a predictor module that predicts the Palmoplantar Pustulosis Global Assessment Score and / or Generalized Pustular Psoriasis Physician Global Assessment Score for the patient (depending on the type of test input and the training of the DNN). The predictor thereby applies one or more individually trained deep neural network model(s) to the test input. That is, the test input serves as an input to the DNN(s). The DNN is trained with a corresponding training dataset that includes a plurality of training images having the same structure as the test input, the training images being captured from a plurality of test patients (90) selected according to predetermined inclusion / exclusion criteria that determine that the training patients (90) have a reciprocal disease-free PPP history and / or GPP history, respectively.

[0047] Again, the training data set includes multiple training images of each test patient captured at different times during a predetermined minimum time interval, and each training image is annotated with one or more severity scores associated with erythema, pustules, and scale as ground truths reflecting the severity of palmoplantar pustulosis condition and / or the severity of generalized pustular psoriasis condition of the respective training patient at the time the training image was captured. In other words, the DNN(s) are trained in a training phase with suitable training data that allows the DNN to classify the severity of the patient's ND condition into either PPP or GPP before being applied to the test input. Advantageous DNN topologies for this purpose include, but are not limited to, Vision Transformer, combinatorial networks that stack convolutional and attention layers, and convolutional neural network topologies such as ResNet, EfficientNet, or ConvNeXT architectures.

[0048] In an optional embodiment, the predictor module may combine a set of features of the patient's tabulated clinical data with features extracted from the patient's individual digital image for severity prediction of that patient, whereby a particular set of features of the tabulated clinical data and each corresponding digital image is associated with the same severity of the patient's condition. Advantageously, the capture of the digital test input images and the corresponding tabulated clinical data test inputs is performed within a given time interval (e.g., during the same day) such that the same ND condition of the patient is reflected by both the image test input and the tabulated data test input.

[0049] For any embodiment, two alternative implementations are disclosed. In one implementation, the predictor is further configured to normalize the tabulated clinical data features and concatenate the normalized features with the corresponding image features extracted from a convolutional layer of a deep neural network (CNN). The concatenated feature set is then used as an input layer of the classification layer of the DNN. In this implementation, the DNN is trained with separate augmented training data including a plurality of training images and associated feature sets of the tabulated clinical data features.

[0050] In an alternative implementation, the predictor further comprises a clinical data classifier trained on the tabular clinical data features associated with each training image of the deep neural network model using the same ground truth. The joint training of the clinical data classifier (CDC) and the deep neural network is performed by using so-called ensemble learning. Furthermore, the predictor is configured to combine the output of the deep neural network model and the output of the clinical data classifier into a single severity score.

[0051] In one embodiment, the system further comprises a severity checker module that compares the predicted severity score to a predetermined medication threshold. The predicted severity score has one of at least three severity score values ​​covering a severity range from none to severe. If the predicted severity score is equal to or greater than the predetermined medication threshold, the severity checker assigns the patient as a candidate for treatment with an anti-interleukin-36 receptor antibody.

[0052] In one embodiment, the system further comprises an antibody dosing module configured to determine a recommended dosage for the candidate treatment based on a pharmacologic effective amount of an anti-interleukin-36 receptor antibody suitable for treating the palmoplantar pustular psoriasis condition and / or generalized pustular psoriasis condition of the candidate treatment based on the predicted severity score, and corresponding dosage instructions are provided to a drug administration entity for administering the treatment to the patient.

[0053] As mentioned above, the invention disclosed herein can also be used in methods for treating patients suffering from neutrophilic skin diseases such as PPP or GPP.

[0054] The disclosed methods and systems may also be used by methods for modifying, discontinuing, or continuing therapy of an individual receiving an anti-IL-36R antibody for the treatment of PPP / GPP, with the predicted severity score serving as a basis for modifying, discontinuing, or continuing treatment, or for characterizing the patient as likely to respond to treatment with an anti-IL-36 antibody.

[0055] Furthermore, if the patient's predicted severity score decreases over time, i.e., if repeated performance of the method according to the first and / or second aspect over a particular time interval results in a decrease in the severity score, the patient is characterized as likely to respond to treatment with an anti-IL-36 antibody based on the previously determined severity score(s).

[0056] In one embodiment of any of the above aspects or embodiments for the treatment of a patient having a PPP and / or GPP condition, the anti-IL-36R antibody comprises: a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 35, 102, 103, 104, 105, 106, or 140 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 or 141 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, 111, or 142 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3).

[0057] In one embodiment of any of the above aspects or embodiments for the treatment of a patient having a PPP and / or GPP condition, the anti-IL-36R antibody comprises: a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 35, 102, 103, 104, 105, 106, or 140 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 141 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, 111, or 142 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3).

[0058] In one embodiment of any of the above aspects or embodiments for the treatment of a patient having a PPP and / or GPP condition, the anti-IL-36R antibody is Ia) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 102 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or II.a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 103 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or III.a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 104 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or IV.a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 105 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or Va) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 106 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or VI.a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 140 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or VII.a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 104 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 141 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, 111, or 142 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3).

[0059] In one embodiment of any of the above aspects or embodiments for the treatment of a patient having a PPP and / or GPP condition, the anti-IL-36R antibody is (i) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 77, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 87; or (ii) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 77, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 88; or (iii) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 77, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 89; or (iv) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 80, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 87; or (v) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 80, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 88; or (vi) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 80, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 89; or (vii) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 85, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 100; or (viii) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 85, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 101; or (ix) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 86, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 100; or (x) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 86, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 101.

[0060] In one embodiment of any of the above aspects or embodiments for the treatment of a patient having a PPP and / or GPP condition, the anti-IL-36R antibody is i. a light chain comprising the amino acid sequence of SEQ ID NO: 115, and a heavy chain comprising the amino acid sequence of SEQ ID NO: 125; or ii. a light chain comprising the amino acid sequence of SEQ ID NO: 115, and a heavy chain comprising the amino acid sequence of SEQ ID NO: 126; or iii. a light chain comprising the amino acid sequence of SEQ ID NO: 115, and a heavy chain comprising the amino acid sequence of SEQ ID NO: 127; or iv. a light chain comprising the amino acid sequence of SEQ ID NO: 118 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 125; or v. a light chain comprising the amino acid sequence of SEQ ID NO: 118 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 126; or vi. a light chain comprising the amino acid sequence of SEQ ID NO: 118 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 127; or vii. a light chain comprising the amino acid sequence of SEQ ID NO: 123, and a heavy chain comprising the amino acid sequence of SEQ ID NO: 138; or viii. a light chain comprising the amino acid sequence of SEQ ID NO: 123 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 139; or ix. A light chain comprising the amino acid sequence of SEQ ID NO: 124, and a heavy chain comprising the amino acid sequence of SEQ ID NO: 138. In one embodiment related to any of the above aspects or embodiments for the treatment of patients with a PPP and / or GPP condition, the anti-IL-36R is spesolimab (BI655130).

[0061] The methods and systems disclosed herein may also be applied to predict the severity score of additional (associated) neutrophilic skin disease conditions of a patient when adapting the image training data and test inputs to the corresponding disease, such as hidradenitis suppurativa (HS); acute generalized exanthematous pustulosis; acute febrile neutrophilic dermatosis (Sweet's syndrome); amicrobial pustulosis of the skin folds (APF); Behcet's disease; intestinal bypass syndrome (gut-associated dermatitis-arthritis syndrome); gut-associated dermatitis-arthritis syndrome (BADAS); CARD14-mediated pustular psoriasis (CAMPS); cryopyrin-associated periodic syndrome (CAPS); interleukin-36 receptor antagonist deficiency (DIRTA); interleukin-I receptor antagonist deficiency (DIRA); persistent erythema elevatum; histiocytic neutrophil disease (HES); These may include globular dermatitis; acropustulosis pediatricus; neutrophilic dermatosis of the dorsum of the hands; neutrophilic eccrine hidradenitis; neutrophilic urticarial dermatosis; palisade neutrophilic granulomatous dermatitis; psoriasis vulgaris; pyoderma gangrenosum, acne, hidradenitis suppurativa (PASH) syndrome; pyoderma gangrenosum (PG); pyoderma gangrenosum and acne (PAPA); septic arthritis; the skin lesions of Behcet's disease; Still's disease; subcorneal pustulosis (Sneddon-Wilkerson disease); synovitis, acne, pustulosis-osteophytosis; osteitis (SAPHO) syndrome; rheumatic neutrophilic dermatitis (RND), and ichthyosis (and its subtypes including Netherton syndrome (NS)).

[0062] In a related embodiment, the neutrophilic skin disease is hidradenitis suppurativa (HS). In a related embodiment, the neutrophilic skin disease is acute generalized exanthematous pustulosis. In a related embodiment, the neutrophilic skin disease is acute febrile neutrophilic skin disease (Sweet's syndrome). In a related embodiment, the neutrophilic skin disease is amicrobial pustulosis of the skin folds (APF). In a related embodiment, the neutrophilic skin disease is Behcet's disease. In a related embodiment, the neutrophilic skin disease is intestinal bypass syndrome (gut-associated dermatitis-arthritis syndrome). In a related embodiment, the neutrophilic skin disease is gut-associated dermatitis-arthritis syndrome (BADAS). In a related embodiment, the neutrophilic skin disease is CARD 14-mediated pustular psoriasis (CAMPS). In a related embodiment, the neutrophilic skin disease is cryopyrin-associated periodic syndrome (CAPS). In a related embodiment, the neutrophilic skin disease is Interleukin-36 Receptor Antagonist Deficiency (DIRTA). In a related embodiment, the neutrophilic skin disease is Interleukin-I Receptor Antagonist Deficiency (DIRA). In a related embodiment, the neutrophilic skin disease is persistent erythema elevatus. In a related embodiment, the neutrophilic skin disease is histiocytic neutrophilic dermatitis. In a related embodiment, the neutrophilic skin disease is acropustulosis pediatricus. In a related embodiment, the neutrophilic skin disease is neutrophilic skin disease of the dorsum of the hands. In a related embodiment, the neutrophilic skin disease is neutrophilic eccrine hidradenitis. In a related embodiment, the neutrophilic skin disease is neutrophilic urticarial skin disease. In a related embodiment, the neutrophilic skin disease is palisade neutrophilic granulomatous dermatitis. In a related embodiment, the neutrophilic skin disease is psoriasis vulgaris. In a related embodiment, the neutrophilic skin disease is pyoderma gangrenosum, acne, and hidradenitis suppurativa (PASH) syndrome. In a related embodiment, the neutrophilic skin disease is pyoderma gangrenosum (PG). In a related embodiment, the neutrophilic skin disease is pyoderma gangrenosum and acne (PAPA). In a related embodiment, the neutrophilic skin disease is septic arthritis. In a related embodiment, the neutrophilic skin disease is the skin lesions of Behcet's disease. In a related embodiment, the neutrophilic skin disease is Still's disease.In a related embodiment, the neutrophilic skin disease is subcorneal pustulosis (Sneddon-Wilkerson disease). In a related embodiment, the neutrophilic skin disease is synovitis, acne, pustulosis-osteitis, and osteitis (SAPHO) syndrome. In a related embodiment, the neutrophilic skin disease is rheumatic neutrophilic dermatitis (RND). In a related embodiment, the neutrophilic skin disease is ichthyosis (and its subtypes including Netherton syndrome (NS)).

[0063] It will be understood that any of the methods, administration schemes, and / or dosing regimens disclosed herein equally apply to the use of any of the disclosed anti-IL-36R antibodies in such methods, administration schemes, and / or dosing regimens, i.e., anti-IL-36R antibodies as disclosed herein for use in the treatment, prevention, amelioration, and / or amelioration of any of the disclosed diseases and / or conditions. In other words, the present invention also provides the use of an anti-IL-36R antibody as disclosed herein for the manufacture of a medicament for the treatment, prevention, amelioration, and / or amelioration of any of the disclosed diseases and / or conditions.

[0064] Additional features and advantages of the invention will be realized and attained by the elements and combinations particularly pointed out in the appended claims or the detailed description and drawings. It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention as described. [Brief description of the drawings]

[0065] [Figure 1] FIG. 1 is a simplified block diagram of a computer system for predicting the severity of a neutrophilic skin disease condition in a patient, according to one embodiment. [Figure 2A] 1 is a simplified flowchart of a computer-implemented method for predicting the severity of a patient's PPP / GPP condition, according to two embodiments of the present invention. [Figure 2B]1 is a simplified flowchart of a computer-implemented method for predicting the severity of a patient's PPP / GPP condition, according to two embodiments of the present invention. [Figure 3A] 1 is a simplified flow chart of a method for treating a PPP / GPP condition in a patient, according to two embodiments of the present invention. [Figure 3B] 1 is a simplified flow chart of a method for treating a PPP / GPP condition in a patient, according to two embodiments of the present invention. [Figure 4] FIG. 2 is a schematic diagram of an example test input grid image. [Diagram 5] FIG. 1 illustrates an exemplary embodiment of a severity score. [Figure 6A] FIG. 2 illustrates details of the deep neural network model used by the predictor module according to one embodiment. [Figure 6B] 13 illustrates an alternative implementation of the predictor module that includes tabulated clinical data as additional test input. [Figure 6C] 13 illustrates an alternative implementation of the predictor module that includes tabulated clinical data as additional test input. [Figure 7A] FIG. 1 is a diagram illustrating an example of a convolution step of a CNN. [Figure 7B] FIG. 13 illustrates an exemplary computation of a filter applied to a window on an input image to create a feature map or to create a new feature map from an input feature map. [Figure 8] FIG. 1 shows an example where three filters are applied to an input to generate three new feature maps. [Figure 9] FIG. 1 illustrates an exemplary embodiment of subsequent convolutions in a CNN for erythema. [Figure 10A] FIG. 13 is a diagram illustrating an example of a feature amount map transformed by an adjustment function. [Figure 10B] FIG. 13 shows an example of flattening a pooled feature map from a 2D array to a 1D array. [Figure 10C] FIG. 13 is a diagram illustrating an example of map pooling. [Figure 11A] FIG. 1 shows an example of erythema severity scores (on a 5-point severity scale from 0 to 4, as indicated) in digital images of skin lesions from a patient with GPP. [Figure 11B] FIG. 1 shows an example of pustule severity scores (on a 5-point severity scale from 0 to 4, as shown) in digital images of skin lesions from a patient with GPP. [Figure 11C] FIG. 1 shows an example of scale severity scores (on a 5-point severity scale from 0 to 4, as shown) in digital images of skin lesions from a patient with GPP. [Figure 12] FIG. 1 illustrates an example of a general-purpose computing device and a general-purpose mobile computing device that may be used with the techniques described herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0066] Before the present invention is described, it is to be understood that this invention is not limited to the particular methods and experimental conditions described, as such methods and conditions may vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the present invention will be limited only by the appended claims.

[0067] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to one skilled in the art that the subject technology may be practiced without some of these specific details. In other instances, well-known structures and techniques are not shown in detail so as not to obscure the present invention.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0069] FIG. 1 is a simplified block diagram of a computer system 100 for predicting the severity of a neutrophilic skin disease in a patient 10, according to one embodiment. In this example, the system is configured to predict the PPP and / or GPP status of the patient. However, the system may also be configured to cover other types of neutrophilic skin diseases. FIGS. 2A and 2B are simplified flowcharts of computer-implemented methods 1000, 2000 for predicting the severity of a PPP / GPP status in a patient 10. The methods can be performed by the computer system 100. FIGS. 3A and 3B are simplified flowcharts of methods 3000, 4000 of treating a PPP / GPP status. The methods 3000, 4000 include the methods 1000, 2000, respectively. All of the steps of the methods 1000, 2000 can be performed by the system 100. Also, most of the method steps of the methods 3000, 4000 can be performed by the system 100. In the following, computer system 100 is described in the context of method steps that may be performed by the system, and the following description therefore refers to the system diagram of Figure 1, and the method diagrams of Figures 2A, 2B, 3A, and 3B.

[0070] The system 100 is communicatively coupled to an image source (e.g., a digital camera 20, or a pre-processor capable of merging (from the system's point of view) a single image of the digital camera into a grid image 21). For the communicative coupling, a person skilled in the art can use a standard interface (not shown) for exchanging digital image data. The digital camera is used to capture images of skin areas on characteristic body parts of the patient 10. Depending on the disease for which the severity level (score) is predicted, different characteristic body parts of the patient are used as appropriate. For example, for the assessment of PPP severity, the characteristic body parts are the left and right palms, and the left and right soles of the feet. For the assessment of GPP severity, the characteristic body parts are the trunk, the left and right lower limbs, and the left and right upper limbs. The camera 20 is used to capture at least a pair of images showing the back and the front of one characteristic body part. If multiple characteristic body parts show disease symptoms, it may be advantageous to record multiple pairs of images, one pair for each affected body part. For example, for PPP severity prediction, a pair of images may be captured for the left palm (back and front) of the patient 10, and a further pair of images may be captured for the right foot (back and front). The image pair(s) are then merged into a grid image 21 that includes multiple pairs of tiles, each pair of tiles including a back image and a front image of a particular characteristic body part. Advantageously, each tile is of the same size. Some cameras may include such a merging function. However, the merging can also be performed by a pre-processor that provides such a merging function. It should be noted that from the camera's point of view, it is actually a post-processing step. However, from the system 100's point of view, the raw images are pre-processed before being received as a grid image.

[0071] Briefly referring to FIG. 4, a grid image 21 is shown, which includes tiles 22-1 to 22-4 showing skin areas on body parts that are characteristic body parts for GPP analysis and thus suitable for GPP severity prediction. In this example, a first pair of tiles 22-1, 22-2 show the anterior 22-f and posterior 22-b of a patient's torso. The black dots are schematic representations of skin lesions showing erythema, pustules, or scales that are symptoms of ND. A second pair of tiles 22-3, 22-4 show the anterior 22-f and posterior 22-b of the patient's lower extremities (left and right). The grid image 21 may also include a pair of tiles showing the posterior and anterior of the patient's upper extremities. To provide an impression of the visual appearance of ND symptoms on a patient's skin lesions, FIGS. 11A to 11C show real-world skin lesion images with erythema, pustules, and scale symptoms for various severity levels. Figure 11A shows two examples of erythema symptoms with associated individual severity scores on a 5-point scale (0-4) of severity levels covering none, almost none, mild, moderate, and severe. Similarly, Figure 11B shows two examples of pustular symptoms with their respective individual severity scores. Similarly, Figure 11C shows two examples of scaly symptoms with their respective individual severity scores.

[0072] The system 100 now receives the grid image 21 as a test input digital image that serves as an input to the predictor module 110 of the system (steps 1100, 1200). The predictor 110 predicts the Palmoplantar Pustular Psoriasis Global Assessment Score (PPPG A) (first aspect) and / or the Generalized Pustular Psoriasis Physician Global Assessment Score (PPPGA) 111 (second aspect) for the patient 10 using at least one respectively trained deep neural network model DNN1. The DNN1 is trained by a module 190, which may be an integrated module of the system 100 or a module provided on a remote computer communicatively coupled to the system 100 so as to provide the trained DNN1 to the predictor 110. In FIG. 1, the DNN1 is illustrated by its classification layer. However, the overall topology of the DNN1 is more complex, as explained in FIG. 6A. The training module 190 is communicatively coupled to or includes a data storage that provides a training image dataset 300. The training dataset 300 includes a plurality of training images having the same structure as the test input 21 (i.e., a grid structure having at least two tiles). The training images are captured from a plurality of test patients 90 selected according to predetermined inclusion / exclusion criteria that determine that the training patients 90 have a history of palmoplantar pustulosis and / or a history of generalized pustular psoriasis, respectively, without any opposing disease. The training dataset thereby includes a plurality of training images 310-360 of each training patient, captured at different time points during a predetermined minimum time interval. In one experimental implementation, approximately 50% of the training patients did not show pustules, scales, or erythema after treatment with spesolimab. However, if treatment for such training patients is stopped, they will relapse. In a test implementation, the training patients remained under study for at least three months.During this period, the skin appearance of each training patient was reassessed weekly (i.e., once a week, but not necessarily always on the same day of the week), i.e., in this implementation of the test, the minimum time interval was 3 months.

[0073] The selection criteria used for the selection of training patients to provide training images showing ND symptoms were as follows (a patient is eligible to be a training patient if he or she meets the following criteria): Patients with a GPPGA score of 1a.0 or 1, with a known and documented history of GPP (per ERASPEN criteria), regardless of IL36RN mutation status, and with evidence of previously experiencing fever, and / or asthenia, and / or muscle pain, and / or elevated C-reactive protein levels, and / or leukocytosis (above the upper limit of normal [ULN]) with peripheral blood neutrophilia, or 1b. Patients with an acute flare of moderate to severe intensity fulfilling the ERASPEN criteria for GPP, with a known and documented history of GPP (per ERASPEN criteria) regardless of IL36RN mutation status, and further with evidence of having previously experienced fever, and / or asthenia, and / or muscle pain, and / or elevated C-reactive protein levels, and / or leukocytosis (above the ULN) with peripheral blood neutrophilia, or Patients experiencing a first episode of acute GPP flare of moderate to severe intensity with evidence of fever and / or asthenia and / or muscle pain and / or elevated C-reactive protein levels and / or leukocytosis (above the ULN) with peripheral blood neutrophilia. For these patients, the diagnosis will be confirmed retrospectively by a central external expert / committee. 2. Patients may or may not be receiving background treatment with retinoids and / or methotrexate and / or cyclosporine. Patients must discontinue retinoids / methotrexate / cyclosporine prior to receiving the first dose of spesolimab or placebo. 3.Male or female patients aged 18-75 years at the time of screening. 4. Signed and dated written informed consent prior to participation in the study, in accordance with the ICHGCP and respective local laws, prior to the initiation of any screening procedures. 5. Women of childbearing potential must be prepared and able to use highly effective contraceptive methods according to ICH M3(R2) that have a low failure rate of less than 1% per year when used consistently and correctly. Women are considered of childbearing potential (WOCBP), i.e., capable of pregnancy until menarche and after menopause, unless permanently infertile. Permanent sterilization methods include hysterectomy, bilateral salpingectomy, and bilateral oophorectomy. Tubal ligation is not a permanent sterilization method. Postmenopausal status is defined as the absence of menstruation for 12 months without another medical cause.

[0074] The exclusion criteria used for the selection of training patients were as follows (i.e., if any of the following criteria applied, the patient was not selected as a training patient): 1. Patients with synovitis, acne, pustulosis, hyperostosis-osteitis syndrome. 2. Patients with primary erythrodermic scalp psoriasis vulgaris. 3. Patients with primary plaque caput psoriasis in the absence of pustules or patients with pustules confined to psoriatic plaques. 4. Drug-induced acute generalized exanthematous pustulosis. 5. Immediate life-threatening exacerbation of GPP or requiring intensive care, as determined by the investigator. Life-threatening complications include, but are not limited to, cardiovascular / cytokine-driven shock, pulmonary distress syndrome, or renal failure. 6. Severe progressive or uncontrolled liver disease, defined as a >3-fold ULN elevation in aspartate transaminase or alanine transaminase or alkaline phosphatase, or a >2-fold ULN elevation in total bilirubin. 7. Having been treated with: a. Any restricted medication as identified in Supplementary Table 1, or any medication that is assessed by the investigator as likely to interfere with the safe performance of the test; b. Prior exposure to spesolimab or another IL36R inhibitor. 8. Patients have received escalating doses of cyclosporine and / or methotrexate and / or retinoid maintenance therapy within 2 weeks prior to receiving the first dose of spesolimab / placebo. 9. Initiation of systemic medications such as cyclosporine and / or retinoids and / or methotrexate within 2 weeks prior to receiving the first dose of spesolimab / placebo. 10. Patients with congestive heart disease, as assessed by the investigator. 11. Active systemic infections (fungal and bacterial diseases) during the last 2 weeks prior to receiving the first drug dose, as assessed by the investigator. 12. Increased risk of infectious complications as assessed by the investigator (e.g., recent, pyogenic infection, any congenital or acquired immunodeficiency [e.g., human immunodeficiency virus (HIV)], previous organ or stem cell transplant). 13. Associated chronic or acute infection, including HIV or viral hepatitis. For patients screened while having a flare (inclusion criteria 1b or 1c), if HIV or viral hepatitis test results from Visit 1 are not available in time for randomization, these patients may receive randomized treatment as long as the investigator rules out active disease based on available documented medical history (i.e., negative test results for HIV and viral hepatitis) within 3 months prior to Visit 2. If the patient is treated and cured of acute infection, the patient may be rescreened. 14. Active or Latent Tuberculosis (TB): QuantiFERON® (or T-Spot®, if applicable) TB testing will be performed at screening. If the result is positive, the patient may enter the study only if further workup (according to local practice / guidelines) conclusively establishes that the patient has no evidence of active TB. Patients with active TB must be excluded. If the presence of latent TB was confirmed, treatment should have been initiated and maintained according to local guidelines. For patients screened while having a flare-up (inclusion criteria 1b or 1c), if the results of the TB test are not available in time for randomization, these patients may receive treatment for randomization (if they meet all other inclusion / exclusion criteria) as long as the investigator has ruled out active disease based on documented medical history available within 3 months prior to Visit 2 (i.e., negative for active TB). 15. History of allergy / hypersensitivity to systemically administered study medication or its excipients. 16. Any documented active or suspected malignancy or history of malignancy within 5 years prior to screening, except for adequately treated basal or squamous cell carcinoma of the skin or cervical intraepithelial neoplasia. 17. Currently enrolled in another investigational device or drug trial, or has been less than 30 days since completing another investigational device or drug trial(s) or receiving other investigational treatment(s). 18. Women who are pregnant, breastfeeding, or planning to become pregnant during the study. Women who have stopped breastfeeding prior to receiving the study drug do not need to be excluded from participation. They should abstain from breastfeeding for up to 16 weeks after receiving the study drug. 19. Major surgery (major surgery as assessed by the investigator) occurring within 12 weeks prior to receiving the first dose of investigational drug or planned during the study, e.g., hip replacement, aneurysm removal, gastric ligation, as assessed by the investigator. 20. Evidence that a current or pre-existing disease, non-GPP medical condition (including chronic alcohol or drug abuse or any condition), surgical procedure, psychiatric or social problem, medical laboratory findings (including vital signs and ECG) or screening laboratory values ​​are outside the reference ranges that, in the opinion of the investigator, are clinically significant and render the study participant unreliable to adhere to the protocol, comply with all study visits / procedures, or complete the study, compromise patient safety, or compromise the quality of the data.

[0075] Each training image is annotated with one or more severity scores associated with erythema, pustules, and scale as ground truths reflecting the severity of the palmoplantar pustular psoriasis condition and / or generalized pustular psoriasis condition of the respective training patient at the time the training image was captured, depending on the ND condition to be evaluated. For example, the Generalized Pustular Psoriasis Physician Assessment (GPPGA) relies on clinical evaluation of the subject's skin symptoms. It is a modified PGA, a physician assessment of psoriasis lesions, adapted for evaluation of patients with generalized pustular psoriasis (GPP). The investigator (or qualified site personnel) can score the erythema, pustules, and scale of all psoriasis lesions on the training image from 0 to 4. In one implementation, each element is graded separately and the training images are annotated with three individual ground truths, respectively. In this implementation, a separate deep neural network can be trained for each symptom (erythema, pustules, and scale). In another implementation, an average is determined from the individual scores as a composite score, and training images are annotated with the composite score as the ground truth. For example, a composite average score may be calculated as the sum of the individual scores for erythema, pustules, and scale divided by 3. For example, the total GPPGA score is "0" if the average = "0" for all three components, "1" if the average is less than the range "0" to "1.5", "2" if the average is less than the range "1.5" to "2.5", "3" if the average is less than the range "2.5" to "3.5", and "4" if the average is equal to or greater than "3.5". Lower scores indicate less severe symptoms, e.g., 0 is none and 1 is almost none. To obtain a score of 0 or 1, the patient must be fever-free in addition to the skin presentation requirement. Table 1 below shows an example of GPPGA scoring using a 5-point severity scale.

[0076] [Table 1]

[0077] FIG. 5 shows the above five-point scale example 112a used as a ground truth value for training the deep neural network model(s). In this implementation, of course, the predicted severity score classification for a given test input image is also one of the score values ​​on the five-point scale 112a. In an alternative implementation, the system was trained using only the three-point scale 112b, where "almost none" and "mild" are represented by a common score value of "1", and "moderate" and "severe" are represented by a common score value of "3". A suitable prediction / classification accuracy can also be achieved by this implementation.

[0078] In one embodiment (first aspect), the predictor module 110 predicts the palmoplantar pustulosis global assessment score of the patient 10 (steps 1200, 2200). In another embodiment (second aspect), the predictor module 110 predicts the total generalized pustular psoriasis physician global assessment score 111. For this purpose, the predictor 110 applies one or more respectively trained deep neural network models DNN1 to the received test input 21. As explained above, the only difference between the first and second aspects lies in the characteristic body parts for each ND condition shown on the training image. FIG. 6A shows an exemplary topology 600 for the deep neural network model used by the predictor 110.

[0079] Now referring to FIG. 6A, this figure shows an example topology 600 of a deep neural network model (DNN1 in FIG. 1), which is trained by respective training data to predict results based on input images. In the figure, an input image 605 is shown with four tiles showing ND symptoms on characteristic body parts of the patient to be evaluated. For training the neural network, skin images of patients with and without generalized pustular psoriasis (GPP) were used as training images. The training data includes different types of symptoms such as scale, and / or pustules, and / or erythema with various severity. For example, in the case of GPP, the neural network model learns whether a person has GPP if erythema and / or pustules and / or scale are present, and if they are present, learns the stage / degree / scaling score of erythema and / or pustules and / or scale, and outputs individual scores for symptoms, as well as a total Generalized Pustular Psoriasis Physician Global Assessment (GPPGA) score and Generalized Pustular Psoriasis Area and Severity Index (GPPASI). Similarly, one skilled in the art can train a deep neural network to identify the presence of PPP without further explanation.

[0080] The topology 600 includes a convolutional neural network (CNN). The first stage of the CNN is feature extraction, which includes convolutional and conditioned linear activation layers 610, 630, and pooling 620. In this stage, the model identifies the important features of erythema, scale, and pustules, as well as the important features that determine the scores of these symptoms, and creates a feature detector to identify whether these features are present in a new input image. The second stage of the convolutional neural network (CNN) incorporates an artificial neural network 640 (fully connected layer of DNN1) using the output / identification from the convolution stage to classify it as including erythema+ score, and / or including pustule+ score, and / or including scale+ score, or asymptomatic, or a single severity score representing a composite average value for the presence of all three symptoms.

[0081] FIG. 7A shows an example of the convolution step in a CNN. This step utilizes input data 701 to create a feature map 703 or feature detector. For example, when using backpropagation, the model learns features and detects these features in new images to create a feature detector that can perform the appropriate classification. "Filter" is another term used to describe a feature detector, and the two terms are used interchangeably. FIG. 7A shows an example of how an image is scanned by a particular feature detector 702. FIG. 7A shows a 7×5 pixel input image 701 scanned using a 3×3 filter window 702 with a stride of 1. In each window, individual numbers representing pixels of the image are multiplied by the number of corresponding positions in the 3×3 feature detector. The average of the products is calculated and substituted into a new feature map 703. The filter moves across the image with a stride of 1, repeating the same process in each window until the image has been completely scanned by the filter and a complete feature map has been created.

[0082] FIG. 7B shows the calculation of the first scan window 701-1 multiplied by the filter 702, which sums to an average of −0.11, which is assigned as the first value in the feature map 703. The output feature map 703 indicates whether a particular feature is detected or not. The above calculation is representative, and the values ​​−1 and 1 are used for simplicity, but values ​​represent pixels with a much larger range. The input image in FIG. 7A, FIG. 7B is either an input image or a feature map, as described below.

[0083] The size of the scan window, the filter, and the number of strides are all variables and hyperparameters that are tuned to train the model. The actual values ​​in the filters are learned by the model using backpropagation and cannot be specified. The filters represent features, and when an image is filtered by a filter, the image is being scanned for that particular feature. Through training, the network determines which features are important and creates a filter for each feature. Although Figures 7A and 7B show the convolution operation for one filter, this operation can be performed to generate any number of filters.

[0084] FIG. 8 embodies an example of applying multiple filters 802 to an input image 801 to generate multiple feature maps 803. More filters are then applied to the output feature maps and the combined feature maps. Generally, in a CNN, lower layers detect simple structures (e.g., different types of edges). While going deeper (i.e., toward layers closer to the output of the CNN), the layers build on each other and learn to encode more complex patterns. Putting this complex pattern together, the model can finally detect a pattern corresponding to erythema or other symptoms. FIG. 9 shows a simplified block diagram of a subsequent filter 94 that is applied to the combination of already filtered feature maps 93 (by filter 92). As shown in FIG. 9, a first layer of filters 92 applied to image 91 detects different types of edges in the image that indicate erythema, and a second filter is applied to the combination of output feature maps that indicate erythema. FIG. 9 simplifies the process of detecting erythema, but in reality there are many more layers of filters before diagnosing erythema. The earlier filters are more similar to the edges, shapes, and features of the erythema, and the more subsequent layers are less detailed in representing the erythema. The filters after the erythema diagnosis detect more features that characterize the erythema score. The model also includes a convolution process with the filters that finally diagnoses the scales and pustules and their condition / scaling scores, as well as a global score for GPP (or PPP according to the first aspect).

[0085] To increase the nonlinearity in the model, an adjustment function, or conditioned linear activation unit (ReLU), is applied to the feature map output from each filter. ReLU essentially replaces all negative values ​​in the feature map with a value of "0" and preserves all positive values. An example of a feature map 703 being conditioned is shown in FIG. 10. The feature map 703 initially contains positive and negative values, and after applying the adjustment function, the negative values ​​are replaced with zeros in the conditioned feature map 704. Images naturally contain nonlinear features such as transitions between pixels and boundaries, but convolution operations can impose linearity. The conditioned works to resolve the linearity and helps preserve the nonlinearity of the model.

[0086] The output from the conditioner is pooled to reduce the size / dimension of the feature map. Figure 10C is an example of reducing a 4x4 feature map IFM to a 2x2 feature map NFM by "max pooling". In this example, the original feature map is scanned using a 2x2 window SW with a stride of 2, and the maximum number from each window is converted to the new feature map. Other types of pooling include average pooling, where the average from the window is assigned to the new feature map, and sum pooling, where the sum of the numbers in the window is assigned to the new feature map. All pooling techniques effectively reduce the dimension of the feature map.

[0087] The feature map data is flattened from a 2D array to a 1D array so that it can later be inserted into an artificial neural network for classification. FIG. 10B is an example of a 3×3 2D array flattened to a 1D array. The 1D array is finally passed through an artificial neural network (ANN) for classification (see artificial neural network 640 in FIG. 6A). Convolution handles the feature extraction part of the network, and the ANN at the latter end of the model handles the variety in the inputs so that it can be classified correctly. Convolution detects features, and the ANN handles the variety in the location of those features. To further account for the variety, data augmentation may be applied to the initial input images of people's skin to generate more samples with more variety. Data augmentation includes rotating the images and changing the scale of the images, which helps train the model for variety.

[0088] In one implementation, an internal dataset of images from patients with GPP was augmented and used to train a deep neural network model. Data augmentation included rotating, zooming, and shrinking the images to generate variation in the input data and create more samples. In this implementation, an existing pre-trained model was used for transfer learning and trained with the GPP images. Possible embodiments of the pre-trained deep neural network model include, but are not limited to: - Convolutional Neural Networks (CNNs), such as the ResNet, EfficientNet, and ConvNeXt architectures; - Visual Integrator (VIT); and -Combinatorial networks that stack convolutional and attention layers.

[0089] In one embodiment of the deep neural network, the output is classified as a score of GPP. Another embodiment includes using the score to classify GPP, normal skin, and / or skin lesions that are neither GPP nor normal. This embodiment flags the image as containing skin areas that are not normal and potentially at risk for other infections and should be further analyzed.

[0090] During the convolution stage of the deep neural network, the model according to the second aspect determines which features represent GPPs, and embodiments of this layer include, but are not limited to, edges, colors, and shapes that represent GPPs, as well as pustule counts. These features are identified and used to train the model during training. These features are also determined during training using alternative algorithms mentioned above, including, but not limited to, convolutional neural network (CNN) architectures such as ResNet (residual neural network), EfficientNet architectures, and ConvNeXt architectures, vision transformers (VIT), and combinatorial networks that stack convolutional and attention layers. In some implementations, image processing may also include background removal, skin lesion localization (via insight-based designed features including color, texture, boundary irregularities, and asymmetries), and learned classifiers, such as U-Net architectures for semantic segmentation.

[0091] Returning to FIG. 1, in one embodiment, the classification of ND conditions described above using the deep neural network DNN1 trained on the training image data set 300 may be enhanced by further using tabulated clinical data 200 for training the classifier. When the predictor 110 is used in this embodiment to evaluate a test input of a patient 10 in addition to the image grid 21, the test input of the patient further includes the patient's tabulated clinical data 30. Thereby, a particular set of features 30 of the tabulated clinical data and the respective digital image 21 are associated with the same severity of the patient's condition. This is usually achieved by capturing such data on the same day. Of course, the same is true for the training data used to train the classifier. In this example, each training image 310-360 has such an associated tabulated clinical training data set 210-260, respectively.

[0092] Figures 6B and 6C show two alternative implementations 110-1, 110-2 of the "tabular clinical data" embodiment. In Figure 6B, the predictor 110-1 is configured to combine a feature set 30 of the tabular clinical data 30-1, 30-2, 30-3 of a patient with features EF1 to EFn extracted by the CNN from each digital image 21 of the patient for the severity assessment of said patient. That is to say, the input layer CFS of the classification neural network is extended compared to Figure 6A with the tabular clinical data features 30-1, 30-2, 30-3. Advantageously, the tabular clinical data features 30-1, 30-2, 30-3 are normalized by a normalizer N1 before being concatenated with the respective features EF1 to EFn extracted from the convolutional layer CNN of the deep neural network. The concatenated feature vector CFS is then processed by a fully connected layer FC representing the classification layer FCL of the deep neural network. A classification output CO is then provided by the final classification layer. In this embodiment, the training module trains a deep neural network DNN1 using respective augmented training data including a plurality of training images 310-360 and associated feature sets 210-260 of tabular clinical data features (see FIG. 1).

[0093] Examples of tabular clinical data for augmenting a patient's feature set may include, but are not limited to, fever, C-reactive protein (CRP), white blood cells (WBC), and albumin.

[0094] By combining such tabulated clinical data features with features extracted from each digital image, the classifier achieves higher prediction accuracy compared to using the image alone.

[0095] FIG. 6C shows an alternative implementation 110-2 for the tabular clinical data embodiment. The predictor 110-2 includes an image path (top of FIG. 6C) that essentially implements a deep neural network as described in FIG. 6A. The CNN is used to extract features EF1-EFn from the received grid image 21. A fully connected classification layer FC provides a first classification output CO1 (severity level determined by the trained deep neural network). In this implementation, the predictor 110-2 further includes a clinical data classifier CDC that is trained on the tabular clinical data features associated with each training image of the deep neural network model using the same ground truth. That is, if the ground truth of a training image specifies a particular severity score, the ground truth of the associated tabular clinical dataset is also set to this particular severity score. The training module in this implementation uses ensemble learning for the clinical data classifier CDC and the deep neural network of the image path. The predictor finally combines the first output CO1 of the deep neural network model and the second output CO2 of the clinical data classifier CDC into a single severity score CO. The CDC does not use a neural network but is based on a different classifier, such as, for example, logistic regression, gradient boosting, random forest, support vector machine, or other suitable classifier.

[0096] The above disclosed method and system are dedicated to the assessment of GPP and / or PPP severity of a patient's ND condition. However, those skilled in the art will understand that the approach disclosed herein for the assessment of a patient's ND condition can also be applied to other types of neutrophilic skin diseases when using appropriate training data to train the underlying deep neural network model. For example, such other types of neutrophilic skin diseases include hidradenitis suppurativa (HS); acute generalized exanthematous pustulosis; acute febrile neutrophilic skin disease (Sweet's syndrome); amicrobial pustulosis of the skin folds (APF); Behcet's disease; intestinal bypass syndrome (gut-associated dermatitis-arthritis syndrome); gut-associated dermatitis-arthritis syndrome (BADAS); CARD14-mediated pustular psoriasis (CAMPS); cryopyrin-associated periodic syndrome (CAPS); interleukin-36 receptor antagonist deficiency (DIRTA); interleukin-I receptor antagonist deficiency (DIRA); persistent erythema elevatum; histiocytic erythematosus. These may include neutrophilic dermatitis; acropustulosis pediatricus; neutrophilic dermatosis of the dorsum of the hands; neutrophilic eccrine hidradenitis; neutrophilic urticarial dermatosis; palisade neutrophilic granulomatous dermatitis; psoriasis vulgaris; pyoderma gangrenosum, acne, hidradenitis suppurativa (PASH) syndrome; pyoderma gangrenosum (PG); pyoderma gangrenosum and acne (PAPA); septic arthritis; the skin lesions of Behcet's disease; Still's disease; subcorneal pustulosis (Sneddon-Wilkerson disease); synovitis, acne, pustulosis-osteophytosis, and osteitis (SAPHO) syndrome; rheumatic neutrophilic dermatitis (RND), and ichthyosis (and its subtypes including Netherton syndrome (NS)).

[0097] Returning to FIG. 1, optional embodiments are now described that allow the use of the systems and methods disclosed herein for assessing the severity of neutrophilic skin diseases with visible skin manifestations in the context of providing appropriate treatment to patients suffering from ND. In these optional embodiments, the system may further include a severity checker module 120 and may further include an antibody dose module 130. The corresponding methods 3000, 4000 are shown in FIG. 3A, 3B, where the method 3000 is represented by a simplified flowchart for treating patients suffering from palmoplantar pustulosis and the method 4000 is represented by a simplified flowchart for treating patients suffering from generalized pustular psoriasis. The two methods 3000 and 4000 differ only in the first step. In the method 3000, the first step 3100 is to execute the computer-implemented method 1000 according to the first aspect to predict the severity score of the palmoplantar pustulosis condition of the subject. In method 4000, the first step 4100 is to perform the computer implemented method 2000 according to the second aspect to predict the severity score of the patient's generalized pustular psoriasis condition. The remaining steps 3200-3400 of method 3000 correspond to steps 4200-4400 of method 4000 (see FIG. 3B). Thus, the details of the method are described herein for method 4000 (GPP), but apply equally to method 3000 (PPP).

[0098] After execution of the computer-implemented method 2000 (step 4100), the system 100 has determined a severity score 111 of the generalized pustular psoriasis condition of the patient 10. The severity checker module 120 can compare the predicted severity score 111 with a predefined medication threshold 2 (step 4200). The predicted severity score has one of at least three severity score values ​​1, 2, 3, covering a severity range from none to severe (see FIG. 5). Note that the reference numbers 1, 2, 3 do not represent the amount of each value, but simply refer to the position of each value in the severity range. In this example, the severity checker defines a second severity score value 2 (dotted background) as the predefined medication threshold. According to the range 112b of FIG. 5, the amount of this value may correspond to the intermediate value "1". If a five-point scale is used as in the range 112a, the threshold may be set to "3" for moderate. In other words, when the predictor has determined the patient's GPP severity score, the severity checker checks whether the determined (predicted) severity score is greater than or equal to a predefined drug administration threshold 2 (step 4240). If not (no), the process ends (step 4280) as there is no need for treatment given the patient's current GPP status. If so (yes), the severity checker can assign the patient 10 as a candidate for treatment with an anti-interleukin-36 receptor antibody.

[0099] If the patient is a candidate for treatment, the system 100 can use the antibody dosage module 130 to determine (step 4300) a recommended dosage based on a pharmacologic effective amount of the anti-interleukin-36 receptor antibody. The pharmacologic effective amount is determined in a clinical study. Based on the patient's severity score and the drug administration threshold, the computer can now determine whether the patient should be treated with a respective dosage (e.g., additional dosage) and make a respective recommendation. If the severity score is equal to or greater than the predetermined drug administration threshold, the recommended dosage can be provided to the respective drug administration entity 400 to administer (step 4400) the recommended dosage of the anti-interleukin-36 receptor antibody to the patient for the treatment of generalized pustular psoriasis. The antibody dosage module can enable the user to interact with a physician. The recommended dosage may be initially communicated to the physician for approval, and the recommended dosage or correction dosage may be provided to the drug administration entity 400 only upon approval by the physician.

[0100] With respect to the method 3000 for treating a patient suffering from palmoplantar pustulosis, the determined severity score may be a Palmoplantar Pustulosis Global Assessment (PPPGA) score or a Palmoplantar Pustulosis Global Pustulosis (PPPGP) score. When using a 5-point scale implementation, the following severity scores may be applied: 0 if the total PPPGA score or PPPGP score is equal to 0, 1 if the total PPPGA score or PPPGP score is greater than 0 but less than 1.5, 2 if the total PPPGA score or PPPGP score is 1.5 or more but less than 2.5, 3 if the total PPPGA score or PPPGP score is 2.5 or more but less than 3.5, and 4 if the total PPPGA score or PPPGP score is 3.5 or more. If the patient shows a score of 2 or more for the total PPPGA score or PPPGP score, a recommended dose based on a pharmacologic effective amount of anti-IL-36R may be administered to the patient for PPP treatment.

[0101] The antibody according to one embodiment can be incorporated into a pharmaceutical composition suitable for administration to a patient. The compounds of the invention can be administered alone or in combination with pharma- ceutically acceptable carriers, diluents, and / or excipients, in single or multiple doses. The pharmaceutical composition for administration is designed to be appropriate for the selected mode of administration, and pharma- ceutically acceptable diluents, carriers, and / or excipients (e.g., dispersants, buffers, surfactants, preservatives, solubilizers, isotonicity agents, stabilizers, etc.) are used as appropriate. The composition is designed according to conventional techniques, e.g., Remington, The Science and Practice of Pharmacy, 19th Edition, Gennaro, Ed., Mack Publishing Co., Easton, PA 1995, which provides an overview of formulation techniques as generally known to those skilled in the art.

[0102] Pharmaceutical compositions comprising the anti-IL-36R monoclonal antibodies of the present invention can be administered to patients suffering from neutrophilic skin diseases as described herein using standard administration techniques, including oral, intravenous, intraperitoneal, subcutaneous, pulmonary, transdermal, intramuscular, intranasal, buccal, sublingual, or as a suppository.

[0103] The route of administration of the antibody of the present invention may be oral, parenteral, inhalation, or topical. Advantageously, the antibody of the present invention may be incorporated into a pharmaceutical composition suitable for parenteral administration. The term parenteral as used herein includes intravenous, intramuscular, subcutaneous, rectal, vaginal, or intraperitoneal administration. Peripheral systemic delivery by intravenous or intraperitoneal or subcutaneous injection is preferred. Vehicles suitable for such injection are known in the art.

[0104] Pharmaceutical compositions must typically be sterile and stable under the conditions of manufacture and storage in the containers provided (including, for example, sealed vials or syringes). Thus, pharmaceutical compositions may be sterile filtered after formulation is made or otherwise rendered microbiologically acceptable. A typical composition for intravenous infusion may have a volume of fluid on the order of 250-1000 mL, such as sterile Ringer's solution, saline, dextrose solution, and Hank's solution, and an antibody concentration of a therapeutically effective dose (e.g., 1-100 mg / mL or more). Doses may vary depending on the type and severity of the disease. As is well known in the medical art, the dose for any one patient depends on many factors, including the patient's size, body surface area, age, the particular compound administered, sex, time and route of administration, general health, and other drugs administered concomitantly. A typical dose may be, for example, in the range of 0.001-1000 mg. However, doses below or above this exemplary range are also envisioned, especially considering the aforementioned factors.

[0105] The system 100 disclosed herein can automatically assess the severity of skin lesions in ND patients with visible signs and symptoms such as skin erythema, pustules, and / or scales to assist treating physicians in quickly and easily forming an objective assessment of the subject's disease severity required for effective treatment of the patient. Treatment is advantageously applied when the patient has a severity score corresponding to at least a mild severity level. That is, in the example of FIG. 5, the drug administration threshold is set to "2" for the example 5-point scale 112a and "1" for the example 3-point scale 112b.

[0106] Neutrophilic dermatoses (NDs) are a heterogeneous group of conditions, but they share common features and overlapping pathophysiology involving the IL-36 pathway. Neutrophilic dermatoses are primarily associated with the development of cutaneous symptoms due to the accumulation of neutrophils, but IL-36 plays an important role in driving disease manifestations, especially in pustular NDs, such as generalized pustular psoriasis (GPP), palmoplantar pustulosis (PPP), subcorneal pustulosis (Sneddon-Wilkerson disease), pustular psoriasis, acute generalized exanthematous pustulosis (AGEP), acropustulosis of childhood (IA), Behçet's disease, pustulosis, osteophytosis and osteitis (SAPHO) syndrome, gut-associated dermatosis-arthritis syndrome (BADAS), neutrophilic dermatosis of the dorsum of the hands (NDDH), and amicrobial pustulosis of the subcutaneous fat (APF). Furthermore, blocking the IL-36 pathway may be beneficial in acute febrile neutrophilic dermatosis (Sweet's syndrome), rheumatic neutrophilic dermatitis (RND), neutrophilic eccrine hidradenitis (NEH), dermatitis with the formation of papules such as erythema elevatum elevans (EED), and / or skin ulcers, nodules, and plaques such as pyoderma gangrenosum (PG). Blocking IL-36 is also beneficial in patients with ichthyosis (and its subtypes, including Netherton syndrome (NS)).

[0107] Without wishing to be bound by this theory, the anti-IL-36R antibody binds to human IL-36R and thus interferes with the binding of IL-36 agonists, thereby at least partially blocking the signal transduction cascade from IL-36R to inflammatory mediators involved in neutrophilic skin diseases. The anti-IL 36R antibody of the present invention is disclosed herein, for example, in U.S. Patent No. 9,023,995, the entire contents of which are incorporated herein by reference.

[0108] In some embodiments, anti-IL-36R antibodies, particularly humanized anti-IL-36R antibodies, as well as compositions and articles of manufacture comprising one or more anti-IL-36R antibodies of the invention, particularly one or more humanized anti-IL-36R antibodies, are described and disclosed herein. Binding agents comprising antigen-binding fragments of anti-IL-36 antibodies, particularly humanized anti-IL-36R antibodies, are also described. In one embodiment, the anti-IL-36R is spesolimab (BI655130).

[0109] The approaches described herein can also be used in the context of modifying, discontinuing, or continuing treatment of an individual receiving an anti-IL-36R antibody for the treatment of ND. Based on the predicted severity score, the severity checker and antibody dose module can make recommendations to modify, discontinue, or continue the patient's treatment.

[0110] The approach described herein can also be used in the context of monitoring whether a patient who has received an anti-IL-36R antibody for the treatment of ND responds to the treatment. Based on the predicted severity score, the severity checker can characterize the subject as being responsive to treatment with an anti-IL-36R antibody if the subject shows a decrease in the predicted severity score compared to the previous predicted score.

[0111] Severity Scoring Scoring the severity of erythema, pustules and scale, and calculating the total Neutrophilic Dermatopathy Area Severity Index (NDASI) and total Neutrophilic Dermatopathy Global Assessment (NDGA) are further described in the following paragraphs in the context of GPP, but this is not intended to limit the scope of the present invention. The methods described below for GPPASI and GPPGA are equally applicable to score the severity in any ND condition (e.g., PP) that has skin manifestations including erythema, pustules and / or scale. Such scoring methods can be used, for example, to determine the correct answer for training images in which scoring is performed manually.

[0112] The components of the ND or GPP / PPP severity score are erythema, pustules, and scale.

[0113] For subsequent calculation of the Generalized Pustular Psoriasis Area Severity Index (GPPASI) and Generalized Pustular Psoriasis Physician Global Assessment (GPPGA), the severity of each component is typically scored on a 5-point scale: 0=none, 1=almost none, 2=mild, 3=moderate, 4=severe. Other scales are possible (see FIG. 5).

[0114] For GPPASI, typically each component is scored separately for each body area. For GPPGA, typically each component is scored separately for each lesion or for all lesions. If one lesion is identified and scored, calculate the sum of the severity scores of erythema, pustules, and scale to calculate the total GPPGA, then divide the result by 3. If two or more lesions are scored to calculate the total GPPGA, calculate the sum of the mean, median, or maximum severity scores of erythema, pustules, and scale, then divide the result by 3.

[0115] For example, for patients with light-skinned skin (e.g., Caucasian, Asian), an erythema severity score of 0 indicates a skin lesion or area with normal or post-inflammatory hyperpigmentation, an erythema severity score of 1 indicates a skin lesion or area with a faint, diffuse pink or slightly red color, an erythema severity score of 2 indicates a skin lesion or area with a light red color, an erythema severity score of 3 indicates a skin lesion or area with a bright red color, and an erythema severity score of 4 indicates a skin lesion or area with a dark red color (see FIG. 11A). Furthermore, a pustule severity score of 0 indicates a skin lesion or area with no visible pustules, a pustule severity score of 1 indicates a skin lesion or area with occasional small, individual pustules (not coalescing) with low density, a pustule severity score of 2 indicates a skin lesion or area with medium density grouped small, individual pustules (not coalescing), a pustule severity score of 3 indicates a skin lesion or area with dense pustules with some coalescence, and a pustule severity score of 4 indicates a skin lesion or area with very dense pustules with a pustular crimson color (see FIG. 11B). Furthermore, a scaling severity score of 0 indicates a skin lesion or area that is devoid of scaling or crusting, a scaling severity score of 1 indicates a skin lesion or area with superficial, localized scaling or crusting restricted to the periphery of the lesion, a scaling severity score of 2 indicates a skin lesion or area with primarily fine scaling or crusting, a scaling severity score of 3 indicates a skin lesion or area with moderate scaling or crusting covering most or all of the lesion, and a scaling severity score of 4 indicates a skin lesion or area with severe scaling or crusting covering most or all of the lesion (see FIG. 11C).

[0116] As mentioned above, the respective severity of erythema, pustules, and / or scale on a skin area or lesion of a subject with ND can be scored by computer vision using a suitable AI (artificial intelligence) algorithm trained accordingly. Examples of such algorithms are described above. For example, a computer-implemented method according to the present invention scores digital images with ND lesions for severity of erythema, pustules, and / or scale by classifying the digital images according to a training image dataset. The training image dataset includes digital images of skin lesions that have been previously annotated or scored for severity of erythema, pustules, and / or scale by a specialist dermatologist. In another example, scoring a digital image having ND lesion(s) for severity of erythema, pustules, and / or scaling can be accomplished by analyzing the level, intensity, and / or extent of skin redness (in the case of erythema), or by counting pustules, measuring their size, analyzing their color, and / or assessing how diffuse or dense they appear in a given area of ​​skin (in the case of pustules), or by outlining the edges of the scales and / or measuring their level in terms of the degree of fineness or crusting (in the case of scales).

[0117] Calculation of GPPASI Body area factor: head = 0.1x, upper limbs = 0.2x, trunk = 0.3x, lower limbs = 0.4x.

[0118] Body region area score: 0 = 0% (no involvement), 1 = greater than 0 and less than 10% involvement, 2 = 10-30%; 3 = 30-50%; 4 = 50-70%; 5 = 70-90%; 6 = 90-100%.

[0119] Individual score per body domain = body domain coefficient × body domain area score * × sum of component severity scores for a body area (*body area area score is the area affected by erythema and / or pustules and / or scale; each component is not assessed separately).

[0120] Total GPPASI score = sum of individual scores from all body regions.

[0121] Calculating GPPGA Individual component scores: As shown in Figures 11A-11C, the severity of each component (erythema, pustules, and scale) is graded individually using a severity scale of 0 to 4 (5-point severity scale), where 0 is absent, 1 is almost absent, 2 is mild, 3 is moderate, and 4 is severe.

[0122] Composite average score: Calculate the average of the individual component scores:

[0123]

number

[0124] Total GPPGA Score: 0 = Mean = 0 for all three elements 1 = if the mean is greater than 0 and less than 1.5 2 = if the average is greater than or equal to 1.5 and less than 2.5 3 = if the average is greater than or equal to 2.5 and less than 3.5 4 = if the average is 3.5 or higher

[0125] The present invention is further described in the following examples, which are not intended to limit the scope of the invention.

[0126] FIG. 12 illustrates an example of a general-purpose computing device 900 and a general-purpose mobile computing device 950 that may be used with the techniques described herein. The computing device 900 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. The general-purpose computing device 900 may correspond to the computer system 100 of FIG. 1. The computing device 950 is intended to represent various forms of mobile devices, such as personal digital assistants, mobile phones, smartphones, and other similar computing devices. For example, the computing device 950 may be used as a GUI front end for a user to interact with the computing device 900, for example, to receive predicted severity score values ​​and recommended dosage information from the computing device 900, and / or to input certain data into the computing device, such as, for example, a correction dosage to override the recommended dosage. The components, their connections and relationships, and their functions shown herein are merely exemplary and are not intended to limit the implementation of the invention described and / or claimed herein.

[0127] The computing device 900 includes a processor 902 (e.g., CPU, GPU), a memory 904, a storage device 906, a high-speed interface 908 that connects to the memory 904 and a high-speed expansion port 910, and a low-speed interface 912 that connects to a low-speed bus 914 and the storage device 906. Each of the components 902, 904, 906, 908, 910, and 912 are interconnected using various buses and may be mounted on a common motherboard or otherwise as needed. The processor 902 can process instructions for execution within the computing device 900, including instructions stored in the memory 904 or the storage device 906, to display visual information for a GUI on an external input / output device, such as a display 916 coupled to the high-speed interface 908. In other implementations, multiple processing units and / or multiple buses may be used, along with multiple memories and multiple types of memories, as needed. Multiple computing devices 900 may also be connected, with each device providing a portion of the required operations (e.g., as a server bank, a group of blade servers, or a processing device).

[0128] The memory 904 stores information within the computing device 900. In one implementation, the memory 904 is one or more volatile memory units. In another implementation, the memory 904 is one or more non-volatile memory units. The memory 904 may be another form of computer-readable medium, such as a magnetic disk or an optical disk.

[0129] The storage device 906 can provide mass storage for the computing device 900. In one implementation, the storage device 906 can be or include a computer-readable medium such as a floppy disk device, a hard disk device, an optical disk device, or a tape device, a flash memory or other similar solid-state memory device, or an array of devices including devices in a storage area network or other configuration. The computer program product can be tangibly embodied in an information carrier. The computer program product can also include instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer-readable or machine-readable medium, such as the memory 904, the storage device 906, or a memory on the processor 902.

[0130] The high-speed controller 908 manages bandwidth-intensive operations for the computing device 900, while the low-speed controller 912 manages lower bandwidth-intensive operations. Such an allocation of functions is merely exemplary. In one implementation, the high-speed controller 908 is coupled to the memory 904, a display 916 (e.g., via a graphics processor or accelerator), and a high-speed expansion port 910 that can accept various expansion cards (not shown). In this implementation, the low-speed controller 912 is coupled to the storage device 906 and the low-speed expansion port 914. The low-speed expansion port, which may include various communication ports (e.g., USB, Bluetooth, Ethernet, wireless Ethernet), may be coupled to one or more input / output devices, such as a keyboard, pointing device, scanner, or a networking device, such as a switch or router, for example, via a network adapter.

[0131] Computing device 900 may be implemented in a number of different forms, as shown. For example, it may be implemented as a standard server 920, or multiple times in a cluster of such servers. It may also be implemented as part of a rack server system 924. It may also be implemented in a personal computer, such as a laptop computer 922. Alternatively, components from computing device 900 may be combined with other components in a mobile device (not shown), such as device 950. Each such device may include one or more of computing devices 900, 950, and the entire system may be composed of multiple computing devices 900, 950 in communication with each other.

[0132] Computing device 950 includes, among other components, a processor 952, memory 964, input / output devices such as a display 954, a communication interface 966, and a transceiver 968. Device 950 may also include a storage device, such as a microdrive or other device, to provide additional storage. Each of the components 950, 952, 964, 954, 966, and 968 are interconnected using various buses, and some of the components may be mounted on a common motherboard or in other manners as desired.

[0133] The processor 952 may execute instructions within the computing device 950, including instructions stored in the memory 964. The processor may be implemented as a chipset of chips including separate analog and digital processing units. The processor may coordinate other components of the device 950, such as, for example, the user interface, applications run by the device 950, and control of wireless communications by the device 950.

[0134] The processor 952 may communicate with a user via a control interface 958 and a display interface 956 coupled to a display 954. The display 954 may be, for example, a Thin-Film-Transistor Liquid Crystal Display (TFT LCD) or an Organic Light Emitting Diode (OLED) display, or other suitable display technology. The display interface 956 may include appropriate circuitry for driving the display 954 to present visual and other information to the user. The control interface 958 may receive commands from the user and convert them for delivery to the processor 952. Additionally, an external interface 962 may be provided in communication with the processor 952 to enable short-range communication of the device 950 with other devices. The external interface 962 may, for example, be provided for wired communication in some implementations and wireless communication in other implementations, and multiple interfaces may be used.

[0135] The memory 964 stores information within the computing device 950. The memory 964 may be implemented as one or more of one or more computer readable media, one or more volatile memory units, or one or more non-volatile memory units. An expansion memory 984 may also be provided and connected to the device 950 via an expansion interface 982, which may include, for example, a Single In Line Memory Module (SIMM) card interface. Such expansion memory 984 may provide additional storage space for the device 950 or may also store applications or other information for the device 950. In particular, the expansion memory 984 may include instructions for performing or supplementing the processes described above, and may also include secure information. Thus, for example, the expansion memory 984 may function as a security module for the device 950 and may be programmed with instructions that enable secure use of the device 950. Furthermore, secure applications may be provided via a SIMM card along with additional information, such as placing identifying information on the SIMM card in an unhackable manner.

[0136] The memory may include, for example, flash memory and / or NVRAM memory, as described below. In one implementation, a computer program product is tangibly embodied in an information carrier. The computer program product includes instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer-readable or machine-readable medium, such as memory 964, expansion memory 984, or memory on processor 952, and may be received, for example, via transceiver 968 or external interface 962.

[0137] Device 950 may communicate wirelessly via communication interface 966, which may include digital signal processing circuitry, as needed. Communication interface 966 may provide communications under various modes or protocols, such as GSM voice calls, SMS, EMS, or MMS messaging, CDMA, TDMA, PDC, WCDMA, CDMA2000, or GPRS, among others. Such communications may occur, for example, via radio frequency transceiver 968. In addition, short-range communications may occur using, for example, Bluetooth, WiFi, or other such transceivers (not shown). In addition, a Global Positioning System (GPS) receiver module 980 may provide additional navigation- and location-related wireless data to device 950, which may be used as appropriate by applications executing on device 950.

[0138] Device 950 may also communicate audibly using audio codec 960, which may receive spoken information from a user and convert it into usable digital information. Audio codec 960 may also generate audible sounds for the user, such as through a speaker in a handset of device 950. Such sounds may include sounds from a voice call, may include recorded sounds (e.g., voice messages, music files, etc.), and may include sounds generated by applications running on device 950.

[0139] The computing device 950 may be implemented in a number of different forms, as shown in the figure. For example, it may be implemented as a mobile phone 980. It may also be implemented as part of a smartphone 982, personal digital assistant, or other similar mobile device.

[0140] Various implementations of the systems and techniques described herein may be realized in digital electronic circuitry, integrated circuits, specially designed application specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs executable and / or interpretable on a programmable system including at least one programmable processor, which may be special purpose or general purpose, coupled to receive data and instructions from and transmit data and instructions to a storage system, at least one input device, and at least one output device.

[0141] These computer programs (also known as programs, software, software applications or codes) contain machine instructions for a programmable processor and may be implemented in high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus and / or device used to provide machine instructions and / or data to a programmable processor (e.g., magnetic disks, optical disks, memories, Programmable Logic Devices (PLDs)), including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0142] To provide for interaction with a user, the systems and techniques described herein may be implemented on a computer having a display device (e.g., a cathode ray tube (CRT) or liquid crystal display (LCD) monitor) for displaying information to a user, and a keyboard and pointing device (e.g., a mouse or trackball) by which the user can provide input to the computer. Other types of devices may be used to provide interaction with a user. For example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback), and input from the user may be received in any form, including acoustic, speech, or tactile input.

[0143] The systems and techniques described herein can be implemented in a computing device that includes a back-end component (e.g., as a data server), or includes a middleware component (e.g., an application server), or includes a front-end component (e.g., a client computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or includes any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0144] Computing devices may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

[0145] Several embodiments have been described. Nevertheless, it will be understood that various modifications can be made without departing from the spirit and scope of the invention.

[0146] Additionally, the logic flows depicted in the figures do not require the particular order shown, or sequential order, to achieve desired results. Additionally, other steps may be provided or steps may be eliminated from the described flows, and other components may be added to or removed from the described systems. Accordingly, other embodiments are within the scope of the following claims.

Claims

1. 1. A computer-implemented method (1000) for predicting the severity of a palmoplantar pustulosis condition in a patient using a deep neural network model, comprising: receiving (1100) a test input digital image (21, 22) representing a skin area of ​​the patient (10), the test input digital image comprising a grid having at least one pair of tiles (21-1, 21-2, 22-1, 22-2), the first tile (21-1, 22-1) representing a skin area on a front (22-f) of a characteristic body part of the patient, and the second tile (21-2, 22-2) representing a skin area on a back (22-b) of the characteristic body part of the patient, the characteristic body part being selected from the palms of the left and right hands, and the soles of the left and right feet, and at least one of the skin areas representing at least one of erythema, pustules, and scales on the skin; predicting (1200) a Palmoplantar Pustulosis Global Assessment (PPPGA) score for the patient, wherein the predicting comprises applying the deep neural network model (DNN1) to the test input (21), the deep neural network model being trained with a training dataset including a plurality of training images having the same structure as the test input (21) and captured from a plurality of test patients (90) selected according to predetermined inclusion / exclusion criteria that establish the training patients as having a history of generalized pustular psoriasis without any opposing disease, the training dataset including a plurality of training images (310-360) for each test patient captured at different time points during a predetermined minimum time interval, each training image annotated with one or more severity scores associated with erythema, pustules, and scaling as ground truths reflecting the severity of the disease state of the respective training patient at the time the training image was captured; A method comprising:

2. 1. A computer-implemented method (2000) for predicting the severity of generalized pustular psoriasis condition in a patient (10) using a deep neural network model (DNN1), comprising: receiving (2100) a test input digital image (21, 22) showing a skin area of ​​the patient (10), the test input digital image comprising a grid having at least one pair of tiles (21-1, 21-2, 22-1, 22-2), the first tile (21-1, 22-1) showing a skin area on a front (22-f) of a distinctive body part of the patient (10), the second tile (21-2, 22-2) showing a skin area on a back (22-b) of the distinctive body part of the patient, the distinctive body part being selected from the trunk, the left and right lower limbs, and the left and right upper limbs, and at least one of the skin areas showing at least one of erythema, pustules, and scale on the skin; predicting (2200) a Generalized Pustular Psoriasis Physician Global Assessment (GPPGA) score for the patient (10), wherein the predicting comprises applying the deep neural network model (DNN1) to the test input (21), the deep neural network model being trained on a training dataset (300) including a plurality of training images having the same structure as the test input (21) and captured from a plurality of test patients (90) selected according to predetermined inclusion / exclusion criteria that establish the training patients as having a history of generalized pustular psoriasis without any contradictory disease, the training dataset including a plurality of training images (310-360) for each test patient captured at different time points during a predetermined minimum time interval, each training image annotated with one or more severity scores associated with erythema, pustules, and scaling as ground truths reflecting the severity of the generalized pustular psoriasis condition of the respective training patient (90) at the time the training image was captured; A method comprising:

3. 2. The method of claim 1, wherein each training image is annotated with one of at least three severity score values ​​covering a severity range (112) from none to severe, the annotated severity score values ​​reflecting an average of individual erythema severity scores, pustule severity scores, and scaling severity scores for each training image, and the trained deep neural network model provides a single severity score value as output for a test input of the patient.

4. 2. The method of claim 1, wherein each training image is annotated with an individual erythema severity score value, a pustule severity score value, and a scaling severity score value, each individual severity score value being one of at least three severity score values ​​covering a severity range (112, 112b) from none to severe for the respective erythema severity, pustule severity, and scaling severity on the training image, the trained deep neural network model providing as output the individual severity score value for each of erythema severity, pustule severity, and scaling severity, and a single severity score for the test input for the patient is determined based on averaging the determined individual severity score values.

5. 4. The method of claim 3, wherein the severity ranges (112, 112b) include the following score values ​​for severity levels: none, almost none, mild, moderate, and severe.

6. 2. The method of claim 1, wherein for the severity prediction of the patient, a set of features (30) of tabulated clinical data of the patient (10) is combined with features extracted from each digital image (21) of the patient, and a particular set of features of tabulated clinical data and each digital image is associated with the same severity of the patient's condition.

7. 7. The method of claim 6, wherein the tabulated clinical data features (30) are normalized and concatenated with the respective features (EF1 to EFn) extracted from the convolutional layer (CNN) of the deep neural network, and the concatenated feature set (CFS) is used as an input layer of the classification layer (FCL) of the deep neural network model (DNN1), and the deep neural network has been trained with respective augmented training data including the plurality of training images (310 to 360) and associated feature sets (210 to 260) of tabulated clinical data features.

8. 7. The method of claim 6, wherein a clinical data classifier (CDC) is trained on tabular clinical data features associated with each training image of the deep neural network model using the same ground truth, the training using ensemble learning for the clinical data classifier (CDC) and the deep neural network (DNN1), and the method further comprises combining the output (CO1) of the deep neural network model and the output (CO2) of the clinical data classifier into a single severity score (CO).

9. 2. The method of claim 1, wherein the deep neural network model implements one of the following algorithms: a convolutional neural network selected from the ResNet architecture, the EfficientNet architecture, or the ConvNeXT architecture; a Vision Transformer; and a combination network having a convolutional layer and an attention layer.

10. 10. A computer program product for predicting the severity of a palmoplantar pustulosis condition or a generalized pustular psoriasis condition in a patient, the computer program product, when loaded into a memory of a computing device and executed by at least one processor of the computing device, causing the at least one processor to perform the steps of the computer-implemented method of claim 1.

11. 1. A computer system (100) for predicting the severity of a palmoplantar pustulosis condition and / or a generalized pustular psoriasis condition in a patient (10) using one or more respectively trained deep neural network models (DNN1), comprising: an interface configured to receive a test input digital image (21, 22) of a skin area of ​​the patient (10), the test input digital image comprising a grid having at least one pair of tiles (21-1, 21-2, 22-1, 22-2), a first tile (21-1, 22-1) representing a skin area on the front (22-f) of a characteristic body part of the patient (10) and a second tile (21-2, 22-2) representing a skin area on the back (22-b) of a characteristic body part of the patient, at least one of the skin areas exhibiting at least one of erythema, pustules, and scales on the skin, if the patient is suffering from palmoplantar pustulosis, the characteristic body parts selected from the left and right palms and the left and right soles of the feet, and if the patient is suffering from generalized pustular psoriasis, the characteristic body parts selected from the trunk, the left and right lower legs, and the left and right upper legs; and a predictor module (110) configured to predict a Palmoplantar Pustulosis Global Assessment Score and / or a Generalized Pustular Psoriasis Physician Global Assessment Score (111) for the patient (10), wherein the predictor applies the one or more respective trained deep neural network models (DNN1) to the test input (21), and determines whether the one or more deep neural network models have the same structure as the test input (21) and whether a plurality of training patients (90) meet predetermined inclusion / exclusion criteria that determine that the training patients (90) each have a history of palmoplantar pustulosis and / or a history of generalized pustular psoriasis without any opposing disease. and training using a training dataset (300) comprising a plurality of training images captured from the plurality of test patients (90) selected accordingly, the training dataset comprising a plurality of training images (310-360) for each test patient captured at different times during a predetermined minimum time interval, each training image annotated with one or more severity scores associated with erythema, pustules, and scaling as ground truths that respectively reflect the severity of the palmoplantar pustulosis condition and / or generalized palmoplantar pustulosis condition of the respective training patient at the time the training image was captured. Computer system.

12. 12. The system of claim 11, wherein the one or more deep neural network models implement any of the following algorithms: a convolutional neural network selected from ResNet, EfficientNet, or ConvNeXT architectures; a Vision Transformer; and a combinatorial network having convolutional and attention layers.

13. 12. The system of claim 11, further comprising a severity checker module (120) configured to compare the predicted severity score with a predetermined medication threshold, the predicted severity score having one of at least three severity score values ​​covering a severity range (112) from none to severe, and further configured to assign the patient as a candidate for treatment with an anti-interleukin-36 receptor antibody if the predicted severity score is equal to or greater than the predetermined medication threshold.

14. 14. The system of claim 13, further comprising an antibody dosing module (130) configured to determine, for the candidate treatment, a recommended pharmaceutically effective dose of the anti-interleukin-36 receptor antibody suitable for treating the candidate's palmoplantar pustulosis condition and / or the generalized pustular psoriasis condition based on the predicted severity score, and provide corresponding dosing instructions (131) to a drug administration entity (400).

15. 12. The system of claim 11, wherein the predictor module is configured to combine a feature set of the patient's tabulated clinical data with features extracted from each digital image of the patient for severity prediction of the patient, wherein a particular feature set of the tabulated clinical data and each digital image is associated with the same severity of the patient's condition.

16. 16. The system of claim 15, wherein the predictor (110) is further configured to normalize the tabulated clinical data features (30) and concatenate the normalized features to the respective features (EF1 to EFn) extracted from the convolutional layer (CNN) of the deep neural network, and the concatenated feature set (CFS) is used as an input layer of the classification layer (FCL) of the deep neural network model (DNN1), and the deep neural network has been trained with respective augmented training data including the plurality of training images (310 to 360) and associated feature sets (210 to 260) of tabulated clinical data features.

17. 16. The system of claim 15, wherein the predictor (110) further comprises a clinical data classifier (CDC) trained on tabulated clinical data features associated with each of the training images of the deep neural network model using the same ground truth, the training using ensemble learning for the clinical data classifier (CDC) and the deep neural network (DNN1), and the predictor is further configured to combine the output (CO1) of the deep neural network model and the output (CO2) of the clinical data classifier into a single severity score (CO).

18. The anti-interleukin-36 receptor antibody is I. a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 102 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region (H-CDR1) comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or II. a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 103 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3), or III. a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 104 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or IV. a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 105 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or V. a) a light chain variable region comprising the amino acid sequence of SEQ ID NO:26 (L-CDR1), the amino acid sequence of SEQ ID NO:106 (L-CDR2), and the amino acid sequence of SEQ ID NO:44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO:53 (H-CDR1), the amino acid sequence of SEQ ID NO:62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO:72 (H-CDR3); or VI. a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 140 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 53 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, or 111 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3); or VII. a) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 26 (L-CDR1), the amino acid sequence of SEQ ID NO: 104 (L-CDR2), and the amino acid sequence of SEQ ID NO: 44 (L-CDR3); and b) a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 141 (H-CDR1), the amino acid sequence of SEQ ID NO: 62, 108, 109, 110, 111, or 142 (H-CDR2), and the amino acid sequence of SEQ ID NO: 72 (H-CDR3).

18. The method or system of any one of claims 1 to 17, comprising:

19. The anti-interleukin-36 receptor antibody is (i) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 77, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 87; or (ii) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 77, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 88; or (iii) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 77, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 89; or (iv) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 80, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 87; or (v) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 80, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 88; or (vi) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 80, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 89; or (vii) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 85, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 100; or (viii) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 85, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 101; or (ix) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 86, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO: 100; or (x) a light chain variable region comprising the amino acid sequence of SEQ ID NO: 86, and a heavy chain variable region comprising the amino acid sequence of SEQ ID NO:

101.

20. The anti-interleukin-36 receptor antibody is i. a light chain comprising the amino acid sequence of SEQ ID NO: 115 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 125; or ii. a light chain comprising the amino acid sequence of SEQ ID NO: 115 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 126; or iii. a light chain comprising the amino acid sequence of SEQ ID NO: 115 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 127; or iv. a light chain comprising the amino acid sequence of SEQ ID NO: 118 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 125; or v. a light chain comprising the amino acid sequence of SEQ ID NO: 118 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 126; or vi. a light chain comprising the amino acid sequence of SEQ ID NO: 118 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 127; or vii. a light chain comprising the amino acid sequence of SEQ ID NO: 123 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 138; or viii. a light chain comprising the amino acid sequence of SEQ ID NO: 123 and a heavy chain comprising the amino acid sequence of SEQ ID NO: 139; or ix. a light chain comprising the amino acid sequence of SEQ ID NO: 124, and a heavy chain comprising the amino acid sequence of SEQ ID NO:

138.

18. The method or system of any one of claims 1 to 17, comprising:

21. The method or system according to claim 13 or 14, wherein the anti-interleukin-36 receptor is spesolimab (BI655130).

22. The method or system of claim 13 or 14, wherein the anti-interleukin-36 receptor antibody is in the range of about 0.001 to about 1200 mg.

Images